CN118401177A - Diagnosis support program - Google Patents

Abstract

A diagnosis support program capable of displaying the activity of an organ is provided. A diagnosis support program for analyzing an image of a human organ and displaying the analysis result, characterized by causing a computer to execute the following processing: acquiring a plurality of frame images; calculating the frequency for representing the state of the organ based on the frames of images; calculating a phase difference between a waveform of an organ model acquired in advance and a waveform corresponding to the calculated frequency; and outputting a signal representing the phase difference.

Description

Diagnosis support program

Technical Field

The present invention relates to a diagnosis support program for analyzing images of organs of a human and displaying analysis results.

Background

In recent years, heart disease mortality has increased, and the need for useful and simple diagnostic techniques has increased. MRI diagnostic techniques have rapidly advanced and various cardiac examinations can be performed in a short time, so that the importance of MRI is increasing in image diagnosis of cardiac regions. In the "cardiac MRI", examination was performed by the methods of "cine MRI", "perfusion", "delayed radiography", and "black blood (BB: blackBlood)". In particular, cine magnetic resonance imaging has a feature that it has no limitation in the observation range, can observe an arbitrary cross section, and is excellent in reproducibility, compared with ultrasonic examination, SPECT examination, or the like. Therefore, imaging is generally performed in a plurality of medical facilities. In cine magnetic resonance imaging, the entire left ventricle is subjected to data collection of "about 10 slices/20 phases", for example using an electrocardiogram synchrony. Recently, using the "steady state (STEADY STATE) method", a high contrast between blood and myocardium can be obtained. In addition, in the "cardiac function analysis", there is an increasing demand for obtaining a correct value from the cardiac function evaluation by MRI, compared with CT, LVG (left ventricular imaging), or SPECT.

As such, cardiac MRI is clinically useful, particularly cine magnetic resonance imaging, which, although it can be imaged in a number of medical facilities, uses software to analyze images is rarely used. The reason for this is that software for conventional cardiac function analysis requires a complicated operation for extracting or correcting contours of the inner and outer membrane sides of the myocardium. In addition, since contour drawing on the inner and outer membrane side of the myocardium is easily affected by the individual operator, reproducibility of analysis results is also a problem. In addition, in a case where a medical facility using MRI apparatuses provided by a plurality of manufacturers is increased in popularity of cardiac MRI, there is a problem that data processing is not easy because serial numbers of respective companies are different.

In order to solve such a problem, a software is provided that reduces the labor for drawing a complicated myocardial contour, improves the accuracy thereof, and can reduce the correction work by automatically performing interpolation processing even if unintentional drawing is assumed. The software enables less stressed image viewing by displaying images side by side and performing a flick operation in an MRI cardiac functional analysis viewer.

Prior art literature

Non-patent literature

Non-patent document 1: https:// www.zio.co.jp/ziostation +.2-

Disclosure of Invention

Problems to be solved by the invention

However, as in the technique described in non-patent document 1, a doctor cannot easily grasp the disease simply by displaying MRI images side by side. Therefore, it is preferable to display an image according to the state of the heart. In other words, it is preferable to grasp the heart of the human body as the subject, and display an image representing actual motion based on the waveform or frequency of the heart or the tendency of change of the image.

In addition, the phase shift of the active period of an abnormal organ is known with respect to the activity of a normal organ, but based on this fact, a technique of displaying an image of an organ is not put into practical use.

The present invention has been made in view of such circumstances, and an object thereof is to provide a diagnosis support program capable of displaying an organ activity. More specifically, it is an object to quantify the form of the wave obtained and the matching rate, phase shift, and other mismatch rate with respect to Hz, to calculate the numerical values to be diagnostic assistance, and to generate an image to be diagnostic assistance by imaging these numerical values, with respect to new target data to be measured.

Means for solving the problems

(1) In order to achieve the above object, the present application adopts the following means. In other words, a diagnosis support program according to one aspect of the present application is a diagnosis support program for analyzing an image of a human organ and displaying the analysis result, and is characterized by causing a computer to execute: acquiring a plurality of frame images; calculating the frequency for representing the state of the organ based on the frames of images; calculating a phase difference between a waveform of an organ model acquired in advance and a waveform corresponding to the calculated frequency; and outputting a signal representing the phase difference.

(2) In the diagnosis support program according to one aspect of the present invention, the organ model divides an image of an organ and determines the image by an average value of pixel values in each divided region.

(3) In addition, a diagnosis support program according to an aspect of the present invention is characterized in that an image of the organ is segmented using a Voronoi (Voronoi) segmentation method.

(4) In addition, in one aspect of the present invention, in the case where the organ is a lung, the lung field region is divided according to a rate of change in the volume of the lung.

(5) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, the diagnosis support program causing a computer to execute: acquiring a plurality of frame images; extracting pixel values of specific points contained in the organ image; calculating a time-dependent change in the extracted pixel values; extracting at least one frequency signal from the temporal variation of the pixel values; and outputting the extracted frequency signal as the signal of the specific point.

(6) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, the diagnosis support program causing a computer to execute: under a plurality of photographing conditions, obtaining images of an organ photographed from a plurality of directions; extracting pixel values of specific points contained in an image of the organ; calculating a time-dependent change in the extracted pixel values; extracting at least one frequency signal from the temporal variation of the pixel values; outputting the extracted frequency signal as the signal of the specific point; using the output signals at specific points, a cross-sectional view of all or a portion of the organ is created.

(7) In addition, a diagnosis support program according to an aspect of the present invention is characterized in that an image of the organ is divided into a predetermined region including the specific point by using a (Voronoi) division method.

(8) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, the diagnosis support program causing a computer to execute: obtaining a signal output from the diagnosis support program according to any one of (1) to (7); for the acquired signals, causing AI (Artificial Intelligence) to learn a signal representing a normal organ or a signal representing an abnormal organ; and recording the learning result according to the AI.

(9) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: acquiring a plurality of frame images; and determining an organ from the acquired frame image, comparing the determined organ with the recorded learning result, and outputting the ratio of abnormal values.

(10) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: outputting a difference in phase between a waveform representing the periodic motion of the normal organ and a waveform representing the periodic motion of the abnormal organ.

(11) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: in each of the plurality of frame images, a signal of a portion having a different transmittance is added to a signal of another portion, or a signal of a portion having a different transmittance is subtracted from a signal of another portion.

(12) In addition, an aspect of the present invention is a diagnosis support program including: transforming the waveform of the periodic signal into a trigonometric function; and outputting a signal representing the waveform converted into the trigonometric function.

(13) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, the diagnosis support program causing a computer to execute: acquiring a plurality of frame images captured by periodic signals applied to a human body; performing Fourier transform on the change of the pixel value of the specific area of each frame of image; extracting a spectrum within a certain frequency band from the spectrum obtained after the fourier transform, the spectrum within the certain frequency band including a spectrum corresponding to a frequency of a signal applied to a human body; performing inverse Fourier transform on the extracted spectrum in the certain frequency band; and displaying each image after the inverse Fourier transform on a display.

(14) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying the analysis result, the diagnosis support program causing a computer to execute: acquiring a plurality of frame images captured by periodic signals applied to a human body; calculating the change rate of the pixel value of the specific area of each frame of image; extracting only a region where a tuning rate is within a predetermined range, using the tuning rate which is a ratio between a rate of change of pixel values of the specific region and a rate of change of periodicity of a signal applied to a human body; each image including the extracted region is displayed on a display.

(15) A diagnosis support program according to an aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, the diagnosis support program causing a computer to execute: acquiring a plurality of frame images; calculating a function representing a frequency and a waveform of pulmonary blood flow based on the respective frame images; calculating a function of a waveform representing pulmonary blood pressure based on the function of the frequency and waveform representing the calculated pulmonary blood flow and information representing the thickness of pulmonary blood vessels; a pulmonary blood pressure is estimated from the calculated waveform of pulmonary blood pressure.

(16) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: acquiring a plurality of frame images; and determining an organ and a blood flow flowing in the organ from the acquired frame image, comparing the determined organ and blood flow with the recorded learning result, and outputting the characteristic quantity of the blood flow in the organ.

(17) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: the feature quantity of the blood flow in the main blood vessel of the lung or the blood capillary of the lung and the blood flow in the peripheral lung vessel is outputted.

(18) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: the activities of the lung respiration and the activities of the lung blood flow are compared, and feature quantities representing the linkage of the two are output.

(19) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: acquiring a plurality of frame images; and determining a lung field region from the acquired frame image, comparing a wave representing the motion of the determined lung field region with a reference wave, and outputting a feature quantity representing the linkage of the wave.

(20) In the diagnosis support program according to one aspect of the present invention, the reference wave is a wave representing a respiratory cycle.

(21) Further, a diagnosis support program according to an aspect of the present invention is characterized by further comprising the following processing: calculating a maximum value of the pixel value; the waveform is obtained based on the signal after the calculated maximum.

(22) In addition, a diagnosis support program according to an aspect of the present invention is characterized in that data having periodicity is input, fourier transform processing is performed, and filtering processing for extracting a specific frequency is performed, thereby performing inverse fourier transform processing.

(23) Further, a diagnosis support program according to an aspect of the present invention is characterized in that a plurality of waveforms are superimposed and a peak of a phase in one period of an arbitrary waveform is detected, whereby a phase difference of any other waveform is calculated.

(24) In addition, according to one aspect of the present invention, there is provided a diagnosis support program for dividing an image of a lung into a plurality of regions, calculating an average and a distribution of intensity values of the respective regions, and acquiring correlations between count values in a lung scintigraphy.

(25) Further, a diagnosis support program according to an aspect of the present invention is characterized in that a basic waveform is generated by overlapping a plurality of original waveforms obtained from each image, a main waveform is generated based on the basic waveform, and a waveform corresponding to an organ of each image is generated by setting the bandwidth or weighting the original waveforms.

ADVANTAGEOUS EFFECTS OF INVENTION

According to one aspect of the present invention, a human organ as a subject can be grasped, and an image representing actual movement can be displayed based on the waveform or frequency of the organ or the tendency of change of the image.

Drawings

Fig. 1 is a diagram showing a schematic configuration of a diagnosis support system according to the present embodiment.

Fig. 2A is a diagram showing a cross-sectional view of the heart.

Fig. 2B is a diagram showing a cross-sectional view of the heart.

Fig. 2C is a diagram showing a cross-sectional view of the heart.

Fig. 2D is a diagram illustrating an example of Voronoi (Voronoi) division.

Fig. 3A is a graph showing the "intensity" variation of a particular block and the results of fourier analysis thereof.

Fig. 3B is a graph showing the fourier transform result of extracting a frequency component close to the heart rate and the "intensity" change of the frequency component close to the heart rate by performing inverse fourier transform.

Fig. 3C is a diagram showing an example of extracting a certain frequency band from the spectrum obtained after fourier transform.

Fig. 4 is a flowchart showing an outline of the image processing of the present embodiment.

Fig. 5 is a flowchart showing an outline of the image processing of the present embodiment.

Fig. 6 is a flowchart showing an outline of the image processing of the present embodiment.

Fig. 7A is a schematic diagram showing the left lung of a human body from the front.

Fig. 7B is a schematic diagram showing the left lung of a human body from the left side.

Fig. 8A is a schematic diagram showing the left lung of a human body from the front.

Fig. 8B is a schematic diagram showing a human left lung from the left side.

Fig. 9A is a schematic diagram showing the left lung of a human body from the front.

Fig. 9B is a schematic diagram showing a human left lung from the left side.

Fig. 10A is a schematic diagram showing the left lung of a human body from the front.

Fig. 10B is a schematic diagram showing the left lung of a human body from the left side.

Fig. 11 is a diagram showing an example of a lung field detection method according to the present invention.

Fig. 12 is a diagram showing a state in which the blood flow of the lung is uneven.

Fig. 13 is a diagram showing a state in which the blood flow of the lung is uneven.

Fig. 14 is a diagram showing how the analysis range of the lung is divided into 4 pieces.

Fig. 15 is a graph showing the result of making the change in the pixel value of the lung approximate to a sine wave.

Fig. 16 is a graph showing the result of making the change in the pixel value of the lung approximate to a sine wave.

Fig. 17A is a graph showing the correlation between the first second forced expiratory volume (FEV: forced ExpiratoryVolume in one second) and the phase shift.

Fig. 17B is an example showing a concept of phase difference.

Fig. 18 is a diagram showing how the breath is.

Fig. 19 is a graph showing changes in pulmonary blood flow within a single heart rate of a normal lung.

Fig. 20 is a graph showing changes in pulmonary blood flow within a single heart rate of an abnormal lung.

Fig. 21 is a diagram showing a normal lung and lungs of PAH (pulmonary arterial hypertension: pulmonary arterial hypertension (including pulmonary arteriosclerosis)).

Fig. 22 is a diagram showing an image of a normal lung.

Fig. 23 is a diagram showing a normal lung and a lung of COPD.

Fig. 24 is a diagram showing a three-dimensional structure on the tip-diaphragm side of the lung.

Fig. 25 is a view showing the state of the lung around the field.

Fig. 26 is a diagram showing shallow and deep breaths of a normal lung.

Fig. 27 is a diagram showing the lungs of a patient with cardiac insufficiency.

Fig. 28 is a diagram showing the lungs of a patient with cardiac insufficiency.

Fig. 29A is a diagram showing a low-resolution differential image in which the block size is increased and the variation in luminance value is counted.

Fig. 29B is a diagram showing a high-resolution differential image in which the block size is reduced and the variation in luminance value is counted.

Fig. 29C is a diagram showing an image example of MBDP and ABDP.

Fig. 29D is a diagram showing how an image of the lung is divided into 6 areas.

Fig. 30A is a graph of pulmonary blood flow between individual heart rates of the heart represented by each frame.

Fig. 30B is a diagram showing an example of measurement results of the pulmonary artery (MPA: mainpulmonaryArtery), the distal pulmonary artery (DPA: distal PulmonaryArtery), and the distal pulmonary vein (DPV: distal Pulmonary Vein).

Fig. 30C is a graph showing signal intensities in an ultrasound image of an atrium.

Fig. 30D is a graph showing signal intensities in an ultrasound image of an atrium.

Fig. 30E is a diagram showing the activity of the wave of respiration.

Fig. 30F is a diagram showing a stereoscopic structure and a cross section according to the chest XP image.

Fig. 30G is a diagram showing a normal lung and a patient's lung.

Fig. 30H is a diagram showing the lungs of the patient.

Fig. 30I is a diagram showing the lungs of a patient.

Fig. 30J is a diagram showing a comparison between deep and shallow breaths.

Fig. 30K is a diagram showing how the diaphragm and the like are shifted from the original wave.

Fig. 30L is a diagram showing an image of a pneumonia patient.

Fig. 30M is a diagram illustrating an example of an electrocardiogram.

Fig. 31A is a diagram showing an example of comparing a waveform of aortic blood flow with a waveform of ventricular volume.

Fig. 31B is a diagram illustrating an example of 2 lung images.

Fig. 31C is a diagram showing an example in which a difference is acquired from 2 lung images and a certain threshold is set to visualize the difference.

Detailed Description

The present inventors focused on the fact that conventionally, a technique for visualizing the movement pattern of an organ (for example, the myocardium of a heart) has not been put to practical use, and found that diagnosis by a doctor can be supported by expressing not only the movement of the organ but also the deviation and non-movement of the organ, and reached the present invention.

In other words, a diagnosis support program according to one aspect of the present invention is a diagnosis support program for analyzing an image of a human organ and displaying an analysis result, wherein a process for acquiring a plurality of frame images, a process for calculating a frequency representing an organ state based on each of the frame images, a process for calculating a phase difference between a waveform of an organ model acquired in advance and a waveform corresponding to the calculated frequency, and a process for outputting a signal representing the phase difference are executed on a computer.

Thus, the video can be qualitatively made and objectivity can be given in the display content while simplifying the skill required for diagnosis. In the present specification, a heart is described as an example of an organ, but needless to say, the present invention is not limited to the heart, and can be applied to various organs such as a lung and a blood vessel.

In the present application, the organ movement is regarded as constant and adjusted (normalized), and the organ movement can be determined by comparing the organ movement with a previous image of a subject, by comparing the organ movement with a range to be analyzed and an average of the whole organ, by comparing the organ movement with a normal image, and by comparing the organ movement with an age sample.

[ Basic concept ]

The basic concept of the present invention is explained. In the present invention, in characterizing periodic changes in the state of organs in a human body, for example, the cross-sectional area, the surface area, and the volume of the heart, a certain number of repeated or certain activities (routine) in the time axis are captured as waves and measured in the whole or a certain partial range of activities that are captured repeatedly at a certain period. The measurement result of the wave is (a) the form itself of the wave or (b) the interval (frequency: hz) of the wave.

For example, in a heart image, there may be waves connected in the same way in the same time. For example, if the heart rate is the heart rate, the following approximation can be defined.

(Average of density changes over some approximate range)/(heart rate)/(heart changes)/(electrocardiogram)/(surface area and volume changes of the heart)

In the present invention, for example, regarding the heart, analysis can be performed focusing on deformation (wall thickness relaxation, wall thickness contraction) of a local ventricular wall, and thickness in systole (thickness), thickness in diastole (thickness), wall motion (adventitia and intima) can be captured and expressed periodically. In addition, in the heart, "ejection fraction" obtained from "relative wall thickness (RELATIVE WALL THICKNESS: RWT)" obtained from "(2×back wall thickness)/(left ventricular end diastole diameter)" or "obtained from" (end diastole volume-end systole volume)/end diastole volume "can be captured and expressed periodically.

In the present invention, an image with higher accuracy can be extracted using any of these data or combining these data. In this case, the calculation may be performed a plurality of times. At this time, the artifact (artifact) with respect to the result is removed again, and the function is extracted by extracting a waveform from new data, a data waveform as the initial basis, a waveform of another modality, or the like, and a surrounding or multiple waveform. In this case, the number of times may be one or more.

In creating the basic data, the accuracy is improved by compensating for the extraction of the components of each other by a plurality of waveform measurements such as a plurality of modes (for example, a change amount made up of a constant density (density), a capacity (volumetry), a heart activity, and the like) or a heart rate. This can improve accuracy in light of certain expectations such as reduction of artifacts and lines.

Where "density" translates to "density", but in an image refers to the "absorption value" of pixels in a particular region. For example, in CT, air is "-1000", bone is "1000", and water is "0" for use.

In the present specification, "density" and "intensity" are used separately. As described above, the "density" refers to the absorption value, and in the original image of XP or XP video, the air permeability is high, the high permeability portion is white, the numerical value is marked, the air is "-1000", the water is "0", and the bone is "1000" for display. On the other hand, the "intensity" is displayed by relatively changing from the "density" to the "normalized" width of the density and the degree of the signal. In other words, "intensity" is a relative value such as brightness or emphasis degree in an image. The absorption value of the directly processed XP image is represented as "density" or "change in density" (Δdensity). Since this is expressed on an image, the above conversion is performed and expressed as "intensity". For example, 256 gradations from 0 to 255 are "intensity" when displaying colors. This distinction of terms applies to the case of XP or CT.

On the other hand, in the case of MRI, even if air is defined as "-1000", water is defined as "0", and bone is defined as "1000", there are cases where the values vary greatly due to the pixel value of MRI, the type of measuring equipment, the body condition, the physical constitution, and the measuring time of the person at the time of measurement, and the method for using the signal of MRI such as T1 emphasized photograph is not necessarily different depending on the type of equipment or measuring equipment. Therefore, in the case of MRI, the definition of "density" like XP or CT cannot be defined. Therefore, MRI processes relative values from the initial rendering stage, and is expressed as "intensity" from the beginning. The signal processed by the method is also "intensity".

With the above, the main data can be obtained. The main data is extracted with a certain width and range of the waveform of the main data and the Hz of the wave with respect to a new object to be measured. For example, the extraction is performed in a width or range of only the heart rate or the frame as the blood vessel extraction degree. The waveform and the width of Hz are determined relatively or comprehensively based on statistics using waveform elements, artifacts (artifacts) such as noise, waveforms of other modes (modality) considered to have tuning properties, reproducibility of a plurality of times, and the like in other functions. Here, adjustments and experience are required (machine learning can also be applied). This is because when the width or range is extended, elements of other functions start to enter, and if the width or range is too narrow, elements of the function itself are deleted, and thus adjustment is required for the range. For example, if there are a plurality of times of data, it is easy to specify a range, hz, and measurement matching width, etc. In addition, the fluctuation value and width of the axis, width, range, and Hz extracted from the mutual components can be estimated. In other words, by overlapping a plurality of times, the axis setting of Hz is calculated by averaging, dispersing, and the axis, width, range, and optimum range (range) of Hz. At this time, hz (noise) of other behavior is extracted, and if there is a wave thereof, a degree to which it does not enter is also measured relatively.

Next, the data of the new object to be measured is quantified by quantifying the form of the wave captured originally, the matching rate of Hz, or other mismatch rate, and is calculated as a value for diagnosis assistance. For example, a pulse meter is applied to a diagnosis aid by measuring the waveform matching rate of a main disease while excluding auscultation noise, and accounting for the matching rate of a disease waveform. In this specification, a case where the tuning matching rate of the pixel value can be used will be described.

[ About tuning match Rate ]

In this specification, the tendency of image change will be described as a tuning matching rate. For example, a myocardial region is detected and divided into a plurality of block regions, and the average density (pixel value x) of the block regions in each frame image is calculated. Then, the ratio (x') of the average pixel value of the block region in each frame image of the variation range (0% -100%) from the minimum value to the maximum value of the average density (pixel value x) is calculated. On the other hand, only a block region whose ratio (x '/y ') falls within a predetermined range is extracted using the ratio (x '/y ') of the ratio (y ') of the myocardial change (y) of each frame image from the minimum position to the maximum position of the myocardial (0% -100%).

Where y '=x' or y=ax (a is a coefficient of the value of the myocardial amplitude or the value of the density) is a perfect match. However, not only the case of a perfect match is a meaningful value, a value with a certain amplitude should be extracted. Thus, in one aspect of the invention, a log (log) is used, and a certain amplitude is determined as follows. In other words, when calculated at the ratio (%) of y=x, then the tuned perfect match is "log y '/x' =0". When the range of the tuning matching rate is narrow (mathematically narrow), for example, when the range is close to 0, the range is determined to be "log y '/x' = -0.05 to +0.05", and the range of the tuning matching rate is wide (mathematically wide), for example, when the range is close to 0, the range is determined to be "log y '/x' = -0.5 to +0.5", and the matching value is higher as the range is narrower. When the ratio is obtained for each pixel and the number is counted, the distribution of the positive-coefficient peaks is obtained with perfect matching in the case of a healthy person. In contrast, in the case of a person with a disease, the distribution of the ratio collapses. As described above, the method of determining the amplitude using logarithms is merely an example, and the present invention is not limited thereto.

In other words, the present invention is an invention of "image extraction" as (average of density (density) changes in a certain approximate range)/(heart rate)/(heart change)/(electrocardiogram)/(surface area and volume change of heart), and can be applied to methods other than the logarithmic method. By such a method, a frequency-tunable image can be displayed.

In the case of a blood vessel, in a series of changes in density (x (one waveform in the myocardium)) in response to a series of systoles (y), there is a slight time delay (change in phase) in this form, and so is denoted as y=a' (x-t). In the case of a perfect match, y=x or y=a' x, since t=0. When the tuning matching rate is extracted in a narrow (mathematically narrow) range, for example, a range of "log y '/x' = -0.05 to +0.05" is determined in a range close to 0, and when the tuning matching rate is in a wide (mathematically wide) range, for example, a range of "log y '/x' = -0.5 to +0.5" is determined in a range close to 0. The narrower the range, the higher the number of matches within the range, so to speak, the higher the matching rate.

In the case of other blood vessels, excluding the above "part responding to heart", the density (density) of the central side drawn from the phylum of the lung may be used. In the case of peripheral blood vessels, the same process may be performed.

In addition, the present invention can be applied to a circulator, for example, in which a change in the density (density) of the heart is directly related to a change in the density (density) of the blood flow from the portal to the outside Zhou Fei field, and a series of changes in the density (density) of the heart or a change in the density (density) of the portal are propagated as they are after receiving a single transformation. This is thought to be due to the fact that several phase differences are obtained from the varying relationship between the density (density) of the heart and the density (density) of the lung portals. Further, since a change in density (density) of the portal part or the like is correlated with a change in density (density) of the blood flow to the lung field as it is, the tunability can be expressed as a (relationship of matching rate of y.about.x) reflected in the ratio as it is. In the cervical vasculature, it is considered that the density (density) of the adjacent central cardiac vessels is directly related to each other or is related to a slight phase. Then, the density (density) varies according to the background, and when the propagation is performed, the density (density) is transmitted as a change in density, and can be considered as a tuning matching rate.

In the case where the relative value (STANDARD DIFFERENTIAL SIGNAL DENSITY/density: standard differential signal density/Intensity) when the change amount is 1 is to be displayed as the change amount from the density (density) of the heart among the change amounts per 1 image and the change rate per 1 image, the change amount and the change rate can be plotted for (1) images (normally virtual) when the change amount is 1 per image among the difference images of each image, (2) for 1 heart rate which satisfies the density (change amount or change rate) among the difference images of each image, and (3) for 1 total density (density) among the heart rates in the plurality of shots.

In addition, although it is the case of 3D such as MR, the difference between the value of the intensity (intensity) (in the case of MR) or the value of the density (density) (in the case of CT) of the heart rate (in this case, 1) and the intensity (intensity) or the density (density) can be converted into "peak flow volume data (peak flow data)" of the heart rate (even at rest or under load), and by outputting the value as the ratio of the intensity (intensity) or the density (density), the measured operation amount and the operation rate in the heart can be converted at least when calculating the "3d×time (time)" on MRI or CT or the like. Similarly, by inputting 1 cardiac output, the distribution in the "capillary phase (CAPILLARY PHASE)" in the "flow" of the lung field can also be presented as an estimated value converted into the distribution and volume of the lung blood flow.

In other words, when 1 sheet of change amount of only 10% or 20% is obtained, the estimated value can be calculated by (total number of sheets) x (change amount of time) when (average of density (density) change in a certain approximate range)/(heart rate)/(heart change)/(electrocardiogram)/(surface area and volume change of heart) is satisfied.

Then, the image of the new object to be measured is calculated as a diagnosis-assisted image by imaging the form of the original captured wave, the Hz matching rate, or other mismatch rate. For example, the usual swallowing and patient swallowing differences are visualized and represent differences in actions performed so far and actions that are now being performed. Such as the manner of movement of the walking leg, variations in rocking, discrepancies, etc.

The extracted variation is visualized and depicted as an image. This is the heart function analysis and blood vessel (blood flow) analysis described below. The myocardial rate of change is then visualized. In this case, the artifact with respect to the result is removed again, and the function is extracted from the new data extraction waveform, the data waveform as the initial basis, the waveforms of other modes, and the like, and the surrounding waveforms and waveforms. The method of removing the artifact will be described later.

In addition, even if the variation component extracted from the above-described extraction is excluded, the feature quantity is grasped. For example, when the activity of the abdominal intestine is grasped, the influence of respiration and the influence of blood vessels are removed from the abdomen, and the activity of the abdominal intestine is extracted.

In addition, based on the change rate extracted from the image, an image (CT, MRI, special X-ray imaging, PET/scintigraphy, etc.) requiring a certain imaging time is corrected, thereby providing a clearer and more accurate image. For example, the present invention is effective for ascending aortic heart correction, heart morphology correction, bronchial blurring correction, evaluation around the chest, and photographing of a state where breath holding is impossible (in some cases, photographing of a patient takes several minutes).

In addition, by dividing a plurality of waveforms into one period and superimposing data, a base waveform can be obtained. The waveforms of each converted image are applied by associating the bandwidth (frequency width) of the basic waveform, the weight with respect to the basic waveform, the main waveform component obtained by averaging or dispersing the basic waveform, and the weight or bandwidth component with respect to the waveform obtained from the original image, respectively. The width component of the weighting or logarithmic transformation is applied to the waveform of each transformed image, so that the width component can be used for calculation at the time of image transformation such as logarithmic transformation. In other words, a basic waveform is generated by overlapping a plurality of original waveforms obtained from each image, and a main waveform is generated based on the basic waveform, and on the other hand, a waveform of an organ corresponding to each image is generated by setting the bandwidth or weighting the original waveforms. This can contribute to diagnosis support. In addition, the width of the tuning matching ratio, the difference between the tuning matching ratio and the ratio becomes clear, and a certain bandwidth can be measured. Thus, by forming the bandwidth and varying the phase, the anomaly can be extracted by performing simulation. In the image to be imaged, when the density at the peak of the phase of one cycle is measured, the pixel value of the peak can be calculated, and the position where the phase difference appears firmly can be easily grasped based on the peak. When the phase shifts in one cycle, a place where the phase shift is large is extracted. Even for other components of the period shift, a place where the shift is large can be extracted, and these can be displayed as an abnormality. When there is an abnormality, both the waveform and the amplitude are shifted, and thus the waveform can be clearly extracted. Further, the correlation with the "Wave form" can be further improved, and the correlation can be used for correction (correction). For example, in the case of pulmonary blood flow, the peak starts to shift from the heart rate after the occurrence of the heart rate, after reaching the pulmonary blood vessel for 0.016 seconds. Depending on the person, 0.024 seconds is sometimes required. Therefore, the waveform of the pulmonary blood flow can be obtained from the heart beat. Peaks of pulmonary blood flow, peaks of peripheral blood vessels can be obtained from changes in density. In addition, when the phase shift of the pulmonary blood flow is displayed in the past, the treatment is performed with 2 frames fixed with respect to the central phase shift, but this method does not see the blood flow which occurs with delay. In the present application, these blood flows can be displayed. The present application is not limited to blood flow. In other words, the above method can also be applied to other organs or other breaths.

The search for the tuning matching rate includes not only a transformation that finds a certain phase shift, but also a phase that deviates from a certain level, and the extraction of abnormal lesions is performed by searching for matching rate data and images without performing simulation in all phases. Similarly, with respect to the amplitude, a simulation was performed for different states, and a matching image close to the "wave form" was drawn. This can notify the abnormal drawing and identify from different matching rates. When capturing the active phase, a wide frequency is preferably used. For example, at 60-80 Hz. The "doppler effect" may be a phenomenon in which the frequency is increased or decreased, and as a result, the phase is changed. To capture this, an extended width (range) is required.

In addition, "tune imaging" of the relative lung field space can also be performed based on the structure of the organ and the proportion of its space. In the present CT, even in a large portion such as 320 columns, the entire lung field is not photographed at one time, and therefore, in the case of partial photographing, a space of a portion which has not entered the imaging range at the end of inspiration so far enters the image at the end of expiration, a portion which has entered the image at the end of expiration enters the imaging range at the end of inspiration, and the like, but the relative position, the relative "volume", and the air fluctuation can be similarly made into "tune imaging", only from the entire lung field space. Further, as for the difference of the "CT data", the "reconstruction (reconstruction)" of the clear CT or MRI is performed by eliminating the gap, and the "reconstruction (reconstruction)" is performed for the organ or the like performing the movement by matching the signal with the change of the total "basic data (wave form) -based", so that the difference data and the difference image which are clear with respect to the more periodic movement can be collected.

[ About phase shift ]

It is known that an organ (abnormal organ) with poor condition responds with a delay relative to a histogram of a normal organ. For example, the activity of the abnormal part is delayed from the activity of the other part, whether the heart or the lung. Therefore, by displaying how much the shift from the normal frequency is, the site of the abnormal organ can be displayed as an image. As an example of the appearance of the video, only the activity of the abnormal portion is delayed, and the signal value becomes high. In addition, the initial activity delay is not necessarily the initial activity delay, and the graph is recognized as a graph offset when the graph is formed. Thus, the abnormal organ is delayed in both phase and amplitude from the normal organ, and these differences (gaps) are calculated. In other words, the degree of phase change of the change in the pixel value is displayed with reference to the change in the normal pixel value. In addition, the frequencies are the same, but may also appear as different waveforms. In this specification, these relationships are referred to as "phase relationship image (PHASE RELATIVE IMAGING)" or "phase difference image (PHASE DIFFERENTIAL IMAGING)". As shown in fig. 17B, the phase difference or phase shift refers to not only the phase difference PH1 in the same wave but also the concept of the phase difference PH2 in a plurality of different waves, but in this specification, refers to the phase difference PH2 in a plurality of different waves.

For example, the lung shape is divided into 8 pieces, and the average value (intensity) of the pixel values included in each region is taken and plotted. It is determined that the phase of the normal lung activity exactly coincides, but the phase of the abnormal lung activity is shifted. The phase shift becomes a numerical indicator for judging abnormality and normality. Specifically, the phenomena of "bright line shift of image", "phase shift from the point of time of the start" are identified. Further, when the average value of the pixel values in a predetermined region is plotted, the phase shift can be confirmed.

In the present application, the difference in density is displayed as a phase difference. In the present application, it is shown how the phase difference, i.e., the degree of activity, is. The small moving part changes little, and the large moving part changes big, so it appears in the image. That is, the phase difference appears as a waveform, but appears as an image delayed in time. In particular, it delays when the activity is started and then gradually delays. For example, in the case of an abnormal lung, the upper lung field has followability, but the more toward the middle lung field and the lower lung field, the more remarkable the delay. There are also places that are not active.

Waveforms acquired from the organ are synthesized from multiple waves. The phase difference is extracted from the waveform and imaged. That is, not only where the offset is concerned, but also the delay or offset in time. For example, in the "lung of a patient suffering from pulmonary thromboembolism", a waveform in which blood flow is uneven and phase is shifted appears. There is a temporal offset in which the flow of blood flow becomes delayed. In addition, the offset becomes larger as it goes from the center to the periphery of the lung. In the present application, only DICOM data is used, and images showing phase differences can be displayed. In the prior art, it is difficult to show only the phase difference in the X-ray image. In the present application, diagnosis support is enabled by visualizing information hidden in an X-ray image.

The following method can be adopted in terms of matching the phase with a reference value, for example, an average value. For example, breath-hold of the heart has less impact, but random, involuntary movements are mixed in the breath. Therefore, as shown in fig. 18, a phase (in the above example, the jetting state, in fig. 18, a central plateau region is mixed) which is not related to a part of normal respiration (normal wave continuity) is mixed. In this case, for example, a portion different from a normal tendency in a certain range of a waveform element component (a shin value) is cut out from a certain waveform component deviated from an average of phases, and a phase which does not pass through a certain value (for example, when breathing or exhaling in a region where a correlation (delayed) amplitude value is not up to and down to 50% is performed in an upper and lower wave (wave)), a phase of a wave deviated from a range of 90% to 110% of the average is calculated and removed. Thus, a waveform image of a certain quality can be extracted from a certain clinical need in which the dynamics of the respiration and the blood flow are visible. Note that, AI may be learned by a change in adjacent pixel values, a variance value, or the like.

[ As to the differences in density in organ images ]

In the present application, phase shifts in waveforms of organs are tracked, and density differences in images are displayed. For example, in the case of a lung, instead of tracking the entire lung field, a portion of the phase shift is tracked and the density difference calculated. In other words, the relative density difference of a portion of the lung is calculated. In this case, the Voronoi (Voronoi) division method can be used to enlarge a certain area. In the present application, the relative positional relationship is maintained, and the difference is outputted as compared with the density before and after the movement. For example, it is displayed at each branch of a blood vessel (blood vessel branched from an artery, etc.). Then, the density change in a place is filtered to show the difference. Instead of tracking the relative position of the entire lung field, organs (sites, constituent elements) or organs prone to radial (proximal and middle lobes of the lung, etc.) are tracked.

In other words, since the movement of each blood vessel can be followed, the density in each blood vessel is filtered, and the difference is displayed as it is. Since it is possible to follow the movement of a structure such as an organ, the meat is stuck to the structure, and thus the structure of the lung is formed from the tracking result. Since the change in density is displayed in the position of each organ, whether or not it is a relative positional relationship is determined. Focusing on "how to follow the lung fields", the expansion of the region divided using Voronoi (Voronoi) is performed from a point of following a part of the lung fields. The shape of the entire lung field can be determined to follow the movement of the lung field, but the entire shape can be estimated from the determination of the position of the center (point) of a certain area. Instead of determining the post-frame follow-up, the lung field can be activated by determining the post-point follow-up of the point. Note that, since it is not realistic that the amount of calculation becomes a problem in following all points of the organ in the image, it is preferable to determine points in the image of the organ such as the lung and to follow the points. This point then determines the area of responsibility.

Alternatively, the density may be calculated by removing unnecessary frequencies with a filter. The position where the density is changed may be shifted or may be not shifted. I.e. not a relative positional relationship. In the present application, the waveform of the organ can be separated and grasped by software. This can be applied not only by acquiring biological information from a camera image but also from a chest belt, an electrocardiogram, or the like. In addition, noise may be incorporated into the biological information. When noise is externally added, the amplitude suddenly increases, and a spike waveform appears. It is removed. In addition, noise that is regularly incorporated can be removed. In the image output method, the shape of the lung is fixed, and the change in pixel value can be displayed as a color image in the interior of the shape. By this method, the blood flow can be displayed as a color image.

[ Removal of specific frequencies ]

The frequency known in advance can be removed with a filter. For example, the frequency of occurrence in the electromyogram is removed with a filter. The frequency of 50 Hz-60 Hz in the upper arm, 100-200 Hz in the thigh or abdomen, and 0.2Hz in the breath. Thus, the frequency thereof is removed by the filter. In this way, the frequency that becomes noise contained in the electrocardiogram can be removed, and only the heart element can be extracted. Thus, there are methods for extracting the heart/heart rate component and methods for removing the same in contrast. That is, the natural frequency and natural frequency inherent in the organ can be used to extract the desired frequency and remove the unnecessary frequency. The method can be applied to pulse meters, chest belts, cameras, electrocardiograms and the like.

In order to remove noise and separate noise, FFT processing and band-pass filtering processing are performed. In addition, since the pulse can express a frequency, the frequency can be extracted from the thoracic band, and thus can be removed and extracted from the spectrum. For example, it is possible to extract respiratory elements contained in an image of the lung or heart as respiration or remove respiratory elements as noise.

[ About the combination with AI ]

The AI (ARTIFICIAL INTELLIGENCE: artificial intelligence) can learn normal images and abnormal images for diagnosis support. For example, the lung is thin on the outside and large on the inside. Since the difference in the breath of the pneumograph is taken, the height is known. The normal lung has a "thin periphery and thick inside. The initial inclination of the movement is small and the gradual inclination is larger. "feature of the following claims. Imaging the color of the color image changes gradually with the change of the color image. Among them, as the tilt of the method of pulmonary relaxation, the tilt is expressed by differentiating the lung volume with time. For example, a red-colored portion indicates rapid contraction in the lung at a location where the density is rapidly increased. On the other hand, a blue portion shows rapid relaxation in the lung at a portion where the density rapidly decreases. With mild abnormalities, the lungs do not move normally and produce parts that cannot turn red or blue. In addition, there are cases where red and blue are reversed. In all or part of the lung, the activity of breathing in at the site where breathing in is to be performed can be found. In contrast to the normal case where the variation is constant, the method of variation becomes opposite or uneven. A method of causing AI to learn such a change. Then, the AI output can be made based on the feature quantity of the changed method. By associating the feature quantity with a disease or a case, diagnosis can be supported.

Regarding the creation of the model based on AI learning and the judgment of the use of the model, in the step of collecting information to create the model, the rising and falling portions of the waveform are captured by each phase segment of the waveform. For example, to learn the phases of normal exhalation and inhalation. In other words, a normal model is initially created, and secondly, an abnormal lung that does not match the normal lung is learned. Next, in the step of using the model to apply the model to the phenomenon, the model is used to determine that the output is abnormal in a few percent. For example, it is possible to show the phenomena of a sudden thinning and disappearance of the trunk and branch of the blood vessel, a sudden increase in density after the blood vessel is thinned, a phase shift, a lack of occurrence, and a sparse image. When these phenomena are identified, pulmonary Thromboembolism (PTE) can be diagnosed. Taking PTE as an example, this is part of the disease, and other various diseases can also produce such a phenomenon. For example, the decrease or increase in signal intensity, the deviation in reactivity (sensitivity to signal change), the absence of a signal, or the like can be used to grasp the disease condition.

Such PTE patients are characterized by uneven and offset flow of pulmonary blood as shown in fig. 12 and 13. I.e. a phase shift is generated. In the present application, in terms of morphology, a portion with a high signal is generally radial, and the next phase is seen as a whole, and if a phase shift occurs, it can be determined that an abnormality is caused by a lesion. AI can judge these phenomena. In combination with such AI, tracking something that is periodically active becomes a feature.

"Removal of artifacts"

It is necessary to capture the rib-induced artifacts and remove them. In the lung, the branch and leaf structures of blood vessels are visible from the phylum of the lung to the periphery. It is approximately uniform, symmetrical left and right, and extends to the whole. Imaging the blood vessel leaves and branches of the blood vessel appear, and the space between the leaves and branches is also red. This extracts and displays the change in density as a frequency signal. Among them, the diaphragm and surrounding heart are prone to artifacts. The inner side on the lungs is also prone to artifacts. Because the ascending aorta is visible. The lung activity per cycle is modeled on AI, learning anomalies, and artifact removal is performed.

The inventors of the present invention have devised a method for removing rib-induced artifacts. It is desirable to distinguish between rib-induced artifacts and respiratory changes, either alone or in combination. The method described below can be applied to any one of a perspective image, a differential image of a perspective image, and an image after application of a filter. In addition, as a secondary effect, since the position of the rib can be specified, not only the removal of the rib but also the appearance of the rib movement can be visualized.

The block intensity showing the pixel value of the minute region is calculated every frame, but in the block into which the rib enters only on the minute region of the specific frame, the absolute value of the change in the block intensity is relatively large compared with the change in the respiration. Regarding this relatively large change as a costal artifact, by correction, the costal artifact can be specified and removed. As a correction method, the following logic can be cited as an example: the method includes the steps of "determining the maximum value and the minimum value by pinching (limiting the value x to be within the predetermined range [ a, b ])", "judging the artifact by whether or not the predetermined threshold value is exceeded", "contracting the maximum change to be within the threshold value if the predetermined threshold value is exceeded", and "taking the artifact when the form (Weve form) as the reference wave is deviated from the form by a predetermined ratio.

As described above, in any block, the presence of frames into which ribs enter a minute area and frames into which ribs do not enter are the sources of the generation of artifacts. Therefore, in any block, by changing the shape of the block to include or exclude ribs in all frames, artifacts can be reduced. As a method of changing the shape of the block, the following logic may be cited as an example: "enlarging the size of the block", "making the block a rectangle of about 2 times the rib thickness", "excluding the region taking the offset value from the pixel values contained in the block".

The method has the defect that resolution is reduced when the size of the block is increased. By acquiring the difference between the low-resolution difference image (the image in which the size of the block is increased, and the variation in luminance value is counted) shown in fig. 29A and the high-resolution difference image (the image in which the size of the block is reduced, and the variation in luminance value is counted) shown in fig. 29B, it is possible to extract a block that varies more than the surrounding blocks from the high-resolution difference image. By this method, artifacts can be reduced while maintaining resolution. Further, since the difference between the low-resolution differential image and the high-resolution differential image is switchable as a parameter for characterizing the movement of the lung structure, it has a suggestion that it can be used as a parameter having a certain clinical meaning.

As another artifact removal method, a method of extracting a high-density portion from the outside of the lung contour and identifying it as a bone may be considered. This is a method of measuring density from outside of the lung contour and performing position calculation to confirm a rib as an artifact. In other words, although the description of the rib may be a problem, the original image is an X-ray image, and it can be estimated that a place with a high density (a place with a high density) is the rib. Therefore, it is considered that by performing masking processing, overlapping is performed to cancel out the pixel value, and a portion of the rib can be erased. After removing the ribs, interpolation processing or differential processing is preferably performed.

[ Learning and judging according to AI for pulmonary blood flow ]

From the air intake during inhalation, the lung structure density became low, and the lung permeability in the X-ray image was high, and the image appeared white. On the other hand, since the air is reduced during exhalation and the lung structure density is high, the permeability is low, and the image appears black (dark). In addition, in the case of blood vessels, since the blood in the lumen of the blood vessel in diastole increases, the diameter of the blood vessel increases, and thus the X-ray permeability decreases. As the blood in the lumen of the blood vessel in systole decreases, the diameter of the blood vessel becomes smaller, and thus the X-ray permeability increases. As such, the density (density) of the lung or blood vessel image periodically varies. In the present invention, a specific region (organ or the like) in an image is visualized by adjusting the frequency, which is the period of fluctuation.

For example, when visualizing the movement of a blood vessel of the lung, the following steps are performed.

(1) The blood flow frequency at the time of photographing is identified.

(2) A frequency spectrum composition table was created for the difference value of the front-to-back density (density) of each region of the XP video.

(3) The frequency obtained from (1) is taken out of the spectral components.

(4) The obtained extraction frequency (respiratory frequency) is imaged.

(5) And outputting and displaying the frequency tuning image.

The video obtained in such a step has the following features. I.e. as video extending radially from the center in a healthy lung. On the other hand, in the case of a patient (e.g., PTE: pulmonary thromboembolism), a blood vessel that should be visible becomes invisible and has a video of unevenness.

In the present invention, for example, a video of 12.5 frames per second is set. In other words, 12 images (photographs) can be taken in 1 second. Also, regarding normal lungs, it is considered from two general perspectives.

Lung vessel structure "

In normal lungs, from the main vessels to the peripheral vessels look like branches. It appears to flow radially uniformly. The hub appears thick and thick.

Capillary peripheral pulmonary vascular flow (Normal pulmonary flow: normal pulmonary blood flow) "

With a set of tiny dots arranged consecutively, it looks like a cloud.

Thus, the vascular structure of the lung, the capillaries and peripheral pulmonary vessels are seen because the blood flow velocity is high in the pulmonary artery trunk and the central side, and the blood flow velocity is low in the peripheral vessels.

Fig. 19 is a graph showing changes in pulmonary blood flow within a single heart rate of a normal lung. Wherein a single heart rate is shown in 12 frames. In normal lungs, respiration occurs approximately 45% of the time and global blood flow is observed. In the latter half of a single heart rate, blood flow is not observed as in the first half due to the rest of the heart. In contrast, in the case of a patient suffering from pulmonary thromboembolism, the appearance is different. Fig. 20 is a graph showing changes in pulmonary blood flow within a single heart rate of an abnormal lung. Wherein a single heart rate is shown in 6 frames. As shown in fig. 20, the vascular structure of the lung does not appear to spread radially from the center and is not uniform. In the image, the missing portion is abnormal. The left lung has normal vascular structure, but the right lung is sparse and has little blood flow. The blood vessel itself is not visible. In the case of a normal lung, the vascular structure of the lung should be clearly visible radially, but this is not the case in fig. 20. In this way, the vascular structure of the lung can be diagnosed. In addition, even in the case of an abnormal lung, the MIP image has a feature that the image is missing and a thin portion cannot be seen. The number of voids is large and the density is low. Thus, according to the present invention, the vascular structure can be examined, and the possibility of pulmonary thromboembolism can be grasped from the image display mode.

[ About capillaries, peripheral pulmonary vascular flow (Normal pulmonary flow) ]

In abnormal lungs, no capillary vessel/peripheral pulmonary vascular flow appears in the image. Thus, it can be determined as abnormal. For example, in the periphery of the lung, the thickness of the lung is small, but in the case of an abnormal lung, an image of a portion (a portion shown as red in the image) where an increase in density (density) hardly occurs is formed. In the case of a normal lung, the portion of increased density (density) is distributed by showing a gradual change, but not in the case of an abnormality. Thus, the difference between normal and abnormal is represented.

If an acute pulmonary embolism (acute pulmonary embolism: APE) is suspected, the following features appear in the image. In other words, blood flow occurs after respiratory action at the periphery of the lungs. I.e. delayed blood flow occurs in the image. This is where blood begins to flow in the occluded vessel, so blood flow occurs late in the image. As shown in fig. 20, at "stenosis: in partial occlusion, stenosis ", the blood vessels appear in the latter half of a single heart rate. I.e. can be seen after the heart cycle. As shown in fig. 20, at "occlusion: in total occlusion, no blood flow appears in the image. Nor is it visible in the MIP image. In this way, images of the normal lung and the abnormal lung are taken as learning targets of AI.

Further, if capillary vessel-peripheral pulmonary vascular flow is not delineated (in phase) after the blood vessel of the pulmonary artery, the delineation abnormality caused by thromboembolism is suspected. If the capillary/peripheral pulmonary vascular flow is delayed and plotted, (a) the blood vessel is delayed and plotted on the phase (phase) in a state where the blood vessel is thin by "stenosis (stenosis)" or the like, and (b) the capillary/peripheral pulmonary vascular flow is not plotted on all phases (phase), it is estimated that the blood flow has been interrupted by the complete occlusion, as "occlusion".

In other words, the missing part in the vascular structure can be grasped, and the capillary and peripheral pulmonary vascular flow of the finer blood vessel (TERMINAL VESSELS: peripheral blood vessel) can be examined in combination with the vascular structure. Then, by examining the non-flowing portion on the early phase (phase after vascular delineation) and the portion flowing from the delayed phase, it can be judged whether or not "partial occlusion" is suspected: stenosis% by weight of a metal alloy. Further, by examining the portion that does not flow over the entire phase (phase) of the early phase to the late phase, it is possible to determine whether or not "total occlusion" is suspected: occlusion).

For example, in APE, blood flow to the right lung may occur rapidly. That is, capillary and peripheral pulmonary vascular flow may occur in the area where the occlusion is supposed to occur. This is a phenomenon that occurs when blood rapidly flows into peripheral blood vessels as a result of the main flow of the blood vessels being interrupted and blocked.

In addition, fig. 21 is a diagram showing a normal lung and lungs of PAH (pulmonary arterial hypertension: pulmonary arterial hypertension (including pulmonary arteriosclerosis)). In PAH, the lungs are seen as a whole without large loss, but with non-uniformities. The normal vascular structure takes a structure extending radially from the center of the lung in the peripheral direction. The branches are hardly visible. However, if the lung has PAH symptoms (abnormalities), the branches of the blood vessel become "deformed shapes". In other words, if the lung has PAH-based abnormalities, the number of radial lines in the vascular structure is small, and additionally tapers off at the outer Zhou Jisu. That is, the outer periphery becomes extremely fine due to the rarefaction and deformation of the vascular structure. As a distribution, it appears to be uniform, but if carefully observed, the distribution is random, unlike normal. The vascular structure of the central lung is also missing. This is evidence that the part is not active. In addition, outflow of blood flow is delayed, and appearance of blood vessels also becomes delayed. For example, the blood vessels are delayed by 1 to 2 frames, and the vascular structure is visible starting from the central periphery of the lung. This is unusual. In addition, although the blood vessels on the peripheral side of the lung are thin, blood flow occurs at a delayed timing. This is evidence of a delay in blood flow, indicative of an abnormality in the lungs. In addition, it can be seen that the vascular structure itself may flicker or delay appearance in the image. In addition, if there is no capillary or peripheral pulmonary vascular flow, it is abnormal if these are not visible or only a few are visible. These phenomena were imaged as differences from normal.

The mechanism up to the signal imaging of the blood vessels of the lung will be described. In the blood vessels of the normal lungs, first, a pressure difference is formed across the blood vessels due to the driving of the heart. Second, due to the elasticity of the blood vessel, expansion and contraction are generated. That is, the blood vessel expands and contracts due to the elastic force of the blood vessel. As a result, the vessel diameter fluctuates. That is, under normal circumstances, the size can be seen in the size of the lumen of the blood vessel. The signal is identified as a periodic change in the density (density) of XPs and reflected in the image. Second, in the case where arteriosclerosis such as PAH exists, even if the heart is driven, it is difficult for the blood vessels of the lung to develop a pressure difference. In addition, it is difficult to generate expansion and contraction due to vascular elasticity. This is because the stretching force is reduced due to arteriosclerosis (fibrosis) and the lumen hypertrophy. Thereby, the propagation force of the pressure difference to the peripheral blood vessel by the heart is reduced. In addition, it is difficult to be reflected in the vessel diameter. That is, even if the heart is driven, the change in the size of the blood vessel diameter becomes weak. As a result, the periodic XP density (density) changes, which are expressed as various abnormal elements in the image. For example, an overall decrease in signal value (pixel value) can be seen. In addition, since a pressure difference is difficult to form and expansion and contraction due to elasticity of a blood vessel are difficult to occur, it is depicted that ejection is delayed or the whole is weak due to pulmonary artery. Further, since expansion and contraction due to elasticity of blood vessels are difficult to occur, capillary blood vessels and peripheral pulmonary blood vessel flow from the center to the periphery of the lung are reduced, and the blood vessel phase image becomes scattered. Since the change in the size of the blood vessel diameter becomes weak, it becomes difficult to trace the blood vessel itself from the peripheral blood vessel. The blood vessel on the central side is in a rod-like branched shape, and the "diffuse" around the outer periphery is reduced.

From the above, it can be said that the elasticity of the blood vessel is lost due to the change in the hardenability (elasticity) of the blood vessel, and the drawing of the blood flow becomes difficult. In addition, in the case where the density (density) inside is difficult to perform, no signal (pixel) will be seen. In addition to the PAH described above, the same pattern of images is expected to occur in chronic heart failure, hyperparathyroidism, chronic renal failure, collagen diseases such as amyloidosis, and other changes in pulmonary arteriosclerosis.

[ Regarding ventilation-blood flow mismatch ]

Regarding the ventilation and blood flow mismatch of the lung in the past, diagnosis was performed by pulmonary blood flow scintigraphy. These relationships can be understood physiologically. In the present invention, the ventilation and blood flow are displayed on the screen at the same time, and the determination is made by a combination of respiration and blood flow. For example, if the lung is not in a state of normal motion, it can be determined that the lung is abnormal. For example, when ventilation is normal but blood flow is abnormal, pulmonary hypertension and pulmonary thromboembolism may be considered, and when both are bad and blood flow is bad, emphysema may be considered. On the other hand, there are cases where the blood vessel activity is normal, but the lung activity is abnormal. For example, heart failure and abnormal respiration are equivalent to this. Even in normal people, the heart rate is correct but the breathing is difficult, for example, throat is choked with food. Such a "ventilation-blood flow mismatch" model is generated by machine learning according to AI.

For example, in "cases of hypoxic pulmonary disease (HLD)", capillary vessel peripheral pulmonary vascular flow is lacking. That is, there is a region where the respiration of the lung is absent, and the blood flow in the corresponding region is also deteriorated. The same phenomenon can be seen in COPD as well. In the present invention, diagnosis can be performed by learning AI by associating and comparing ventilation and blood flow.

[ About correlation with pulmonary scintigraphy ]

Fig. 29D is a diagram showing an example of images of MBDP (Maximum Blood-flow-RELATED DIFFERENTIAL project: maximum Blood flow-related differential Projection) and ABDP (Accumulated Blood-flow-RELATED DIFFERENTIAL project: cumulative Blood flow-related differential Projection). With respect to MBDP being "for each block, the maximum value between frames is selected and projected into one image", ABDP being "for each block, the value added with the forward change value is projected into one image". As shown in fig. 29E, as an operation method of MBDP and ABDP, an object is to divide an image of a lung into six regions and calculate an average and distribution of intensity values in each region as a clinical index.

For example, correlations may be obtained between pulmonary scintigraphy (an existing method of assessing pulmonary blood flow). Pulmonary scintigraphy measures "blood integrity" by injecting a radioactive agent and acquiring a "count of radioactive material", which is characterized by a strong signal at the site of blood blockage. In the lung scintigraphy, the image of the lung is also segmented into six regions, and counts of radioactive material are acquired at each region. In the lung scintigraphy, for example, when the counts are "100, 200, 300, 400, 500, and 600", the counts are calculated as "4.7%, 9.5%, 14.3%, 19%, 23.9%, and 28.6%", which are percentages of the total count 2100, by the evaluation of the relativity. In MBDP and ABDP, the sum of the intensities of each region can also be re-read as a count of lung scintigraphy and can be evaluated as a percentage of the total intensity. Thus, an image close to a scintigraphy can be obtained. For example, when counting the number of breaths and blood flows, whether the number of breaths and blood flows is delayed or rapid and whether the number of breaths and blood flows is substantially counted may be counted as a signal. The total amount of the flow can be grasped, and thus the diagnosis support is facilitated.

[ About events with reproducibility ]

In the present invention, the AI machine learning is made to repeatedly occur a plurality of times, that is, an event having reproducibility. Thus, a method of learning a normal state to determine abnormality or a method of learning an abnormal state to output case alternatives can be adopted. As an output method of AI judgment, a method of outputting a feature quantity of each case is adopted. The feature quantity may be determined according to an image pattern (average pixel value of each region or ratio of variation of pixel values, etc.), and includes various diseases.

[ About pulmonary blood flow and AI ]

Fig. 30A is a graph of pulmonary blood flow between individual heart rates of the heart represented by each frame. In fig. 30A, the pulmonary blood flow pattern between individual heart rates is shown in 10 frames. In the flow of a single heart rate of pulmonary blood flow, flow begins from vessels on the central side of the lung and then spreads to peripheral end vessels. Because of the narrow diameter of the end vessel (small vessel lumen), the flow rate is slow and there is much, so in the image, it is delayed after the hub and appears as a cloud. In other words, after seeing the main (Mainpulmonary artery: pulmonary artery, distal pulmonary artery: distal pulmonary artery) blood flow, the end vessel (TERMINAL VESSELS) can be seen.

Fig. 30B is a diagram showing an example of measurement results of MPA (MainpulmonaryArtery: pulmonary artery), DPA (Distal PulmonaryArtery: distal pulmonary artery), and DPV (Distal PulmonaryVein: distal pulmonary vein). After MPA, DPA was observed with a slight delay, but with a sharp drop from MPA. In addition, the TV (TERMINAL VESSELS: terminal blood vessel) described above appears before DPV. In fig. 30A, as the measurement result shown in fig. 30B, after blood vessels are seen from the center to the periphery in the first 2 frames and 3 frames, blood capillaries appear from 4, 5, 6 frames.

As how to learn the AI, a "bulls-eye" that expresses the lungs and the heart in concentric circles is employed. The blood vessels normally spread in a substantially concentric circle from the pulmonary valve, but have phases in which the pixel values of the blood vessels are peaks (third and fourth frames of fig. 30A). Wherein both vascular structures and peripheral blood vessels are accessed. Therefore, in order to grasp a signal (pixel value) of pulmonary blood flow, a peak of the pixel value is first detected. Then, the phase of the vascular structure is seen through the vicinity of the peak thereof, and thus, the vascular structure is extracted. The detection method includes (1) a method of "extracting a vascular structure", and (2) a method of "extracting the vascular structure from the hilum or the vicinity thereof by flowing the pulmonary blood from the hilum to the periphery".

For example, in the lung of a patient suffering from acute pulmonary embolism (acute pulmonary embolism: APE), since the vascular structure becomes sparse, the lung field region is divided into about 17 equal parts in concentric circles of the bulls-eye. When the shape of the lung is distributed in concentric circles, it is possible to see what occurs when the phase of the blood flow and the phase of the blood flow of the distal blood vessel are distributed. Therefore, AI conversion becomes easy by displaying the distribution deviation. This is because the distinction between normal and abnormal is facilitated by the distribution bias. In other words, the regions are spatially divided to quantify the pixel values of regions where blood flow enters and regions where blood flow does not enter. This may be calculated automatically. Since quantization is possible, AI can also be applied. By this method, it is possible to calculate from the segmentation, and a missing part or the like can be detected.

Regarding the end blood vessel, when the pixel value of the blood vessel structure region is subtracted from the whole, the pixel value of the end blood vessel can be obtained. In other words, the remaining portion becomes the end blood vessel in addition to the blood vessel structure, and can represent what blood flow the end blood vessel is.

In addition, in the three-dimensional structure, which part is missing (whether there is blood flow) or the like can be detected. When set to 3D or 4D, it can also be displayed in the same manner. In other words, the blood vessel structure is segmented, the vicinity of a portion where the pixel value (signal) of the blood vessel structure is high is detected, and the phase thereof is determined. By this method, a vascular structure can be detected, a terminal blood vessel can be detected, and what the blood flow (flow) is can be displayed. When these signals are collected, a bar graph showing the rise and fall may indicate what activity is being done in the blood vessels within the area. In addition, it is also possible to indicate what activity is being done in the end vessels within the region. It is also possible to indicate how it flows as a whole. When these are calculated in a single heart rate, the waveform shown in fig. 30B can be obtained.

Fig. 30C and 30D are graphs showing signal intensities in an ultrasound image of an atrium. In fig. 30C, E as a peak signal and a as a signal indicating the influence of atrial activity are observed. In fig. 30D, a so-called L signal having a residual waveform appears between the signal E and the signal a. In the case where L occurs between E as a peak signal and a as a signal representing the influence of atrial activity, it is considered to be likely to be heart failure. Thus, the difference between normal organs can be grasped by patterning and quantization, and lesions can be predicted.

In addition, a method of comparing waveforms of past and current electrocardiographs is known, but for example, by overlapping waveforms shown in red, blue, and green, automatic diagnosis can be performed by comparing before and after treatment, comparing normal persons and patients, and comparing the same patient in the past. For example, by grasping the shift of the T wave, the unstable rise after the peak, or the like, automatic diagnosis can be performed, and a doctor can record even the findings.

[ Learning and judging by AI of respiration of lung ]

With respect to lung respiration, the tunability of changes in the lung fields is tracked. In other words, it is determined whether the lung activity spreads from the center to the periphery of the lung. Specifically, the mobility of the movements of the diaphragm, ribs, and lung tissue was observed. In addition, the action of pectoral major muscles can be combined. In addition, in the image, the degree of the gap can also be evaluated. Furthermore, even with normal lungs, the appearance may be different due to natural or effort to breathe, but these are judged to be normal. The "area of the change in the lung activity", "continuity of blood flow, flow pattern", "continuity of the activity from the center of the lung to the outer periphery", "activity of the diaphragm itself", and the "linkage" can be used to determine the difference.

More specifically, the correlation is used to evaluate a wave (for example, a wave of a respiratory element) extracted from the image. The lung activity is focused on the wave, and the magnitude of the amplitude, the phase shift, and the direction of the vector are examined, and the AI is determined in combination with the anatomical knowledge. In this case, the judgment object will change depending on what wave is extracted from the image. In other words, since there is a frequency or a phase shift unique to the disease (case), the disease (case) is specified based on the detected frequency and phase shift. When the lung is cured due to tuberculosis, the frequency becomes zero. For example, in cases of neurotransmitter difficulty, in cardiomyopathy, WEB forms are delayed and waveforms are different while the phases are shifted. Since there is a frequency unique to the disease, a frequency-tuned image unique to the disease can be output. Thus, a new feature amount can be obtained or a new image can be output.

In the present invention, imaging is performed in conjunction with respiratory rate. That is, when an image matching the breathing frequency is extracted, an image of the movement of the lung due to breathing will be seen. Wherein differential values of the lung activity are imaged. Such images extract only the changes along the respiratory rate in the X-ray image. Since the change in density (density) is tracked, the whitening (decrease in transmittance) in XP is due to the increase in density (green, yellow, red) and the darkening (increase in transmittance) in XP is due to the decrease in density (green, bluish, blue), which change is visualized in the image. The image in the present invention is opposite to the inhalation/exhalation diagram of the precision lung function test.

For signal changes in lung activity, the signal value increases along the trachea-bronchi (central), bronchioles, and alveoli (peripheral) at the beginning of respiration, and decreases along the alveoli (peripheral), bronchioles, and bronchioles (central) at the end of respiration. This can also be said to be inhalation, exhalation. The same is true in XP, and when the permeability increases in the case of air intake, the density decreases, and the difference becomes a negative vector (the color is blue). On the other hand, in the case of exhalation, when the permeability is reduced, the density is increased, and the difference is a forward vector (red in color).

Fig. 22 is a diagram showing an image of a normal lung. As the movement of the breath in the inspiration direction (increased permeability) increases, the transition from bluish to blue (strong negative density vector). On the other hand, as the movement of the exhale in the exhale direction (decrease in permeability) increases, the vector of density changes from yellow to red (positive direction is strong). The thinner the lung is, the weaker the overall signal, while the thicker the lung is, the stronger the overall signal is. The signal intensity becomes greater or lesser depending on the wave of "inspiration, expiration" activity of the breath.

[ About correspondence between images and diseases ]

Fig. 23 is a diagram showing a normal lung and a lung of COPD. It can be grasped that in normal lung, the posture of "bronchus to bronchioles to alveoli of bronchus to lung segment" expansion can be seen, but in COPD lung, only "bronchus to bronchioles" expansion, and "bronchioles to alveoli" do not. In this way, in the lung of COPD, the central portion of the lung is unevenly imaged. The left-right asymmetry is also related to abnormality determination. The same is true for the linkage. The movements of the diaphragm, ribs, lung tissue become scattered. In the present invention, the linkage between the reference wave and each structure is correlated with the anatomical structure and evaluated.

Fig. 24 is a diagram showing a three-dimensional structure on the tip-diaphragm side of the lung. As shown in fig. 24, in the "lung tip", the three-dimensional structure of the lung is thin. Therefore, in the image in the present invention, the variation itself is small. In addition, drawing defects due to overlapping at the time of lesions are few. In "above the septum" the three-dimensional structure of the lung is thick. Therefore, in the image in the present invention, the variation appears strongly. In addition, the drawing defect increases due to overlapping at the time of lesions. In the image of the present invention, the change is still strong in the "slightly more caudal than the lung apex (below the lung apex)". It can be said that the overlap at the time of lesions is also general. For example, depending on the lung structure, a difference between the left and right of a person easily occurs to the extent of thickness. In the case of "under the diaphragm", since the soft shadow is strong, the difference between the XP images on a sheet-by-sheet basis is hardly present in the numerical value, and evaluation is not easy.

Thus, even in a normal lung, the following events may occur.

Reduced rate of change in lung field weeks

Reduction of pulmonary Structure around the leaf Interval

Unevenness of the changes caused by movement of structures in the vicinity of the pulmonary gating

Mild signal reduction inside the lower right lung field

Changes in thickness of the lung on the apex side due to thickness of the anterior and posterior chest, pectoral muscle, etc

Artifacts of bones of the rib

Shadow of only rough branches and leaves

Fig. 25 is a view showing the state of the lung around the field. In the lung field periphery, the leaflet structure is stereoscopic and complex, resulting in rough peripheral variations as it is reflected in the image. The spread around the lung may be relatively reduced due to the shape of the lung, rib delineation, or delineation of pectoral major muscles, etc. In addition, there may be a difference in the depicting ability from the peripheral bronchioles to the alveoli with respect to the appearance of the edges of the lungs. This also varies depending on the degree of respiration and the respiration situation. In other words, the signal variation according to the breathing cycle occurs consistently. In addition, depending on the respiratory cycle, thin, thick changes occur from the tip of the lung to the diaphragm side or from the outside to the inside of the lung. But is approximately side-to-side symmetric. In the present invention, it is evaluated whether or not the wave activity as the respiratory cycle and the lung activity are linked. Whether or not the rib cage, diaphragm, rib, etc. are moving in association with the wave as the respiratory cycle was evaluated.

Regarding the lung structure around the leaf fissures, a decrease in signal value (pixel value) can be seen. This means that, especially on the margin of the leaf fissure around the lower left lung field, there is little signal change due to the low alveolar density change. In addition, a slight decrease in signal was observed in the lower right lung field, and although a negative signal was observed due to the structure below the diaphragm, correction was performed so that these were in the normal range. In the invention, when distinguishing normal videos, the images are read while comprehensively considering the three-dimensional chest structure and the respiratory physiological change rate. In addition, by performing image reading while considering the structure and image conversion in various XPs, various cases such as partial functions of the lung can be distinguished.

In performing machine learning and determination according to AI, it is critical to define a reference wave such as a respiratory cycle and whether or not to link with the reference wave. In other words, the wave linkage is determined from the viewpoint of whether there is a large fluctuation, a small fluctuation, or no fluctuation. If the fluctuation is small at a place where the fluctuation should be large or large at a place where the fluctuation should be small, it can be determined as abnormal. In addition, the phase shift may be determined to be abnormal with respect to the reference wave. That is, waves consisting of respiratory cycles with a certain degree of connectivity are considered normal lungs. If the AI were allowed to learn it, it would be a model of normal lung. In AI determination, it is determined whether its linkage is reserved or lost. In addition, there are cases where the amplitude is abnormal or there is a case where the phase is abnormal.

Fig. 26 is a diagram showing shallow and deep breaths of a normal lung. In the shallow breath of the normal lung, the slope of the wave is weak and the change is small, so that the periphery (peripheral bronchioles to alveoli) cannot be delineated. In this case, only bronchi to bronchioles are depicted. On the other hand, in the deep breath of the normal lung, the gradient of the changed wave is strong, and the change is large, so that the outer periphery (peripheral bronchioles to alveoli) can be depicted. In the abnormal deep breath of the lung, the activity is delayed with respect to the activity of the diaphragm. Thus, in the present invention, respiration can be evaluated by "linkage", and thus can be diagnosed by linkage.

Fig. 30E is a diagram showing the activity of the wave of respiration. When the slope of the waveform is large, it is clearly visible when displayed in the image. This is because everything is linked to the morphology and slope of the wave. In the case of a normal lung, the septum, thorax, trachea and signal output patterns are the same. The degree of coincidence with the wave can be diagnosed by plotting a graph.

In addition, as a three-dimensional structure, the thickness of the lung can be captured. For example, in chest XP images, CT is represented by circular slices, but the volume is superimposed. Fig. 30F is a diagram showing a stereoscopic structure and a cross section according to the chest XP image. The outside of the lung is thin in the three-dimensional structure and the image changes little. In the case of lesions, the degree of defect in image rendering due to superimposition is small. The inner side of the lung is thick in the three-dimensional structure and the image changes greatly. If there is a lesion, the degree of defect in image drawing due to the overlapping becomes large. The change of the image from the outside to the slightly inside (outside of the intermediate layer) increases to some extent, and in the case of a lesion, the defect level of image drawing due to the overlapping also increases to some extent.

Fig. 24 is a diagram showing a side surface of the lung. Also, when viewed from the side, it is related to the front-rear thickness. It can be grasped that the pixel value is increasing or not increasing in the stereoscopic structure. The spatial structure of the lungs, whether each structure is distended or not, can be displayed. In the case of shallow breathing, the signal is not strong because the lung is inflated slightly. But if the wave is in place, it can be displayed by grasping the slope and can be useful for diagnosis. Regarding the structure of the trachea itself, it should also be easier to see at the beginning of inflation (because of the large slope of the change), but if not, this means an abnormality, the possibility of lesions can be predicted. In addition, if it is not seen when it should be seen on the central side of the trachea, it can be judged as abnormal. The amplitude moves along the volume and wave morphology. If not, it can be judged as abnormal. If the same amplitude is obtained, it can be judged that the amplitude is abnormal.

Fig. 30G is a diagram showing a normal lung and a patient's lung. Fig. 30H is a diagram showing the lungs of the patient. In these images of the lung, the "red" portion indicates that the pixel value is increasing, and in respiration, indicates the state of being breathing. The area and volume of the lung are reduced and the pixel value is positive. The "blue" portion indicates that the pixel value is decreasing, and in breathing, indicates the state of inhaling. The area and volume of the lung increase and the pixel value is negative. These represent anomalies in activity. In the patient's lungs, it is shown as blue when it should be red, or as red when it should be blue. That is, the event is performed in reverse to the original event, and it can be determined that an abnormality exists. In addition, if the separator is moving without a color change, it may be determined that there is an abnormality. For example, after the AI learns the moving image of the normal lung and builds the learning model, the presence of the lesion can be determined for the image of the patient's lung. Further, by quantifying the features of the image of the normal lung and comparing the quantified features with the features of the image of the patient's lung, it is also possible to determine abnormality.

Fig. 30I is a diagram showing the lungs of a patient. In this case, the central vascular structure of the lung is visible, but the peripheral vascular structure is not. Peripheral alveoli are not active. Normal lungs have wave correlation, but abnormal lungs have no wave correlation. In this regard, the AI may perform learning and determination. In addition, quantization can be performed in a certain area, and the degree of difference can be grasped. Fig. 30J is a diagram showing a comparison between deep and shallow breaths. The image is shallow during shallow breathing, and the image is deep during deep breathing. A segmentation map is calculated and drawn from the morphology of the wave.

Fig. 30K is a diagram showing how the diaphragm and the like are shifted from the original wave. These deviations are based on the phase shift. For example, in the lung of a patient suffering from interstitial pneumonia, as shown in fig. 30K, the phase is shifted from the original wave. The phase shift is quantized. Particularly effective is the heart. In the case of neurotransmitter delay, myocardial activity may be delayed. This delay in myocardial activity means a phase delay. These phase shifts are quantized and pixelated. In the case where a delay in propagation of (the disorder of) neurotransmitters is visible, there is a high possibility that myocardial infarction occurs after one or two years. In other words, whether the amplitude of the quantized or pixelated wave is shifted, and whether the phase of the wave is shifted.

In the case of COPD, the phase of the pixel changes of the lungs will also shift. For the tip of the lung, the activity deviates and the response is poor. The phase of the change in pixel values of the lungs of patients with interstitial pneumonia will also shift. In contrast, normal persons have no phase shift. The pattern of such phase shifts can be learned and judged by AI. In particular, since it is difficult to grasp with the naked eye, it is quantized, making AI learning and judgment very useful.

Fig. 30L is a diagram showing an image of a pneumonia patient. Indicating that the structure of the lung is lost. Vascular structures extending radially from the lung gate are normal, but are displayed completely randomly. This may be judged as abnormal. In addition, as shown in fig. 30L, the vascular structure of a portion of the hydrothorax patient is also randomly displayed. In this case, since a white spot appears in the X-ray image, a doctor can make a diagnosis in combination therewith. In other words, the combined image and the X-ray image are used together with the image of only the breath and the image of only the blood flow, and thus the comprehensive judgment can be performed.

Fig. 30M is a diagram illustrating an example of an electrocardiogram. As shown in fig. 30M, in addition to the pulse portions shown in a certain period, signals deviating from the period are displayed. The main reasons for this are that muscle activity, lung activity, skeletal creak is data that the noise goes into the electrocardiogram. That is, data on an electrocardiogram deviates due to the influence of other organs or sites. Conventionally, these deviations have been corrected manually, but in the present invention, the correction process is automated. In other words, these data are handled as "frequency tunable data" as a frequency tunable image. More specifically, these data are fourier-transformed, only the frequency of a periodic pulse signal appearing in an electrocardiogram is filtered out, and finally noise-reduced electrocardiogram data can be obtained by inverse fourier transformation. Breathing noise can be removed as well.

[ Lung of patient with heart failure ]

Fig. 27 and 28 are diagrams showing the lungs of a patient suffering from heart failure. Fig. 27 shows an initial state in the first half of a single heart rate, and fig. 28 shows a posterior state in the first half of a single heart rate. The normal vascular structure adopts a structure extending radially from the center of the lung in the peripheral direction, and branches are hardly visible. However, in the case of a patient with heart failure, the branches of the blood vessels of the lungs are in a "distorted shape". In other words, in the case of an abnormality of the lung based on heart failure, the number of radial lines in the vascular structure is small, and in addition, the outer Zhou Jisu tapers. That is, the outer periphery becomes extremely fine due to the sparse and distorted vascular structure. As a distribution, it seems to be uniform, but if carefully observed, the vascular structure becomes "sparse", so to speak, "diffuse sparsity". This is different from normal people. Capillary vessel peripheral pulmonary vessel flow, a blood flow delay was observed but occurred in general. In general, the activity shifts, as waves of the blood flow, produce a phase delay. As shown in fig. 28, even in a period in which the blood flow ends, the blood flow is observed. This indicates that blood flow is delayed from the hilum. In normal lungs, blood flow is observed during periods when blood flow is not observed. These phenomena were imaged as differences from normal.

[ About 3D tomography ]

In the three-dimensional display technology of an image, a projected image based on three-dimensional information is displayed on an image display device such as an LCD which is a two-dimensional plane. The three-dimensional image display technology is classified into a projection method (parallel projection, perspective projection) on a two-dimensional plane as an image display surface and an image display method (surface-rendering, volume-rendering, maximum value display, minimum value display, sum value projection, etc.). It is to be understood that the three-dimensional multi-plane reconstruction Method (MPR) is also included in the three-dimensional display system, in which the display of any cross section is not performed without using the projection method.

Conventionally, in IVR (Interventional Radiology: interventional radiology), a subject is transported to a CT room to take a tomographic image or an IVR-CT apparatus is used. Wherein the C-arm CT imaging technique according to the C-arm-equipped angiography apparatus avoids the transportation risk by acquiring tomographic images in the IVR on the spot, and further, achieves excellent cost performance in terms of cost and use as compared with the IVR-CT apparatus. Among techniques for reconstructing tomographic images from rotational images of a C-arm, there is a 3D angiography (Angio) technique, but the 3D angiography technique mainly depicts contrast vessels, whereas C-arm CT imaging is very excellent in low contrast resolution, and can depict a light tumor depth.

In the image processing in the C-arm CT imaging, a tomographic image is obtained by performing three-stage image processing of conversion processing to a water penetration length, reconstruction preprocessing, and reconstruction processing on the rotational image obtained by imaging. In the conversion process to the water penetration length as the initial stage, the pixel value of the rotated image is converted to the water penetration length. The water penetration length is an amount equivalent to how many millimeters of water is penetrated by converting a pixel value into water under the same X-ray conditions. Specifically, after gain correction of the detector, the scattered ray component is corrected by the scattered ray correction process, and converted into a water penetration length in the beam hardening correction process. The dark part in the rotated image means that the X-ray absorption is large, and therefore the converted water permeation length becomes long, and the converted image looks like a positive-negative inversion.

In the following reconstruction preprocessing, four types of correction processing are performed as follows: c-arm orbit correction, ring artifact correction, intercept correction, and pi correction. The C-arm orbit correction is a correction of the offset of the real C-arm orbit and the theoretical ideal orbit. The ring artifact correction is correction for reducing ring artifacts in tomographic images that become obstructing image observation. The intercept correction is correction to improve flatness of a tomographic image by performing sinogram interpolation of an object portion protruding from a field of view during rotation. Pi correction is for correcting the difference in the amount of collected data generated in the central portion and the peripheral portion due to the cone beam effect at the time of half-scan reconstruction. After the above correction, the tomographic image is calculated by the reconstruction process. The algorithm of the reconstruction process is called the FBP (FilteredBack Projection: filtered back projection) method.

In C-arm CT imaging, about 30 minutes are required from various angles. For example, since the breathing cycle is not constant, the reconstruction is used for correction. Wherein, the part near the bulb is clearly visible, but the part far away is not clearly visible. It can be seen that if the voltage is increased, it is necessary to perform a process of obtaining a difference and eliminating noise while maintaining the difference in permeability. For example, only a near site is observed, then the site is observed from the rear, and by subtracting, the front and rear can be seen. If the voltage is increased, all can be seen, but for unknown reasons, pulling the near side makes the back side easier to see, and conversely, pulling the back side makes the near side easier to see.

Traditionally, no method of making the image visible while correcting it in a breathing state has been achieved. In the present application, the operations of inhalation and exhalation are regulated to be constant. The phase and frequency are combined for reconstruction. In other words, under a plurality of conditions, a plurality of images photographed from a plurality of directions are used to process each other, the plurality of conditions are changed, and a tomographic image is created. And, it is used for X-ray photographing. In other words, a tomographic image in the frequency-analyzed X-ray image is created. Finally, a fine adjustment based on the reconstruction is performed. With respect to a specific organ, tomographic images are created at a plurality of positions and under a plurality of conditions. For example, the waveforms are matched with respect to the active lung and blood flow, phase by phase. With respect to breathing, phase alignment is also performed. Then, when visualized, a positional alignment and a stereoscopic impression are generated. Regarding blood flow, matching (synchronization) in time is also required. In this case, for example, 10 heart rates are taken each, and then the phases are matched.

In the present application, since the images are taken while being moved, the frequency of the organ is considered. Therefore, the positional alignment, the phase alignment, and the angle adjustment are performed in time and in phase. In the C-arm CT imaging, since the imaging positions for the organs are not exactly the same, it is necessary to perform positional alignment and angular alignment. For example, when attempting to photograph the lung from diagonally right and front at an angle of 45 degrees, it may be 43 degrees or 49 degrees in practice. It is necessary to fine-adjust such an offset according to the positional relationship with the ribs or the thorax.

In other words, the CT image, the C-arm CT cross-sectional image, or the XP image obtained from other angles are aligned so as to match the previously obtained 3D image of the structure such as the "vertebral body, heart, mediastinum structure, rib, or the like", or the angle, or other bronchi. The order of high matching is "cone, heart, mediastinum structure, rib", and in particular, the angle and position matching coefficient of the cone is enhanced, the coefficient matching by the lung motion such as rib is slightly reduced, and the most matching angle and position relation of each XP image captured by AI is grasped as an image pattern, and when creating a three-dimensional structure, the creation of a cross section, a three-dimensional 3D structure, 4D creation, and the like are performed based on the information thereof. In particular, since the CT image is an "inhalation image" captured in the inhalation state of the subject, matching is considered to be most easy when the XP image at the time of maximum effort inhalation is caused to coincide with the source of the image. In addition, when the XP is imaged from a plurality of angles, if there is no image peculiar to the subject, for example, CT or C-arm CT tomographic image, etc., there is also a case where such angle imaging is calculated from a standard image model by information such as age or height, and by fine adjustment, there is also a possibility that a cross-sectional structure or a three-dimensional structure and a three-dimensional cross-sectional structure or a three-dimensional structure according to the movement of an organ such as a breath or a heart are created.

In addition, albumin is injected into the patient, and a state of fine pulmonary thrombosis is intentionally created. In normal lungs, albumin is distributed throughout the lungs. In the case of pulmonary thromboembolism, albumin does not reach due to peripheral blockage. Thus, unevenness is generated in the image. As seen using the 3D integral structure, the anterior (anterior side of the lung) entered, but the posterior (dorsal side of the lung) did not. Then, a tomographic image is created from the 3D. If the phase shift in the video is large, it is known to be abnormal because it does not go on.

In addition, in the case of X-ray photographing of AP (anti-posterior: front-rear), (1) found on the lung or front side (A), this may be due to the distance from the detector (detector) near the bulb. (2) Since the photographing is performed several times, there is also photographing in a state where the dose cannot reach the back (P side) under an environment of insufficient dose (low dose). For these reasons, in the default imaging for perspective confirmation, the state of being closer to the bulb side is stronger (the a side in the AP imaging, the P side in the PA imaging), and the relative change and signal of the portion closer to the detector (the far from the bulb (the P side in the AP imaging, the a side in the PA imaging)) are more difficult to distinguish than the bulb side. Therefore, (1) by performing the same photographing (and more uniform image quality and more emphasized image of the P side photographing) at the high dose portion and the conventional low dose photographing, it can be classified into front photographing, overall, and photographing to the rear side. (2) By AP photographing and PA photographing, photographing of each face can be performed. (3) There are cases where signals are delayed in the back side of the lower leaf or the like, and differences in signals in phase are liable to occur. (4) In some cases, a rearward image is easily drawn with respect to breath-hold. (5) Even in breathing, changes occur more easily in deep breathing than in shallow or normal breathing.

With the above trend, the two are superimposed (the overlapping of the AP and PA and the overall signal ratio are objectively drawn), and the above is subtracted, and imaging is performed with emphasis on the front (a side) or the rear (P side) or the like, depending on whether the front is dominant or the overall is positive, whether the rear is dominant, and whether the voltage is high or more visible (voltage high > voltage low). In addition, shooting is performed from a plurality of directions, including RAO or LAO, RL direction, and the like. In addition, by photographing at the time of rotation or spiral movement, various photographing can be performed at a plurality of angles on the bulb side. Since all breaths are not constant, by using normalized dilation, reduction or time-dependent reconstruction (reconstruction) on the basis of the main waveform or the like (base), approximate respiratory, blood flow elements in its cross-section can be imaged by overlapping on previously simultaneous taken tomographic 3D images of a standard lung or patient, matching signals in position, relative phase. In addition, by multi-angle transmission imaging, a multi-angle 4D image or a signal in each section can be generated for a three-dimensional time.

[ Method for pulsing the human body ]

ASL (ARTERIAL SPIN labeling of arterial spin) is a method of imaging by marking arterial blood flowing into an imaged cross section with RF pulses, and obtaining perfusion images or relative cerebral blood flow images without using a contrast agent. The RF pulse is an impact signal that if applied periodically will become a marker knowing how far the neck has been rotated. For example, when an RF pulse is applied to the neck, the pulse spreads throughout the head. Pulse diffusion is normal, but may not be. Among them, frequency tunability can be recognized and predicted even in the head, and it can be predicted that it is possible to grasp that blood flow is easy or difficult to flow. Since the frequency applied to the human body is clear, it is possible to judge whether it is normal or abnormal by deviating or delaying the frequency.

[ About the behavior according to sine waves ]

As shown in fig. 14, the analysis range is divided into four parts in the image of the lung, and whether or not there is a difference in the change in the signal of each region is confirmed by the following method. That is, the "signal change: f (x) "approximates" g (x) =a×sin (cx+b) ", and the correlation between each parameter and the FEV% predicted value was investigated.

As a result, as shown in fig. 15 to 17, in healthy lungs, "r4→r3→r2→r1" and amplitude become large, tending to phase match. In contrast, in COPD cases, it was visually confirmed that the amplitude and phase were different. From studies of the approximation of the sine wave, it is shown that the phase shift is inversely related to the FEV% prediction (r= -0.775). When acquiring the correlation, one point is excluded as an outlier, but this is considered to be due to a method of creating a lung field (analysis range). The artifacts of the diaphragm and rib mainly affect the amplitude, and no trend to contribute to the phase shift was observed. At least in evaluating the phase shift, it is believed that it is not necessary to actively exclude R4/L4. On the other hand, in the case of performing a certain evaluation using the amplitude, attention needs to be paid to the influence of the artifact.

[ Estimation of pulmonary blood pressure ]

Traditionally, in order to measure pulmonary blood pressure, catheters are inserted to measure pulmonary blood pressure, but are highly invasive. In the present application, waveform data may be obtained from a part of an organ, but when an average of a part is obtained, a sine wave is represented. The sine wave is believed to have a correlation with pulmonary blood pressure. In other words, by the image analysis of the present application, the waveform for obtaining the pulmonary blood pressure can be calculated from the frequency, phase and signal value of the waveform. Wherein periodic activities matching heart rate are extracted and imaged. At this time, the change in the fitting blood flow is tracked in the image. For example, the sine wave will not be transmitted to the lung gate or periphery. In addition, a delay from the center of the lung to the outside occurs. Thus, the velocity of the blood flow can be calculated by the delay. Pulmonary blood pressure can be estimated from the thickness and flow rate of the blood vessels. In addition, it is possible to grasp the condition of the blood vessel, the phase shift, the difference in partial view, the respiratory state, etc., as shown in the present specification, and the heart failure, the condition in the whole or part of the disease, and grasp the degree of the disease. For example, learning the heart index or pulmonary arterial pressure inhalation pressure by AI helps the treatment guidelines for heart failure. By calculating the degree and proportion of respiratory activity and overlapping respiratory coefficients from the apex to the diaphragm, outside and inside, an estimated respiratory function can be calculated. In addition, in matching signals (signals) such as respiratory function constants and flow of blood flow, it is possible to measure a mismatch coefficient and predict a state of ventilation blood flow mismatch, ventilation blood flow imbalance, and an obstruction caused together with pulmonary ventilation blood flow (corresponding to pulmonary obstructive pulmonary blood flow obstruction). That is, it can be classified into good lung ventilation and blood flow, good lung ventilation and blood flow abnormality, good lung ventilation and abnormal blood flow abnormality, and countermeasures for controlling the disease state and the treatment policy can be determined according to the degree thereof.

The pulmonary blood flow waveform calculated in the present application is a waveform quantized by using the degree of change of an arbitrary region as a graph, and approximates to a waveform of a pulmonary artery recorded at the time of an actual right heart catheter examination. After the pulmonary valve of the pulmonary arterial pressure waveform closes the notch, the angle until returning to its baseline, corresponding to the right ventricular diastole, reflects the pulmonary vascular resistance. With this feature, the difference between the PVR high value group, the PVR low value group, that is, the CpcPH patient and the IpcPH patient can be detected with respect to the correlation between the angle of diastole of the pulmonary blood flow waveform calculated in the present application and the actual pulmonary arterial pressure waveform.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. Fig. 1 is a schematic diagram of the components of the diagnosis support system according to the present embodiment. The diagnosis support system performs a specific function by executing a diagnosis support program on a computer. The base module 1 is composed of a cardiac function analysis unit 3, a blood flow analysis unit 5, another blood flow analysis unit 7, a fourier analysis unit 9, a waveform analysis unit 10, and a visualization/quantization unit 11. The base module 1 retrieves image data from a database 15 via an input interface 13. The database 15 stores therein images according to DICOM (DIGITAL IMAGING AND Communication IN MEDICINE: digital imaging and communications in medicine). The image signal output by the base module 1 is displayed on the display 19 through the output interface 17. Next, the function of the basic module of the present embodiment will be described. The input image is not limited to the database 15, and may be input from another external device or actively input through the input interface 13.

[ Refinement of dynamic part detection ]

The contrast of dynamic parts of the lung fields, thorax, heart, etc. may not be uniform along the line. In this case, by changing the threshold value for noise reduction and performing the detection processing a plurality of times, the shape of the dynamic part can be detected more accurately. For example, in the left lung, the contrast of the lines of the septum tends to decrease as it enters the interior of the human body. In addition, there are cases where the movement of the apex and the bottom of the heart is small and the movement of the heart is large in the central part of the heart. In this case, by changing the setting of the threshold value by noise reduction or multiplying a different coefficient by a pixel value, the remaining portion of the left half of the diaphragm or a region where the heart activity is large and a region where the heart activity is small can be detected. By repeating this process a plurality of times, the shape of the entire diaphragm or the shape of the entire heart can be detected. Thus, by this method, not only the position of the diaphragm but also the shape of the chest, heart, and dynamic part can be quantified in terms of the rate of change and the number of changes in the line or surface, which can facilitate new diagnosis.

The thus detected position or shape of the diaphragm or heart can be used for diagnosis. In other words, the coordinates of the diaphragm or the heart can be plotted, and the coordinates of the chest, diaphragm or heart can be calculated using the curve (local surface) or the straight line calculated as described above, and the heart rate, vascular beat, the "density" of the lung field, or the like can be plotted as the position or the coordinates corresponding to the cycle. This method can also be applied to dynamic sites that are linked to respiratory or cardiac pulsations.

By this method, not only Hz in inspiration, expiration, and systolic or diastolic phase, but also when the frequency of a dynamic part linked to a diaphragm or respiration or the frequency (Hz) of a dynamic part of the heart is changed, measurement can be performed with a bandwidth according to the change. When the spectrum of the BPF (bandpass filter: band pass filter) is extracted, the axis of the position of the BBF can be varied in the "reconstruction phase (reconstructionphase)" of each of the breath and the heart within a certain range according to the respective states of the breath and the heart, thereby generating an optimal state and creating a BPF with a variability matching the optimal state. Thus, even if the breathing delay, the rhythm of the breathing fluctuates like at the time of stopping (hz=0), the temporal fibrillation of the heart (extremely high frequency), or at the time of stopping (hz=0), an image according thereto can be provided.

In this specification, regarding "(a) the form itself of the wave", the concept of "waveform tunability" is used and an image is displayed based thereon (Wave form tunable imaging: waveform tunable imaging). In addition, regarding the interval (frequency: hz) of the above "(b) wave, the concept of" frequency tunability "is used, and based on this, an image (Frequencytunable imaging: frequency tunable imaging) is displayed.

For example, in the case of a heart, as shown in fig. 31A, "an example of a waveform of contrast aortic blood flow and a waveform of ventricular volume", the peak of aortic blood flow and the peak or waveform of ventricular volume do not match. However, in fig. 31A, if the equidistant time width is determined to be one cycle at times t1 to t2, times t2 to t3, and times t3 to t4., one cycle of the aortic blood flow and one cycle of the ventricular volume are repeated a plurality of times, so to speak, the frequency of each waveform is tuned. With this waveform in mind, the waveform (Wave form) can be predicted by determining one cycle from the actual measurement value shown in fig. 31A and using the model waveform. In other words, as a method of creating the "waveform as basic data", it is possible to actually measure, generate from frequency (cycle), use a model waveform, and average waveforms among individuals. If the circulation (cycle) of an organ having a frequency such as a heart is known, the waveform (Wave form) of the Wave can be predicted, and therefore, the waveform such as the aortic blood flow rate and the ventricular volume can be grasped, and based on the waveform, a dynamic image of the organ can be displayed.

In addition, in the continuously photographed images, the movement of the dynamic part linked to the diaphragm or the respiration can be detected. In continuously photographed images, when images are selected at arbitrary intervals and differences between the images are calculated, the differences increase, particularly for areas having a large contrast. By properly visualizing the difference, the active area can be detected. In the visualization, the waveform (Wave form) may be created by noise reduction according to a threshold value, or by emphasizing the continuity of a region where the absolute value of the difference is large in curve fitting by a least square method or the like. In the lung field, when the contrast of the line where the diaphragm and the heart meet is emphasized, as shown in fig. 31B, and a certain threshold is set to obtain a difference between two lung images, and the difference is visualized, as shown in fig. 31C, the line where the diaphragm and the heart meet can be visualized.

In addition, the overall frequency of exhalation or inhalation may also be calculated based on the proportion of respiratory elements (elements of the breath including all or a portion of the exhalation or inhalation) to exhalation or inhalation. Likewise, the systolic or diastolic phase, frequency element, or other global frequency may be calculated based on the percentage of dynamic elements of the heart (including dynamic elements of the heart in all or a portion of the systolic or diastolic phase) to the systolic or diastolic phase of the heart or to one beat of the heart, the measurement of global beats. In the case of detecting a diaphragm or a heart, the diaphragm or the heart may be performed a plurality of times, and a signal or a waveform may be selected to be stable. From the above, it is possible to calculate at least one frequency of respiratory elements or cardiac pulsation elements or calculate a frequency indicative of heart rate from the detected position or shape of the diaphragm or heart or the position or shape of the dynamic part linked to respiration. If the position or shape of the diaphragm or heart, or of the dynamic site can be grasped, the frequency or heart rate of the respiratory or cardiac pulsation element can be grasped. According to this method, even if a part of the waveform is separated, the subsequent waveform can be tracked. Thus, even if the frequency of the respiratory element or cardiac pulsation element changes halfway, the original respiratory element or cardiac pulsation element can be followed. In addition, although heart beat and the like may change suddenly, it can be equally applied to organs related to cardiovascular, cardiovascular waveforms.

[ Lung field detection ]

In the present invention, as one aspect of the "refinement of dynamic part detection" described above, refinement of lung field detection can be achieved. In this process, after setting the maximum lung fields and the minimum lung fields, the values thereof are then used to calculate other lung fields. Fig. 11 is a diagram showing an example of a lung field detection method according to the present invention. In this method, "coefficient per image" is represented using a "B-spline curve". In fig. 11, a waveform X indicates "a coefficient of each image L showing the lung field", and becomes a coefficient of the first image and a coefficient … … of the second image in the left-to-right direction. When the control point Y in fig. 11 is moved, the "coefficient per image" becomes smooth. In the present invention, the graphs of the coefficients can be directly edited in this way. In fig. 11, "broken line Z of gray" represents "pixel average value per image". When photographed under optimal conditions, the size variation of the lung fields matches the variation of the pixel mean value. Wherein the pixel average can also be smoothed using curve fitting and used directly as a "coefficient". The same applies to organs related to heart, other cardiovascular frequencies.

[ Analysis of Heart function ]

Fig. 2 is a sectional view showing a schematic structure of a heart. The "cardiac function" is generally defined as "pump function that circulates blood to the left ventricle in the body". In "ischemic heart disease", especially "myocardial infarction", analysis of cardiac function is important for estimating prognosis of a patient. For example, when the value of left ventricular Ejection Fraction (EF) decreases, the output of the pump as the heart decreases, and sufficient blood cannot be delivered to the whole body. Other cardiac functions include left ventricular End Diastole Volume (EDV), left ventricular End Systole Volume (ESV), stroke Volume (SV), cardiac Output (CO), and Cardiac Index (CI). As a local evaluation of cardiac muscle, as shown in fig. 2B and 2C, which are cross sections according to a plane orthogonal to the axis a of fig. 2A, a "Bull's eye map" showing wall thickness, wall motion, wall thickness change rate, and the like is used. The "Bull's eye map" is an image as follows: the cross section of the apex of the heart is arranged at the center of the circle, short axis tomographic images are sequentially arranged concentrically to the outside, and the cross section of the bottom of the heart is arranged at the outermost side for display.

In this embodiment, a "Bull's eye map" is used, and the period of the heart activity is analyzed based on the following index. In other words, the density/intensity (Density/intensity) in a certain defined region within the heart region is used to analyze the period of heart activity. In addition, data obtained by other measurement methods such as a range of a certain volume density/intensity (volume density/intensity) measured at a site where the transmission of X-rays (a plurality of types of modalities such as CT and MRI) is high, and a lung volume map, or external input information may be used. It should be noted that it is preferable to compare the analysis results of each individual heart rate, analyze trends from a plurality of data, and improve the accuracy of the data. In addition, the edge of the heart may be determined, and the frequency may be obtained based on a change in the edge of the heart. In addition, the edge of the lung fields can be determined, and the frequency can be derived from the activity of that edge.

[ Vascular beating analysis ]

In the present embodiment, the vascular pulsation is analyzed based on the following index. That is, the measurement results of other modalities such as an electrocardiogram and a pulse meter, or the changes in density/intensity (density/intensity) of each site from the lung contour to the heart/pulmonary portal site/main blood vessel are determined, and the vascular beat is analyzed. In addition, the change in density/intensity of the target site may also be manually plotted and analyzed on the image. Also, heart rate elements obtained from heart rate or vascular beating may be used. It should be noted that it is preferable to compare the analysis results of each individual heart rate, analyze trends from a plurality of data, and improve the accuracy of the data. In addition, density/intensity (density/intensity) extraction for each site may be performed multiple times, and accuracy may be improved by performing the extraction over a certain range. In addition, there are methods of inputting cardiovascular beat frequencies or frequency bands.

[ Identification of cardiac region ]

Images are extracted from a Database (DICOM) and cardiac regions (in particular the myocardium) are automatically detected using the results of the cardiac function analysis described above. Next, the myocardium is divided into a plurality of block areas, and a change of each block area is calculated. Wherein the size of the block area can be decided according to the photographing speed. When the imaging speed is low, it is difficult to specify a corresponding portion in the next frame image of a certain frame image, and thus the block area is enlarged. On the other hand, in the case of a fast photographing speed, since the number of frame images per unit time is large, even if the block area is small, it is possible to follow. In addition, the size of the block area may be calculated based on selecting a certain time in the period of cardiac activity. Wherein it may be necessary to correct for deviations in the myocardial region. At this time, the heart activity is discriminated, the relative position of the outline of the heart is grasped, and the relative evaluation is performed based on the activity. If the block area is too small, flickering of the image may occur. To prevent this, the block area needs to have a certain size.

[ Creation of Block region ]

Next, a method of dividing the myocardium into a plurality of block regions will be described. Fig. 2B and 2C are diagrams illustrating a method of radially dividing the myocardium from the center of the heart. The region of the heart should be the region as follows: the relative position of the outline of the heart is grasped by discriminating the positional relationship between the heart activity and the blood vessel, and the relative evaluation is performed based on the activity. Therefore, in the present application, after the heart contour is automatically detected, the myocardial region is divided into a plurality of block regions, and the change value (pixel value) of the image included in each block region is averaged. As a result, even if the morphology of the heart changes with the passage of time, the change in the region of interest with time can be tracked.

On the other hand, if the heart region is divided into block regions and not specified, the region of interest deviates from the heart region due to temporal changes in the heart, and becomes a meaningless image. In addition, there are methods of inputting heart rate or frequency bands. It should be noted that these methods may also be applied to three-dimensional stereoscopic images. By making the pixels in the 3D stereoscopic image constant, the calculation of region segmentation can also be performed for 3D. The relative evaluation of the motion based on such relative positions may be performed between adjacent frame images, or may be performed in integer multiples of 2 or 3. Alternatively, a plurality of sheets may be combined into one group, and each group may be processed.

Fig. 7A is a schematic diagram showing a human left lung from the front, and fig. 7B is a schematic diagram showing a human left lung from the left side. Fig. 7A and 7B both show the lungs in an inhaled, i.e. inhaled, breath state. Fig. 8A is a schematic diagram showing a human left lung from the front, and fig. 8B is a schematic diagram showing a human left lung from the left side. Fig. 8A and 8B each show the lungs in an expired or spitted state. As shown in these figures, the morphology of the lung fields changed greatly during respiration, but the change rate of the lung fields on the diaphragm side was large, and the change rate of the lung fields on the opposite side of the diaphragm was small. In the present invention, the position of each region in the lung field is changed according to the change rate. Thus, relative evaluation can be performed based on the relative positional relationship of each region in the lung field region. The relative positional relationship may be expressed by a constant change rate in the lung field region based on the change rate (for example, average change rate) of the lung field, and the change rate may be adaptively changed in the lung field region according to the distance from the diaphragm. Thus, by using the variability in the lung field region, an image synchronized with the respiratory cycle can be displayed.

Fig. 9A is a schematic diagram showing a human left lung from the front, and fig. 9B is a schematic diagram showing a human left lung from the left side. Fig. 9A and 9B both show the lungs in an inhaled, i.e. inhaled, breath state. Fig. 10A is a schematic diagram showing a human left lung from the front, and fig. 10B is a schematic diagram showing a human left lung from the left side. Fig. 10A and 10B each show the lungs in an expired or spitted state. For example, as shown in fig. 9A and 9B, in the inhalation state, a mark P1 is drawn somewhere in the lung field region. If the mark P1 is a fixed point determined by two-dimensional coordinates, the coordinates do not change even in the exhalation state, and thus the mark P1 exists at the same position as shown in fig. 10A and 10B. On the other hand, in the present invention, as described above, in order to evaluate the relative positional relationship of the entire lung field region, the position of the marker P2 is moved instead of the position of the marker P1 in the exhalation state. The point plotted at the time of inhalation and the point at which the movement destination is determined at the time of exhalation may also be evaluated using the vector at this time.

In the case of dividing a region, voronoi (Voronoi) division (taylon division) may be applied. As shown in fig. 2D, the Voronoi (Voronoi) division is a "method of drawing a perpendicular bisector on a straight line connecting adjacent parent points and dividing the nearest-adjacent region of each parent point". By applying such Voronoi (Voronoi) division, the calculation time can be shortened. Further, in the case of drawing a straight line connecting adjacent parent points, weighting may be performed according to the analysis object. For example, in segmenting regions of the pulmonary artery, the weighting may be increased in the coarse places and decreased in the fine places. This reduces the processing load and allows division according to the analysis object. The classification process may be performed on the basis of an index such as a change (periodic change) in pixel value for a plurality of block areas generated by division.

After dividing the region into a plurality of block regions in this way, the change in the image in each block region is calculated based on the relative position of each block region and the dynamic part of the heart or the like. In this case, the difference of signals is acquired only in a range of the mass itself, which is a pixel, as one unit, and the difference of signals can be acquired not only in a range smaller than the mass or in a large range larger than the mass and surrounding the periphery. Further, the vertical range may be increased only in the vicinity of the diaphragm, or the horizontal range may be increased only in other dynamic regions, or the shape of the range may be deformed, or the pixel regions may be connected. It is preferable that the mass morphology is redefined in accordance with the whole lung field or heart morphology after calculating one difference or a plurality of differences. For example, after processing the first to second images, the mass pattern may be created again in the shape on the second image, and compared with the second to third images.

In the above description, the "relative positional relationship" in consideration of the movement of the organ has been described, but the present invention is not limited to this, and the image may be processed in a state where the "absolute positional relationship per pixel" is maintained in a plurality of frame images. The "absolute positional relationship of each pixel" is a relationship between pixels having coordinates specified based on a two-dimensional coordinate axis when the two-dimensional coordinate axis is defined on a frame image. That is, the method of performing the pixel processing without changing the pixel of interest is employed. The processing in the state of maintaining the "absolute positional relationship for each pixel" is premised on a plurality of frame images, but the number of frame images is not specified. The plurality of frame images are classified into groups, each of which may include an equal number of frame images, or each of which may include a different number of frame images.

In other words, the plurality of frame images are classified into a plurality of groups, and a change in pixel value of the organ, a change in distance from the center to the outer edge of the organ, or a change in volume of the organ is calculated in a state in which an absolute positional relationship of each pixel within the plurality of frame images belonging to each group is maintained. Thus, even if the pixel value is slightly changed, it can be treated as a uniform pixel value, and the data amount and the processing steps can be reduced.

In addition, a positional relationship other than the relative positional relationship and the absolute positional relationship may be considered. In other words, a point P having a specific coordinate in a specific frame image may be determined, and another point Q having a coordinate different from the specific point in the following frame image may be determined, but in this case, the size of the vector PQ is small with respect to the movement of the organ. Further, with respect to the point Q, another point R having slightly different coordinates from the point Q in the following frame is determined. By repeating this operation, other points slightly offset from the specific point P are extracted for each frame image, and the present embodiment is applied. Specifically, a plurality of frame images are acquired, pixels having different coordinates from the pixels are extracted in the following frame image with respect to each pixel within a specific frame image, and a periodic variation characterizing the state of the organ is calculated.

Then, the fourier transform characterizes a periodic change in the state of the organ, and, from the spectrum obtained after the fourier transform, a spectrum within a certain frequency band including a spectrum corresponding to the frequency of the movement of the organ is extracted, the spectrum within the extracted certain frequency band is subjected to the inverse fourier transform, and each image after the inverse fourier transform is output. In addition, it is also possible to extract pixels that change corresponding to the frequency of the periodic change characterizing the organ state in the image with a digital filter, and output an image containing the pixels extracted with the digital filter. Alternatively, a color corresponding to the periodic rate of change characterizing the organ state may be selected and the selected color may be appended to the rate of change of the pixel values to display the image on the display. Thus, the activity of the organ can be expressed.

Next, artifacts are eliminated and image data is interpolated. In other words, noise reduction using a noise cut filter is preferable because noise appears if bones or the like are included in the analysis range. In an x-ray image, since air is usually set to-1000 and bone is usually set to 1000, a portion having high transmittance has a low pixel value, and is displayed in black, and a portion having low transmittance has a high pixel value, and is displayed in white. For example, in the case of representing a pixel value in 256 gradations, black is 0 and white is 255.

In the heart region, since X-rays hardly penetrate through a position where a blood vessel or a bone exists, pixel values of an X-ray image become high, and the X-ray image becomes white. The same can be said for other CT and MRI. Wherein based on the waveform of each individual heart rate, using the same phase value to interpolate data, artifacts can be excluded from the results of the above described cardiac functional analysis. When "coordinates are different", "extreme fluctuation of pixel values", "abnormal high frequency or density (density)" is detected, those are truncated, and the image obtained by the rest can be used for Hz calculation of heart activity and adjustment of myocardial region by discriminating the form of continuous and smooth waves by using the least square method or the like, for example. In addition, in the case of overlapping images, there are methods of (1) overlapping the acquisition comparative images of the images on the front and rear acquisition sides as they are in terms of their coordinates, (2) after the images on the front and rear sides are acquired in the base, expanding the images relatively, and overlapping the relative position information thereof in the base. By the above method, the morphology of the heart region or the change of the image in the block region can be corrected.

Wherein the "reconstruction (reconstruction)" in the time axis is described. For example, in the case where the suction time of 15f/s is 2 seconds, 30+1 sheets of images can be obtained. In this case, every 10% of the "reconstruction (reconstruction)" can be performed by simply overlapping three sheets. At this time, for example, in the case where 0.1 seconds is 10% and images thereof take only photographs of 0.07 seconds and 0.12 seconds, "reconstruction (reconstruction)" of 0.1 seconds is required. In that case, a median value (average of both) of about 10% of the image is given for "reconstruction (reconstruction)". In addition, it may be captured on a time axis, and the coefficient may be changed at a ratio of the time. For example, when there is a difference in the time axis, there is no photographing value of 0.1 seconds, and photographing times are 0.07 seconds and 0.12 seconds, "reconstruction (reconstruction)" can be performed by recalculating "(the value of 0.07 seconds) ×2/5+ (the value of 0.12 seconds) ×3/5"). It is preferable that the calculation be performed with a thickness of "reconstruction (reconstruction)" of 10% to 20% or "reconstruction (reconstruction)" of 10% to 40%, which is 0 to 100% of the "maximum differential intensity projection (MaximumDifferential Intensity Projection)". Thus, the portion not photographed may also be subjected to a "reconstruction (reconstruction)" at a single heart rate scale. The present invention may also be used to "reconstruct (reconstruction)" the heart, blood flow, and other series of activities linked to these.

[ Fourier analysis ]

Based on the period of the heart activity and the period of the vascular beating analyzed as described above, fourier analysis is performed on the value of density/intensity (density) in each block region and the amount of change thereof. Fig. 3A is a graph showing the intensity (intensity) change of a specific block and the result of fourier analysis thereof. Fig. 3B is a graph showing the fourier transform result of extracting the frequency component close to the heart rate and the intensity (intensity) change of the frequency component close to the heart rate by inversely transforming the fourier transform result. For example, when fourier transform (fourier analysis) is performed on the intensity (intensity) change of a specific block, the result shown in fig. 3A is obtained. When a frequency component close to the heart rate is extracted from the frequency components shown in fig. 3A, the result shown on the right side of the paper surface of fig. 3B is obtained. By performing inverse fourier transform on the result, as shown on the left side of the paper surface of fig. 3B, an intensity (intensity) change tuned to a change in heart rate can be obtained.

Wherein when performing an inverse fourier transform on a spectrum consisting of frequency components, both frequency elements (heart rate, cardiovascular beat frequency) specified by the density (density) of heart rate or blood flow and spectrum bands (BPF may be used) are combined, or an inverse fourier transform may be performed based on any of them.

In addition, when fourier transform is performed, an AR method (Autoregressive Moving average model: auto-regressive moving average model) may be used so that calculation can be performed in a short time. The AR method has a method using You Er-Wobbe equation (Yule-walker equiation) or Kalman filter in an autoregressive moving average model, and the You Er-Wobbe estimated value (Yule-WALKER ESTIMATES) derived therefrom, the PARCOR method, and the least square method can be used to supplement the calculation. Thus, near real-time images can be taken faster to assist in calculating or correcting artifacts (artifacts). By such fourier analysis, the attribute of the image in each block region can be extracted and displayed.

Wherein the change of the image of each block region in each frame image can be fourier-transformed, and a spectrum within a certain frequency band including a spectrum corresponding to the period of the heart activity can be extracted from the spectrum obtained after the fourier transform. Fig. 3C is a diagram showing an example in which a certain frequency band is extracted from a spectrum obtained after fourier transform. The spectral frequency f of the synthesized wave is established in a relationship of "1/f=1/f1+1/f 2" between the frequencies f1 (heart rate component) and f2 (pathological blood flow component) which are the synthesized elements, and the following method can be adopted in extracting the spectrum.

(1) The portion with the high heart rate spectrum ratio is extracted.

(2) The spectrum is extracted by dividing between the peak of the spectrum corresponding to the heart rate/pathological blood flow and the peaks of a plurality of synthetic waves in the vicinity thereof.

(3) Spectral peaks corresponding to heart rate/pathological blood flow and trough portions of the spectra of a plurality of synthetic waves in the vicinity thereof are divided to extract spectra.

(4) A spectrum included in a certain frequency bandwidth may be extracted from a heart rate component (blood flow component). In this case, a plurality of spectra are obtained with overlapping spectra, but each spectrum can be recovered by separating the components.

As described above, in the present application, a fixed BPF is not used, but a spectrum within a certain frequency band including a spectrum corresponding to a period of heart activity is extracted. Further, in the present application, from the spectrum obtained after fourier transform, a spectrum within a certain frequency band including a spectrum (for example, a spectral model) corresponding to a frequency other than the heart activity obtained from the frame image (for example, density/intensity of each site), a heart rate element obtained from heart rate or vascular beat, or a frequency input from the outside by an operator may be extracted.

The spectral components of the composite wave are 50% +50% if they are only two components (heart rate, pathological blood flow), and are distributed 1/3 in the case of three components. Thus, the spectrum of the composite wave can be calculated to some extent from the percentage of the spectrum of the heart rate component, the percentage of the spectrum of the pathological blood flow component, the components of the spectrum and the height thereof. Spectra can be extracted at a high point of this ratio (%). In other words, the ratio between the pathological blood flow component/heart rate component and the synthetic wave component is calculated, and the high spectral value of the pathological blood flow component/heart rate component is calculated and extracted. In the case of the discrimination of the diaphragm, only a portion having a relatively constant Hz (frequency), in other words, a spectrum corresponding to a region having a small change in Hz, or a superposition thereof may be extracted from data (data) of the heart rate or the cardiovascular frequency. In addition, when determining the spectral band, the spectral band is also determined in a range (range) where HZ changes and its surrounding area. By the above, not only in the case of a perfect match with the cycle of heart activity or the cycle of vascular beating, but also a spectrum which is considered to be good can be extracted, and image diagnosis can be facilitated.

It is to be noted that it is well known that "heart rate" or "respiration" is included in a specific frequency band. Therefore, for example, "0 to 0.5Hz (respiratory rate of 0 to 30 times/min)" is used in the case of respiration, and for example, a filter of "0.6 to 2.5 (heart rate/pulse rate of 36 to 150 times/min)" is used in the case of a circulator, and the respiratory rate or the frequency of the circulator may be determined beforehand by the filter. This enables a frequency-tunable image to be displayed. This is because when the density (density) of the heart is changed, the density (density) of the breath (lung) may be changed, or when the density (density) of the lung is changed, the density (density) of the heart may be changed.

[ Waveform analysis ]

Waveform analysis is performed on heart, blood vessels, brain waves, and others identified as certain waveforms in the examination. Including repetitive movements of the foot under certain conditions. Further, the Hz of the repeated operation is superimposed, and whether the same trend exists is analyzed. Waveform data are compared and the matching rate of the two data is calculated. The fourier analyzed data are then compared.

[ Digital Filter ]

It should be noted that a digital filter may be used instead of the fourier analysis described above. The digital filter is a filter as follows: in order to adjust the frequency components of the signal, a transformation between the time region and the frequency region is performed based on a mathematical algorithm "fast fourier transform and inverse fast fourier transform" that extracts the frequency components of the signal. Thus, the same effects as those of the fourier analysis described above can be obtained.

[ Vision/quantization ]

The results analyzed as described above are visualized and quantified. As standard absorption

(Standard uptake) it is possible to average 1 from the measured density/intensity of the whole heart region and display the values relatively/logarithmically. In addition, since only the blood flow direction is used, it is possible to cut out a change in a specific direction. Thus, only data of a meaningful method can be extracted. Using the discrimination result of the heart region, pseudo-colorization is performed following the change in the analysis range. In other words, the analysis result of each individual (subject) is applied to the relative region along a specific form (minimum, maximum, average, median) of the matching phase. Further, the shape and phase of the sample are changed to a specific shape and phase that can compare a plurality of analysis results.

Further, when creating a "standard heart", the relative positional relationship in the heart (myocardium) is calculated using the analysis result of the above-described heart activity. The "standard heart" is created using lines that integrate the outer contour lines, density (density), and the like of the hearts of a plurality of patients. It should be noted that the idea can be applied not only to the heart but also to the lungs (standard lungs) or other organs (standard organs). For example, an "organ model" may be created by different ages, sexes, countries, and degrees of disease.

In addition to the above-described change in the cardiac pixel value, a change in the distance from the center of the heart to the myocardium (distance L shown in fig. 2B and 2C) may be calculated and fourier analysis may be performed, and further, a change in the volume of the heart may be calculated and fourier analysis may be performed.

As described above, when a "standard heart" can be created, tunability, matching rate, and mismatch rate (display of frequency-tunability image) can be quantified and prompted. In addition, a deviation from the normal state may be displayed. According to the present embodiment, by performing fourier analysis, it is possible to perform a possibility of finding a new disease, a comparison with itself in general, a comparison with the hand and the foot, or a comparison with the hand and the foot on the other side. Furthermore, by quantifying the tunability, it is possible to grasp how the foot is moving, where the swallowing, and the like are abnormal. In addition, it is possible to judge whether or not a person in a diseased state has changed after a certain period of time has elapsed, and in the case of a change, the before-change and after-change are compared.

[ Description of the heart ]

In the present specification, the outline of the heart is temporarily drawn using a combination of a bezier curve and a straight line, and a method of adjusting the heart is adopted to improve the anastomosis. For example, if the heart contour is represented by four bezier curves and one straight line, the heart contour can be drawn by finding 5 points on the heart contour and 4 points of the control points. By shifting the positions of the points, drawing a plurality of cardiac contours, and evaluating the coincidence using the conditions such as "maximizing the total value of the density (density) in the contours" and "maximizing the difference between the total densities (densities) of several pixels inside and outside the contour lines", the contours of the heart can be detected with high accuracy. The control point position of the bezier curve can be adjusted by a least square method or the like by extracting points near the outer edge by classical binarization extraction profiles. The above method is not limited to the heart, and may be applied to other organs as "detection of organs". In addition, it is applicable not only to a planar image but also to a stereoscopic image (3D image). By defining a surface equation and setting control points thereof, a target object surrounded by a plurality of surfaces can be estimated as an organ.

[ Analysis of cardiac Functions Using Fourier analysis ]

Next, a cardiac function analysis using fourier analysis according to the present embodiment will be described. Fig. 4 is a flowchart showing an overview of cardiac function analysis of the present embodiment. The base module 1 extracts an image of DICOM from the database 15 (step S1). Wherein at least a plurality of frame images contained in a single heart rate are acquired. Next, in each acquired frame image, a period of cardiac activity is specified using a change in pixel value, for example, a change in density (density) in a certain region having at least the myocardium (step S2). Next, a heart (myocardium) region is detected (step S3), and the detected myocardium is divided into a plurality of block regions (step S4). Among them, as described above, the myocardium is segmented radially from the heart center using Voronoi (Voronoi) segmentation (taisen segmentation). Then, a change in pixel value of each block region in each frame image is calculated (step S5). Wherein the variation value in each block area is averaged and expressed as one data.

It is to be noted that the display pixels may be blurred to some extent, and the whole may be displayed in a blurred state. In particular, in the case of a blood vessel, a low signal is mixed between high signal values, but if only a high signal value can be roughly grasped, it is irrelevant that the whole is blurred. For example, in the case of blood flow, only signals above a threshold may be extracted. Specifically, when the numbers in the following table are taken as one pixel, the ratio of the central value is taken, and when averaging is performed in one pixel, the average value can be smoothly expressed between adjacent pixels.

[ Table 1]

Note that, for the change value in each block region, noise reduction by truncation may be performed. Next, with respect to the value of density/intensity (density/intensity) and the amount of change thereof in each block region, fourier analysis is performed based on the cycle of the above-described heart activity (step S6). Thereby, the attribute of the image in each block area can be extracted and displayed.

Wherein, a spectrum within a certain frequency band including a spectrum corresponding to the period of the heart can be extracted from the spectrum obtained after the fourier transform. The frequency f of the spectrum of the synthesized wave is established in the relation of "1/f=1/f1+1/f 2" between the frequencies f1 and f2 which are the synthesized elements, and the following method can be adopted in extracting the spectrum.

(1) The high spectral ratio portions of cardiac activity are extracted.

(2) The spectrum is divided between a peak of the spectrum corresponding to the heart rate/blood flow and a peak of a plurality of synthetic waves in the vicinity thereof, and the spectrum is extracted.

(3) The spectral peaks corresponding to the heart rate/blood flow and the trough portions of the spectra of the plurality of synthetic waves in the vicinity thereof are divided, and the spectra are extracted.

The spectral components of the composite wave are 50% +50% if they are only two components (heart rate and blood flow), and are distributed 1/3 in the case of three components. Therefore, from the percentage of the spectrum of the heart rate component, the percentage of the spectrum of the blood flow component, the components of the spectrum and the height thereof, the spectrum of the composite wave can be calculated to some extent. Spectra can be extracted at a high point of this ratio (%). In other words, the ratio between the blood flow component/heart rate component and the synthetic wave component is calculated, and the spectral value of the blood flow component/heart rate component high is calculated and extracted.

Next, regarding the result obtained by fourier analysis, noise reduction is performed (step S7). Wherein truncation or artifact (artifact) removal as described above may be performed. The operations of the above steps S5 to S7 are performed more than once, and it is judged whether or not they are completed (step S8). In the case of incompletion, the process proceeds to step S5, and in the case of completion, the result obtained by fourier analysis is displayed as a pseudo-color image on the display (step S9). Note that a black-and-white image may be displayed. Thus, by repeating a plurality of cycles, it is possible to improve the accuracy of the data. Thus, a desired video can be displayed. In addition, the desired video may also be obtained by modifying the image displayed on the display.

In addition to the myocardial segmentation processing in steps S4 and S5 described above, a change in the distance from the center of the heart to the myocardium may be calculated and fourier analysis may be performed instead of steps S4 and S5. In addition, instead of steps S4, S5, the change in the volume of the heart may be calculated and fourier analysis performed.

[ Analysis of cardiac function Using digital Filter ]

Next, a functional analysis using the digital filter according to the present embodiment will be described. Fig. 5 is a flowchart showing an outline of the cardiac function analysis of the present embodiment. Steps S1 to S5 and steps S7 to S9 are the same as the above-described "cardiac function analysis using fourier analysis", and therefore, the description thereof is omitted. In step T1 of fig. 5, digital filter processing is performed (step T1). The digital filter is a filter as follows: in order to adjust the frequency components of the signal, a transformation between the time region and the frequency region is performed based on a mathematical algorithm "fast fourier transform and inverse fast fourier transform" that extracts the frequency components of the signal. Thus, the same effects as those of the fourier analysis described above can be obtained.

[ Cardiac function analysis Using tuned match Rate ]

Next, a cardiac function analysis using the tuning matching rate of the present embodiment will be described. Fig. 6 is a flowchart showing an outline of the cardiac function analysis of the present embodiment. Steps S1 to S5 and steps S7 to S9 are the same as the above-described "cardiac function analysis using fourier analysis", and therefore, the description thereof is omitted. In step R1 of fig. 6, analysis of the tuning matching rate is performed (step R1). In other words, after detecting a heart (myocardium) region (step S3) and dividing the myocardium into a plurality of block regions (step S4), an average density (pixel value x) of the block regions in each frame image is calculated, and a ratio (x') of average pixel values of the block regions in each frame image with respect to a variation range (0% to 100%) of a minimum value to a maximum value of the average density (pixel value x) is calculated (step S5). On the other hand, a ratio (x '/y ') of a ratio (y ') of heart change (y) to each frame image of minimum to maximum change amplitude (0% to 100%) with respect to the heart surface area (or volume) is calculated (step S5). Using these, only block areas within a certain range of the predetermined ratio (x '/y') can be extracted (step R1).

Where y '=x' or y=ax (a is a coefficient of the value of the amplitude or the value of the density (density) of the heart surface area or volume), is a perfect match. However, not only the case of a perfect match is a meaningful value, but a value having a certain width should be extracted. Thus, in one aspect of the present invention, a log (log) is used and a certain width is determined as follows. In other words, when calculated at the ratio (%) of y=x, the tuned perfect match is "log y '/x' =0". When the range of the tuning matching rate is narrow (mathematically narrow), for example, the range is determined to be "log y '/x' = -0.05 to +0.05" in the range close to 0, and the range of the tuning matching rate is wide (mathematically wide), for example, the range is determined to be "log y '/x' = -0.5 to +0.5" in the range close to 0, and it can be said that the narrower the range is, the higher the matching value in the range is. When the ratio is obtained for each pixel and the number is counted, the distribution of the peaks is obtained with perfect matching in the case of the heart of a healthy person. In contrast, in the case of a person with a disease, the distribution of the ratio collapses. As described above, the method of determining the width using logarithms is merely an example, and the present invention is not limited thereto. In other words, the present invention is an invention of "image extraction" as (average of density (density) changes in a certain approximate range)/(heart rate)/(heart change)/(electrocardiogram)/(surface area and volume change of heart), and can be applied to methods other than the logarithmic method.

In the case of taking into consideration 3D, by measuring the heart rate, the surface area or volume of the heart, the cardiac output, and the central blood flow rate on other devices, it is possible to measure "partial heart surface area", "partial heart volume", "blood flow rate" from these ratios in the respective regions. As these quantitative measurements, in the case where the heart surface area, heart volume, cardiac output, and central blood flow can be measured in other modes (modality) or the like, the estimated functional amount can be estimated by the component of one frame or the ratio thereof, the region change amount ratio. In other words, in the case of cardiac function analysis, the heart volume can be estimated from the heart activity, and in the case of blood flow analysis, the lung blood flow can be estimated from the cardiac output, and the estimated blood flow (ratio) in the drawn branch vessel can be estimated from the center side blood flow (ratio).

In addition, as described above, if the acquired database (database) can calculate all the contents, a higher-precision judgment can be made, but it may take time to perform computer analysis. Therefore, only a certain number of sheets (phase) may be extracted and calculated. For example, the image is not automatically acquired from the front end of the acquired image, but is automatically acquired (may be manually acquired) in the latter half, in other words, not the last image, but with the image centered on the rear side of the middle. Thus, by cutting off the photographing in the tension state which occurs for the first time when photographing the patient, it is easy to extract a more stable image. Instead of directly calculating the acquired image (for example, 300 images), the turning point of "heart activity" may be selected at the first measured heart position or the like, and then calculated. Thus, when it is necessary to repeat video identification or the like, image identification such as continuous heartbeat can be performed. Here too, it can be calculated by means of image identification. In identifying the heart (myocardium) region, it is preferable to use a noise cut-off and a least squares correction chart even if only a part of the shape is manually changed or only a part of the graphic mark is manually changed.

As described above, according to the present embodiment, an image of a human body can be evaluated with an X-ray video apparatus. If digital data can be obtained, the calculation can be performed substantially well in existing facilities and facilities, and the introduction cost is low. For example, in an X-ray video apparatus using a flat panel detector (FLATPANEL DETECTOR), inspection of an object can be simply completed. In cardiac functional analysis, myocardial infarction can be screened. For example, in an X-ray video apparatus using a flat panel detector (FLATPANEL DETECTOR), by executing the diagnosis support program of the present embodiment before performing CT, unnecessary examination can be excluded. In addition, since the examination is simple, a disease with high urgency can be found early and dealt with priority. In the conventional imaging method, although there are several problems in other modalities (modality) such as CT and MRI, if this is solved, detailed diagnosis for each region can be performed.

In addition, the method can be also applied to screening of various blood vessels such as neck blood vessel stenosis, and can also be applied to blood flow evaluation or screening of large blood vessels. In addition, the method can be applied to grasp the state before and after the operation. Further, fourier analysis is performed on the cycle of the heart activity and the blood flow cycle, and abnormal changes in the residual biological motion such as ileus can be observed by removing the waveform of the heart activity and the blood flow waveform in the X-ray image of the abdomen.

In the case where the first image to be acquired is high definition to some extent, the calculation time may be required due to the large number of pixels. In this case, the calculation may be performed after reducing the image to a certain number of pixels. For example, by actually converting pixels of "4000×4000" into pixels of "1028×1028", the calculation time can be reduced.

In addition, conventionally, the width of the contrast of the target image to be judged is manually adjusted. Or a method of relatively drawing the target image to be judged based on a certain criterion is adopted. But is not strictly performed with the identification of the framework of the analysis object (e.g., lung fields). By performing the detection of the heart region and the filtering (filtering) of the bone density (the portion with low permeability cut off a certain range) in the present invention, the internal permeability width in the remaining region will be strictly limited to the width of the identified heart region. Thus, when detecting a heart region, the adjustment of the permeability that is easy to judge becomes more strict when a judgment person such as a doctor or an engineer observes. In addition, the adjustment of the transmittance for evaluating the color (color) becomes more strict.

In addition, in XP or CT, the transparency of X-rays is changed in accordance with the physical state of the subject to be irradiated, and the transparency may be changed depending on the movement of the lung at the time of photographing. In addition, even MRI photographs with a certain non-uniform signal due to a change in magnetic field in a specific direction or the like. For those, the change in "density/intensity" of a particular organ can be measured more precisely by matching the calculation of restoring the transparency of the change in transparency from the surrounding "background" to a certain form with the transparency of the particular organ, or by correcting the signal value for some certain correction by uniformly correcting for the non-uniformity of the overall change in magnetic field. In addition, by changing the degree of transparency by the characteristics of each organ in accordance with the numerical value change in the photographing condition and by more precisely correcting the change in the "density/intensity" of a specific organ, a more accurate change amount can be calculated.

In the present invention, although the average value, the variation, and the like of the intensity (intensity) may be calculated, the "Gamma correction" may be applied to the data (intensity: intensity) obtained from the X-ray, the CT, or the like, and the density (intensity) may not be accurately reflected. The gamma correction refers to a process for correcting the chroma or brightness of an image or the like displayed on a display. Generally, in a computer, an image is displayed on a display according to an input signal, but the brightness or chroma of the display is different according to the characteristics of the display. Therefore, in order to correct those errors, gamma correction is used. Gamma correction performs near-natural display by adjusting the relative relationship of signals and color data as they are input and output to the display. However, since the gamma-corrected image is not an original image, there is a case where a defect occurs when the image processing of the present invention is performed. Therefore, an image before gamma correction is obtained by applying "inverse gamma correction", in other words, an inverse filter corresponding to the gamma correction value, to the image after gamma correction. Thus, an image before gamma correction can be obtained, and the image processing of the present invention can be appropriately performed.

In the present invention, the image subjected to other image processing may be restored and supplied to the processing, not limited to the gamma correction. For example, the image processing performed on the image is analogically performed on the basis of the pixel value change of the region where the density (density) is constant in photographing, such as the space outside the human body, and a function for making the pixel value change constant is applied to all the pixels. Thus, an image close to the original image can be obtained, and the image processing of the present invention can be appropriately performed.

In addition, in the present invention, a single image can be restored from overlapping images. Conventionally, an "X-ray differential image technique" is known. In this technique, "past and present X-ray images" of the same patient photographed in a physical examination or the like are superimposed, and a portion that changes from past to present, that is, a portion where abnormality is assumed is emphasized. Thus, for example, cancer can be found early. As the overlapping method, there is a method as follows: when a plurality of images are captured as "first, second, third, and fourth sheets," for example, "first, second, and third sheets" are regarded as "first superimposed images," second, third, and fourth sheets "are regarded as" second superimposed images, "and" third, fourth, and fifth sheets "are regarded as" third superimposed images. In such superimposed images, the pixel value of a portion where a plurality of images are superimposed is high, and the pixel value of a portion where there is no superimposition is low. Since such superimposed images have a blurred outline of the organ, it is preferable to restore them to each original image. According to the present invention, since the outline of an organ can be specified, the original image can be restored from the superimposed image.

In addition, in the present invention, the frequency difference of each region in the organ can be imaged. In other words, in the organ in which the periodic activity is performed, the change value of each region is fourier-transformed, and the coloring distinction and the like are weighted based on the constituent ratio of the frequency components, so that the characteristics of each region can be shown. For example, the frequency component to be the peak in each region may be determined and each region may be discriminated by coloring. In addition, in each region, the proportion of each frequency component in a specific frequency band may be determined, and coloring distinction may be made for each region according to the proportion. In the specific region, even when the frequency composition ratio to be the reference is, for example, "10Hz is 50%, and 20Hz is 50%", the deviation rate from the frequency arrangement ratio to be the reference can be visualized. Thus, for example, the heart can grasp at a glance whether or not to perform a correct activity or an incorrect activity by coloring and distinguishing the spectral distribution of frequencies.

Description of the reference numerals

1 Basic Module

3 Heart function analysis unit

5 Blood flow analysis unit

7 Other blood flow analysis unit

9 Fourier analysis unit

10 Waveform analysis unit

11 Visualization/quantization unit

13 Input interface

15 Database

17 Output interface

19 Display

Claims (25)

1. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: Obtain multiple frame images; Based on each frame of image, calculating the frequency representing the state of the organ; calculating a phase difference between a waveform of a previously acquired organ model and a waveform corresponding to the calculated frequency; A signal representing the phase difference is output. 2. The diagnostic support program according to claim 1, wherein: The organ model segments the image of the organ and is determined by the average value of the pixel values in each segmented area. 3. The diagnostic support program according to claim 2, characterized in that: Using the Voronoi segmentation method, the images of the organs were segmented. 4. The diagnostic support program according to claim 2, wherein: When the organ is a lung, the lung field region is segmented based on the rate of change of the volume of the lung. 5. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: Obtain multiple frame images; Extracting pixel values of specific points contained in the organ image; Calculating the variation of the extracted pixel values over time; extracting at least one frequency signal from the change in the pixel value over time; The extracted frequency signal is output as the signal of the specific point. 6. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: Under multiple shooting conditions, images of the organ are obtained from multiple directions; extracting pixel values of specific points contained in the image of the organ; Calculating the variation of the extracted pixel values over time; extracting at least one frequency signal from the change in the pixel value over time; Outputting the extracted frequency signal as the signal of the specific point; Using the output signals at specific points, a cross-sectional view of all or a portion of the organ is created. 7. The diagnostic support program according to claim 5 or 6, characterized in that: The image of the organ is segmented into a certain area including the specific point by using the Voronoi segmentation method. 8. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: Obtaining a signal output by the diagnostic support program according to any one of claims 1 to 7; For the acquired signals, causing the artificial intelligence to learn signals representing normal organs or signals representing abnormal organs; The learning results according to the artificial intelligence are recorded. 9. The diagnostic support program according to claim 8, characterized in that: It also includes the following processing: Obtain multiple frame images; An organ is determined from the acquired frame image, the determined organ is compared with the recorded learning result, and the ratio of abnormal values is output. 10. The diagnostic support program according to claim 8, characterized in that: The method further includes a process of outputting a phase difference between a waveform representing the periodic motion of the normal organ and a waveform representing the periodic motion of the abnormal organ. 11. The diagnostic support program according to any one of claims 1 to 8, characterized in that: The method further includes a process of adding a signal of a portion having different transmittance to a signal of another portion, or subtracting a signal of a portion having different transmittance from a signal of another portion, in each of the plurality of frame images. 12. The diagnostic support program according to any one of claims 1 to 8, characterized in that: It also includes the following processing: Transform the waveform of a periodic signal into a trigonometric function; A signal representing the waveform converted into the trigonometric function is output. 13. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: A plurality of frame images are captured by obtaining a periodic signal applied to a human body; Performing Fourier transform on changes in pixel values in a specific area of each frame of image; Extracting a spectrum within a certain frequency band from the spectrum obtained after the Fourier transform, the spectrum including a spectrum corresponding to the frequency of the signal applied to the human body; Performing inverse Fourier transform on the extracted spectrum within the certain frequency band; The images after the inverse Fourier transformation are displayed on a display. 14. A diagnosis support program that analyzes an image of a human organ and displays the analysis result, characterized in that: Causes the computer to execute the following processing: A plurality of frame images are captured by obtaining a periodic signal applied to a human body; Calculating the change rate of the pixel values of the specific area of each frame image; using a tuning rate which is a ratio between a rate of change of pixel values in the specific area and a periodic rate of change of a signal applied to the human body, and extracting only an area where the tuning rate is within a predetermined range; Each image including the extracted area is displayed on a display. 15. A diagnosis support program for analyzing an image of a human organ and displaying the analysis result, characterized in that the program causes a computer to execute the following processing: Obtain multiple frame images; Based on the image frames, calculating a function representing the frequency and waveform of pulmonary blood flow; calculating a function representing a waveform of pulmonary blood pressure based on the function representing the calculated frequency and waveform of pulmonary blood flow and information representing the thickness of pulmonary blood vessels; The pulmonary blood pressure is estimated from the waveform of the calculated pulmonary blood pressure. 16. The diagnostic support program according to claim 8, characterized in that: It also includes the following processing: Obtain multiple frame images; An organ and a blood flow flowing in the organ are identified from the acquired frame image, the identified organ and blood flow are compared with the recorded learning result, and a feature value of the blood flow in the organ is output. 17. The diagnostic support program according to claim 16, characterized in that: The method further includes the following processing: outputting the characteristic amount of blood flow in the main blood vessels of the lung, or blood flow in the capillaries of the lung and the peripheral pulmonary blood vessels. 18. The diagnostic support program according to claim 16, wherein: The method further includes the following processing: comparing the activity of the lungs during breathing with the activity of the pulmonary blood flow, and outputting a feature quantity representing the linkage between the two. 19. The diagnostic support program according to claim 8, characterized in that: It also includes the following processing: Obtain multiple frame images; A lung field region is identified from the acquired frame image, a wave representing the activity of the identified lung field region is compared with a reference wave, and a feature quantity representing the linkage between the waves is output. 20. The diagnostic support program according to claim 19, wherein: The reference wave is a wave representing a respiratory cycle. 21. The diagnostic support program according to claim 1, characterized in that: It also includes the following processing: Calculate the maximum value of pixel value; The waveform is obtained based on the signal after the calculated maximum value. 22. The diagnostic support program according to claim 1, wherein: Data with periodicity is input, Fourier transform processing is performed, and filtering processing is performed to extract specific frequencies, thereby performing inverse Fourier transform processing. 23. The diagnostic support program according to claim 1, characterized in that: By superimposing a plurality of waveforms and detecting the peak of the phase in one cycle of an arbitrary waveform, the phase difference of any other waveform is calculated. 24. The diagnostic support program according to claim 1, characterized in that: The lung image is divided into a plurality of regions, and the average and distribution of the intensity value of each region are calculated, while obtaining correlation between count values in lung scintigraphy. 25. The diagnostic support program according to claim 1, wherein: A basic waveform is generated by overlapping a plurality of original waveforms obtained from each image, and a main waveform is generated based on the basic waveform. On the other hand, a waveform corresponding to an organ in each image is generated by setting the bandwidth or performing weighting on the original waveform.