CN105574882A - Lung segmentation extraction method and system based on CT image of chest cross section - Google Patents

Abstract

The present invention provides a lung segmentation and extraction method and system based on a chest cross-sectional CT image, the method comprising: acquiring a chest cross-sectional CT image; preprocessing the CT image; preprocessing the preprocessed CT image Carry out threshold segmentation; perform lung region extraction on the CT image after threshold segmentation. The present invention can realize the precise segmentation of the lung region, ensure the integrity of the segmentation of the lung parenchyma region, and avoid the problem of missed diagnosis in the subsequent diagnosis process due to the lack of edges and regions of the lung region.

Description

Method and system for lung segmentation and extraction based on chest cross-sectional CT images

technical field

The present invention relates to the technical field of medical image information, in particular to CT image processing technology of chest cross-section, specifically a lung segmentation and extraction method and system based on chest cross-sectional CT images.

Background technique

At present, lung cancer has become one of the most common malignant tumors in the world. Although the clinical multidisciplinary comprehensive treatment technology based on lung cancer has made great progress, the 5-year survival rate of most lung cancer patients is still less than 15%. The best time for treatment. Therefore, how to improve the early diagnosis rate of lung cancer is particularly important.

With the development of computer technology and medical image information technology, computer-aided diagnosis (Computer Aided Diagnosis, CAD) based on medical imaging has greatly helped to improve the correct diagnosis rate of doctors (especially doctors in primary hospitals). Most of the scientific research institutions engaged in the research of pulmonary nodule CAD system are in the United States and Japan. In the lung parenchyma extraction stage of the pulmonary nodule CAD system, Hu et al. used threshold and region growing methods to extract the lung parenchyma, then separated the left and right lungs based on the dynamic programming algorithm, and smoothed the lung wall based on the opening and closing operations based on mathematical morphology. Ukil uses the lung region segmentation method that extracts the tracheal tree to improve the segmentation accuracy of the hilum. Araiato and Sensakovic studied the importance of lung region segmentation as a computer-aided diagnosis system, and the study proved that incorrect lung region segmentation methods would cause 5%-17% nodules to be lost.

At present, there are 8 lung cancer CAD systems based on CT images certified by the US Food and Drug Administration (FDA) in the world, one of which is the ImageCheckerCTLN-1000 system of R2Technology, which was certified by the FDA in 2004. The ImageCheckerCTLN-1000 system provides a real-time, fully automatic nodule detection function for CT images with a slice thickness between 0.5mm and 3min. The nodules that can be detected are solid nodules with a diameter of more than 4min. Also includes Pulmopackage launched by Philips, GE RapidScreenDigitalLungVCAR launched by GE, etc.

Automatic segmentation of lung regions is an essential process for any lung computer-aided diagnosis system, especially for pulmonary nodule CAD systems. At present, most lung region segmentation methods are based on the large difference in gray value between the lung region and the background region. These algorithms mainly include: threshold segmentation, region growth, and connectivity marking, etc., and finally remove the Lung area islanding and filling of voids in and around the lung area.

Due to the complex background of CT images, CT images with lesions are more complex, which increases the difficulty of segmentation. Therefore, if only morphological operations are used to process the extracted lung region, it is easy to cause loss of the edge and region of the lung region, and the missing region is often the area where the lesion is located, resulting in missed diagnosis in the subsequent diagnosis process.

Therefore, how to research and develop a new solution to avoid the problem of missed diagnosis in the subsequent diagnosis process due to the loss of the edge of the lung region and the loss of the region is a technical problem to be solved urgently in this field.

Contents of the invention

In order to overcome the above-mentioned technical problems existing in the prior art, the present invention provides a lung segmentation and extraction method and system based on chest cross-sectional CT images. By preprocessing the chest cross-sectional CT images, threshold segmentation and lung Region extraction can achieve accurate segmentation of lung regions and ensure the integrity of lung parenchymal region segmentation.

One of the objects of the present invention is to provide a method for lung segmentation and extraction based on chest cross-sectional CT images, the method comprising: obtaining a chest cross-sectional CT image; preprocessing the CT image; Threshold segmentation is performed on the CT image after threshold segmentation; lung region extraction is performed on the CT image after threshold segmentation.

In a preferred embodiment of the present invention, the CT image is preprocessed using a joint denoising method of median filtering and wavelet denoising.

In a preferred embodiment of the present invention, performing threshold segmentation on the preprocessed CT image includes: determining pixels with grayscale values less than 0 from the preprocessed CT image; setting pixels with grayscale values less than 0 as 0 , to obtain the first image; determine the gray histogram of the first image; determine the valley between the two peaks in the gray histogram, as the segmentation threshold; according to the segmentation threshold for the first image Carrying out preliminary binary segmentation to obtain a second image; inverting the second image to obtain a third image; using a morphological opening operation to remove the third image to obtain a binary image; The connected regions in the binary image of are marked according to the strategy from left to right and from top to bottom, and the CT image after threshold segmentation is obtained.

In a preferred embodiment of the present invention, performing lung region extraction on the CT image after threshold segmentation includes: extracting connected regions with region labels from the CT images after threshold segmentation; Whether the column of the pixel is greater than 20; when it is judged to be yes, whether the column of the rightmost pixel of the connected region with the region label 2 is less than 290 columns; when judged to be yes, whether the area of the connected region with the region label 2 is greater than 2000; when the judgment is yes, the connected region with the region number 2 is the left lung region.

In a preferred embodiment of the present invention, extracting the lung region from the thresholded CT image further includes: selecting a connected region with an area greater than 2000 from the connected regions whose region labels are not 1 and 2, that is, the right lung region.

In a preferred embodiment of the present invention, performing lung region extraction on the CT image after threshold segmentation includes: extracting connected regions with region labels from the CT images after threshold segmentation; Whether the column where the pixel is located is greater than 20; when the judgment is yes, determine whether the area of the connected region with the region number 2 is greater than 2000; Row; when the judgment is no, the connected area labeled 2 in the area is the left and right lung adhesion lung area.

In a preferred embodiment of the present invention, performing lung region extraction on the CT image after threshold segmentation includes: extracting connected regions with region labels from the CT images after threshold segmentation; Whether the column where the pixel is located is greater than 20; when it is judged to be yes, whether the column where the rightmost pixel of the connected region whose area label is 2 is located is less than 290 columns; when it is judged to be no, determine that the area label is Whether the connected region of 3 is the left lung region.

In a preferred embodiment of the present invention, the extraction of the lung region from the CT image after threshold segmentation includes: extracting the connected region after the region label from the CT image after threshold segmentation; Whether the column where the pixel is located is less than 290 columns; when it is judged to be yes, whether the column where the leftmost pixel of the connected region whose area label is 2 is located is greater than 20; when it is judged to be no, determine that the area label is Whether the connected region of 3 is the left lung region.

In a preferred embodiment of the present invention, determining whether the connected region labeled 3 is the left lung region according to the CT image includes: judging whether the column of the leftmost pixel of the connected region labeled 2 is less than 2; When the judgment is yes, it is judged whether the column of the rightmost pixel of the connected region with the region number 3 is less than 290; when the judgment is yes, whether the area of the connected region with the region number 3 is greater than 2000; The connected region labeled 3 in the above region is the left lung region.

In a preferred embodiment of the present invention, determining whether the connected region labeled 3 is the left lung according to the CT image further includes: selecting from the connected regions not labeled 1, 2, 3 and 4 The connected region with an area greater than 2000 is the right lung region.

In a preferred embodiment of the present invention, performing lung region extraction on the CT image after threshold segmentation includes: extracting connected regions with region labels from the CT image after threshold segmentation; determining a region set from the connected regions, The set of regions includes a plurality of regions, and each region satisfies that the column of the leftmost pixel is less than 2 and the area is greater than 100,000; the region with the largest region label is determined from the set of regions, which is called the starting region; from the connected A suspected area is screened out from the area, and the area label of the suspected area is greater than the area label of the starting area; the target area is determined from the suspected area, and the area of the target area is greater than 2000, and the target area is the lung area.

One of the objects of the present invention is to provide a system for lung segmentation and extraction based on chest cross-sectional CT images, the system includes a CT image acquisition device for obtaining CT images of chest cross-sections; a preprocessing device for The method is to preprocess the CT image; the threshold segmentation device is used to perform threshold segmentation on the preprocessed CT image; the lung region extraction device is used to perform lung region extraction on the thresholded CT image.

In a preferred embodiment of the present invention, the CT image is preprocessed using a joint denoising method of median filtering and wavelet denoising.

In a preferred embodiment of the present invention, the threshold segmentation device includes: a determination module, configured to determine pixels with a gray value less than 0 from the preprocessed CT image; a first image determination module, configured to convert the gray Pixels with a degree value less than 0 are set to 0 to obtain the first image; the histogram determination module is used to determine the gray histogram of the first image; the segmentation threshold determination module is used to determine the two grayscale histograms in the gray histogram The valley between the peaks is regarded as a segmentation threshold; the second image determination module is used to perform preliminary binary segmentation on the first image according to the segmentation threshold to obtain a second image; the third image determination module is used to Invert the second image to obtain a third image; a binary image determination module is used to remove the third image by using a morphological opening operation to obtain a binary image; a region label module is used to The connected regions in the binary image are labeled according to the strategy from left to right and from top to bottom to obtain the CT image after threshold segmentation.

In a preferred embodiment of the present invention, the lung region extraction device includes: a connected region extraction module, used to extract connected regions with region labels from the CT image after threshold segmentation; a first judging module, used to judge Whether the column where the leftmost pixel of the connected region whose region label is 2 is greater than 20; the second judging module is used to judge whether the rightmost pixel of the connected region whose region is marked as 2 is located when the first judging module is judged to be yes. Whether the column is less than 290 columns; the third judging module is used to determine whether the area of the connected region whose area label is 2 is greater than 2000 when the second judging module judges as yes; the first lung determining module is used when When the third judging module judges yes, it determines that the connected region with the region number 2 is the left lung region.

The threshold segmentation device further includes: a second lung determination module, configured to select a connected region with an area greater than 2000 from the connected regions whose region numbers are not 1 and 2, namely the right lung region.

In a preferred embodiment of the present invention, the threshold segmentation device further includes: a fourth judging module, used for judging whether the area of the connected region labeled 2 is larger than 2000; the fifth judging module, used to determine whether the column of the rightmost pixel of the connected region whose region label is 2 is less than 290 columns when the fourth judging module judges as yes; the third lung determining module, used when When the fourth judging module judges no, it determines that the connected region with the region number 2 is the region of left and right lung adhesions.

In a preferred embodiment of the present invention, the lung region extraction device includes: a sixth judging module, used to judge whether the column of the rightmost pixel of the connected region whose region number is 2 is less than 290 columns; a seventh judging module, using When the sixth judging module judges yes, it judges whether the column of the leftmost pixel of the connected region labeled 2 is greater than 20; when the seventh judging module judges no, the lung region extraction device It also includes: a fourth lung determining module, configured to determine whether the connected region labeled 3 is the left lung region according to the CT image.

In a preferred embodiment of the present invention, when the second judging module judges no, the lung region extraction device further includes: a fourth lung determining module, configured to determine the region according to the CT image Whether the connected region labeled 3 is the left lung region.

In a preferred embodiment of the present invention, the fourth lung determination module includes: a first judging unit for judging whether the column of the leftmost pixel of the connected region with the region number 2 is less than 2; a second judging unit for When the first judging unit judges yes, judging whether the column of the rightmost pixel of the connected region labeled 3 is less than 290; the third judging unit is used to, when the second judging unit judges yes, Judging whether the area of the connected region marked as 3 is greater than 2000; the first lung determining unit is used to determine that the connected region marked as 3 is the left lung region when the third judging unit judges as yes .

In a preferred embodiment of the present invention, the fourth lung determination module further includes: a second determination unit, configured to select an area larger than 2000 from connected areas whose area numbers are not 1, 2, 3, and 4; The connected area of is the right lung area. In a preferred embodiment of the present invention, the lung region extraction device includes: a region prediction and determination module, configured to determine a region set from the connected regions, the region set includes a plurality of regions, each region All satisfying that the column where the leftmost pixel is located is less than 2 and the area is greater than 100,000; the start area determination module is used to determine the area with the largest area label from the set of areas, which is called the start area; the suspected area determination module is used to determine from the A suspected area is screened out from the connected area, and the area label of the suspected area is greater than the area label of the starting area; the target area determination module is used to determine the target area from the suspected area, and the area of the target area is greater than 2000 , the target area is the lung area.

The beneficial effect of the present invention is that it provides a lung segmentation and extraction method and system based on chest cross-sectional CT images, by preprocessing the chest cross-sectional CT images, and then performing threshold segmentation and lung region extraction, it is possible to realize the The precise segmentation of the lung area ensures the integrity of the segmentation of the lung parenchyma, and avoids the problem of missed diagnosis in the subsequent diagnosis process due to the lack of edges and areas of the lung area.

In order to make the above and other objects, features and advantages of the present invention more comprehensible, preferred embodiments will be described in detail below together with the accompanying drawings.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a flow chart of a lung segmentation and extraction method based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Fig. 2 is the specific flowchart of step S103 in Fig. 1;

FIG. 3 is a flowchart of Embodiment 1 of step S104 in FIG. 1;

FIG. 4 is a flowchart of the second embodiment of step S104 in FIG. 1;

FIG. 5 is a flow chart of the third embodiment of step S104 in FIG. 1;

FIG. 6 is a flowchart of Embodiment 4 of step S104 in FIG. 1;

FIG. 7 is a flowchart of Embodiment 5 of step S104 in FIG. 1;

FIG. 8 is a flowchart of Embodiment 1 of step S704 in FIG. 7;

FIG. 9 is a flowchart of the second embodiment of step S704 in FIG. 7;

FIG. 10 is a flowchart of Embodiment 6 of step S104 in FIG. 1;

FIG. 11 is a structural block diagram of a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Fig. 12 is a structural block diagram of a threshold segmentation device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

13 is a structural block diagram of Embodiment 1 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

14 is a structural block diagram of Embodiment 2 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

15 is a structural block diagram of Embodiment 3 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Fig. 16 is a structural block diagram of Embodiment 4 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Fig. 17 is a structural block diagram of Embodiment 5 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Fig. 18 is a structural block diagram of Embodiment 1 of the fourth lung determination module in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention;

19 is a structural block diagram of Embodiment 2 of the fourth lung determination module in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention;

Fig. 20 is a structural block diagram of Embodiment 6 of a lung region extraction device in a lung segmentation and extraction system based on a chest cross-sectional CT image provided by an embodiment of the present invention;

Figure 21 is a first schematic diagram of the connected area labels in the specific embodiment provided by the present invention;

Fig. 22 is a second schematic diagram of the connected area labels in the specific embodiment provided by the present invention;

Fig. 23 is a schematic flow chart of lung region extraction in a specific embodiment provided by the present invention;

Figure 24 is a schematic diagram of the lung region extracted in Example 1 provided by the present invention;

Figure 25 is a schematic diagram of the lung parenchyma region in Example 1 provided by the present invention;

Figure 26 is a schematic diagram of the lung region extracted in Example 2 provided by the present invention;

Figure 27 is a schematic diagram of the lung parenchyma region in Example 2 provided by the present invention;

Figure 28 is a schematic diagram of the lung region extracted in Example 3 provided by the present invention;

Fig. 29 is a schematic diagram of the lung parenchyma region in Example 3 provided by the present invention.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In the computer-aided diagnosis of lung cancer based on CT images, the correct and complete extraction of lung regions is particularly important, which is the basis of lung nodule extraction. Aiming at the problem in the prior art that the chest cross-sectional CT image has a complex background and it is extremely difficult to correctly and completely extract the lung region, the present invention proposes a lung segmentation and extraction method and system based on the chest cross-sectional CT image.

The key terms of the present invention are firstly introduced below.

Morphology operations are image processing methods developed for binary images based on the set theory method of Mathematical Morphology. Usually, morphological image processing is performed as a form of neighborhood operation. A specially defined neighborhood is called a "structural element" (StructureElement). Specific logic operations, the result of the logic operation is the corresponding pixel of the output image, mainly including: erosion, expansion, opening operation and closing operation

Pulmonary nodules (Lungnodule), nodular lesions appear on lung CT images, solitary solid nodules, pleural nodules, adherent vascular nodules, ground glass nodules, and cavitary nodules.

Computer-aided diagnosis (CAD) refers to the use of imaging, medical image processing technology, and other possible physiological and biochemical means, combined with computer analysis and calculation, to assist in the detection of lesions and improve the accuracy of diagnosis.

Fig. 1 is the specific flowchart of a kind of lung segmentation extraction method based on chest cross-sectional CT image that the present invention proposes, as can be seen from Fig. 1, described method comprises:

S101: Acquire a CT image of a chest cross section.

In a specific embodiment, the chest cross-sectional CT image is generally in DICOM format.

S102: Preprocessing the CT image;

In a specific implementation manner, the CT image may be preprocessed to remove noise by using a joint denoising method of median filtering and wavelet denoising.

S103: Perform threshold segmentation on the preprocessed CT image. FIG. 2 is a specific flowchart of step S103.

S104: Perform lung region extraction on the CT image after threshold segmentation. In a specific implementation, the present invention can perform lung region extraction on CT images through a preset template.

Fig. 2 is the specific flowchart of step S103, as can be seen from Fig. 2, this step specifically comprises:

S201: Determine the pixels whose gray value is less than 0 from the preprocessed CT image;

S202: Set pixels whose grayscale values are less than 0 to 0 to obtain a first image.

Since the gray value range of the DICOM image is [-1024,1024], and the gray value of the human body area including the lung area is greater than 0, the gray value of less than 0 is set to 0 by fixing the threshold value 0 first.

S203: Determine the grayscale histogram of the first image;

S204: Determine a valley between two peaks in the grayscale histogram as a segmentation threshold;

S205: Perform preliminary binary segmentation on the first image according to the segmentation threshold to obtain a second image;

In a specific embodiment, an Ostu algorithm can also be used to estimate the segmentation threshold, and initially segment the lung region (the empirical threshold is 500).

S206: Invert the second image to obtain a third image;

S207: Perform a removal process on the third image by using a morphological opening operation to obtain a binary image;

S208: Label the connected regions in the binary image according to the strategy from left to right and from top to bottom to obtain a CT image after threshold segmentation.

The morphological opening operation is used to remove the smaller connected regions in the binary image, and then the image is marked with regions (the regions refer to the connected regions). In a specific embodiment, an opening operation is performed on the third image, and the structural element is a circle with a radius of 2, with the purpose of removing isolated "islands" to obtain a binary image. Fig. 21 and Fig. 22 are schematic diagrams of the labels of the connected areas in the specific embodiments provided by the present invention.

In other embodiments of the present invention, the segmentation of lung parenchyma region can also implement threshold segmentation through the probability density distribution of the double Gaussian mixture model, and then remove islands and repair defects through mathematical morphological operations.

Fig. 3 is a flow chart of the first embodiment of step S104. It can be seen from Fig. 3 that in the first embodiment, the lung region is extracted through a preset template, and this step specifically includes:

S301: Extract connected regions with region labels from the CT image after threshold segmentation. Taking FIG. 22 as an example, the extracted connected regions are the connected region with the region number 1 and the connected region with the region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S302: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000;

S303: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000;

S304: When the judgment is yes, whether the area (the number of pixels included) of the connected region whose region label is 2 is greater than 2000 is judged;

S305: When the judgment is yes, the connected region whose region number is 2 is the left lung region. Observer-referenced here, not consistent with anatomy.

In Embodiment 1, it is judged whether the connected region labeled 2 is the left lung, that is, it is judged whether the connected region satisfies condition 1 (judging whether the column of the leftmost pixel of region 2 is greater than 20 (the corresponding index is greater than 10000)), condition 2 ( Judging whether the column of the rightmost pixel in area 2 is less than 290 columns (the corresponding index is less than 150000)), condition 3 (judging whether the area of area 2 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that region 2 is the left lung region.

FIG. 4 is a flow chart of the second embodiment of step S104. It can be seen from FIG. 4 that in the second embodiment, the lung region is extracted through a preset template, and this step specifically includes:

S401: Extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S402: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000;

S403: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000;

S404: When the judgment is yes, judge whether the area (the number of pixels included) of the connected region whose region label is 2 is greater than 2000;

S405: When the judgment is yes, the connected region whose region number is 2 is the left lung region. Observer-referenced here, not consistent with anatomy.

S406: Select a connected region with an area larger than 2000 from the connected regions whose region labels are not 1 and 2, which is the right lung region.

In the second embodiment, it is firstly judged whether the connected region marked with 2 is the left lung, that is, it is judged whether the connected region satisfies condition 1, condition 2, and condition 3. If and only if the three conditions are met, it can be judged that the area marked 2 is the left lung area, and at the same time, the area marked 1 is the gap area between the human body and CT, and then search from the area marked 3 to find the area larger than 2000 If there is a connected area, it is regarded as the right lung, otherwise the tomographic image only finds one lung area.

FIG. 5 is a flow chart of the third embodiment of step S104. It can be seen from FIG. 5 that in the third embodiment, the lung region is extracted through a preset template, and this step specifically includes:

S501: Extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S502: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000. This step judges whether the connected region labeled 2 satisfies condition 1.

S503: When the judgment is yes, it is judged whether the area (the number of pixels included) of the connected region labeled 2 is greater than 2000, and this step judges whether the connected region labeled 2 satisfies the condition 3.

S504: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000, and this step judges whether the connected region labeled 2 satisfies condition 2.

S505: When the judgment is no, the connected area with the area number 2 is the left and right lung adhesion lung area.

In Embodiment 3, it is first judged whether the connected region labeled 2 satisfies condition 1, condition 3, and condition 2. If and only if the condition 2 is not satisfied, that is, the column of the rightmost pixel of the 2 area is not less than 290, it can be judged that the area labeled 2 is the lung area with the adhesion of the left and right lungs.

FIG. 6 is a flow chart of the fourth embodiment of step S104. It can be seen from FIG. 6 that in the fourth embodiment, the lung region is extracted through a preset template, and this step specifically includes:

S601: Extract connected regions with region labels from the thresholded CT image. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S602: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000. This step judges whether the connected region labeled 2 satisfies condition 1.

S603: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000, and this step judges whether the connected region labeled 2 satisfies condition 2.

S604: When the judgment is no, determine whether the connected region with the region number 3 is the left lung region according to the CT image.

In Embodiment 4, it is first judged whether the connected region labeled 2 satisfies condition 1 and condition 2. If and only if condition 1 is satisfied and condition 2 is not satisfied, determine whether the connected region labeled 3 is the left lung region according to the CT image.

FIG. 7 is a flow chart of the fifth embodiment of step S104. It can be seen from FIG. 7 that in the fifth embodiment, the lung region is extracted through a preset template, and this step specifically includes:

S701: Extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S702: Determine whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000. This step determines whether the connected region labeled 2 satisfies condition 2.

S703: When the judgment is yes, it is judged whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000. This step judges whether the connected region labeled 2 satisfies condition 1.

S704: When the judgment is no, determine whether the connected region with the region number 3 is the left lung region according to the CT image.

In Embodiment 5, it is first judged whether the connected region labeled 2 satisfies condition 1 and condition 2. If and only if condition 1 is not satisfied and condition 2 is satisfied, it is determined whether the connected region labeled 3 is the left lung region according to the CT image.

Fig. 8 is a flow chart of the first embodiment of step S604 in Fig. 6 and step S704 in Fig. 7. It can be seen from Fig. 8 that in the first embodiment, the connectivity of the area labeled 3 is determined according to the CT image. Whether the area is the left lung area specifically includes:

S801: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is less than 2, that is, whether the corresponding index is less than 1000;

S802: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns, that is, whether the corresponding index is less than 150000;

S803: When the determination is yes, determine whether the area (the number of pixels included) of the connected region whose region number is 3 is greater than 2000;

S804: When the judgment is yes, the connected region whose region number is 3 is the left lung region.

In this embodiment, judging whether the connected region labeled 3 is the left lung, that is, judging whether the connected region satisfies whether the column of the leftmost pixel is less than 2, and judging whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns (the corresponding index is less than 150000), and judge whether the area of the connected region labeled 3 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that the area 3 is the left lung area.

Fig. 9 is a flow chart of the second embodiment of step S604 in Fig. 6 and step S704 in Fig. 7. It can be seen from Fig. 9 that in this embodiment, the connectivity of the region labeled 3 is determined according to the CT image. Whether the area is the left lung area specifically includes:

S901: Determine whether the column of the leftmost pixel of the connected region whose region number is 2 is less than 2, that is, whether the corresponding index is less than 1000;

S902: When the judgment is yes, judge whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns, that is, whether the corresponding index is less than 150000;

S903: When the judgment is yes, it is judged whether the area (the number of pixels included) of the connected region whose region number is 3 is greater than 2000;

S904: When the judgment is yes, the connected region whose region number is 3 is the left lung region.

S905: Select a connected region with an area greater than 2000 from the connected regions whose region labels are not 1 and 2, which is the right lung region.

In this embodiment, judging whether the connected region labeled 3 is the left lung, that is, judging whether the connected region satisfies whether the column of the leftmost pixel is less than 2, and judging whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns (the corresponding index is less than 150000), and judge whether the area of the connected region labeled 3 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that the area marked 3 is the left lung area, the area marked 1 and the area marked 2 are the gap area between the human body and CT, and then start from the area marked 4 Search to find a connected region with an area greater than 2000, if it exists, it will be regarded as the right lung, otherwise, only one lung region will be found in the tomographic image.

Fig. 10 is a flow chart of the sixth embodiment of step S104. It can be seen from Fig. 10 that in the sixth embodiment, none of the conditions of the first embodiment to the fifth embodiment are satisfied, and the lung region is extracted through a preset template. The steps specifically include:

S1001: Extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

S1002: Determine a region set from the connected regions, the region set includes a plurality of regions, and each region satisfies that the column where the leftmost pixel is located is less than 2 and the area is greater than 100,000.

In a specific embodiment, it is judged sequentially whether the column where the leftmost pixel of each connected region is located is less than 2, and when the judgment is yes, continue to judge whether the area of the connected region is greater than 100,000, when the judgment is yes, the connected region is Available as a region in a region collection.

S1003: Determine the area with the largest area label from the set of areas, which is called the starting area;

S1004: Screen out a suspected region from the connected regions, where the region label of the suspected region is greater than the region label of the starting region.

S1005: Determine a target area from the suspected area, where the area of the target area is larger than 2000, and the target area is the lung area.

In Embodiment 6, first judge the label of the gap area between the human body and the CT, and then traverse all the connected areas sequentially, and search for the last one that satisfies the fact that the leftmost pixel of the area is located in a column less than 2 (the corresponding index is less than 1000), and the area area is greater than 100,000 area, the start area is the gap area between the human body and the CT, and this area is marked as start. Start from the start+1 region to traverse all connected regions, arrange their areas from large to small, and select all block combinations with an area greater than 2000 as lung regions.

As mentioned above, it is a method of lung segmentation and extraction based on chest cross-sectional CT images proposed by the present invention, by preprocessing the chest cross-sectional CT images, then performing threshold segmentation, and finally performing lung segmentation according to a preset template. The lung area extraction can realize the precise segmentation of the lung area and ensure the integrity of the lung parenchymal area segmentation.

Fig. 11 is a structural block diagram of a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 11 that the system includes:

The CT image acquisition device 100 is configured to acquire a CT image of a chest cross section.

In a specific embodiment, the chest cross-sectional CT image is generally in DICOM format.

A preprocessing device 200, configured to preprocess the CT image;

In a specific implementation manner, the CT image may be preprocessed to remove noise by using a joint denoising method of median filtering and wavelet denoising.

The threshold segmentation device 300 is configured to perform threshold segmentation on the preprocessed CT image. Fig. 12 is a specific structural block diagram of the threshold segmentation device.

The lung region extracting device 400 is configured to extract lung regions from CT images after threshold segmentation. In a specific implementation, the present invention can perform lung region extraction on CT images through a preset template.

FIG. 12 is a structural block diagram of a threshold segmentation device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from FIG. 12 that the threshold segmentation device 300 specifically includes:

A determination module 301, configured to determine pixels with a gray value less than 0 from the preprocessed CT image;

The first image determining module 302 is configured to set pixels whose grayscale values are less than 0 to 0 to obtain the first image.

Since the gray value range of the DICOM image is [-1024,1024], and the gray value of the human body area including the lung area is greater than 0, the gray value of less than 0 is set to 0 by fixing the threshold value 0 first.

A histogram determining module 303, configured to determine the grayscale histogram of the first image;

A segmentation threshold determination module 304, configured to determine a valley between two peaks in the grayscale histogram, which is regarded as a segmentation threshold;

The second image determination module 305 is configured to perform preliminary binary segmentation on the first image according to the segmentation threshold to obtain a second image;

In a specific embodiment, an Ostu algorithm can also be used to estimate the segmentation threshold, and initially segment the lung region (the empirical threshold is 500).

The third image determination module 306 is configured to invert the second image to obtain a third image;

The binary image determining module 307 is configured to perform removal processing on the third image by using a morphological opening operation to obtain a binary image;

The region labeling module 308 is configured to perform region labeling on the connected regions in the binary image according to a strategy from left to right and from top to bottom to obtain a CT image after threshold segmentation.

The morphological opening operation is used to remove the smaller connected regions in the binary image, and then the image is marked with regions (the regions refer to the connected regions). In a specific embodiment, an opening operation is performed on the third image, and the structural element is a circle with a radius of 2, with the purpose of removing isolated "islands" to obtain a binary image. Fig. 21 and Fig. 22 are schematic diagrams of the labels of the connected areas in the specific embodiments provided by the present invention.

In other embodiments of the present invention, the segmentation of lung parenchyma region can also implement threshold segmentation through the probability density distribution of the double Gaussian mixture model, and then remove islands and repair defects through mathematical morphological operations.

Fig. 13 is a structural block diagram of Embodiment 1 of a lung extraction device in a lung segmentation extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 13 that in Embodiment 1, lung region extraction The device 400 specifically includes:

The connected region extraction module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The first judging module 402 is used to judge whether the column where the leftmost pixel of the connected region whose region label is 2 is located is greater than 20, that is, whether the corresponding index is greater than 10000;

The second judging module 403 is used to determine whether the column where the rightmost pixel of the connected region whose region label is 2 is located is less than 290 columns, that is, whether the corresponding index is less than 150000 when the judgment of the first judging module is yes;

The third judging module 404, for when the second judging module is judged to be yes, whether the area (the number of pixels included) of the connected region whose judgment region label is 2 is greater than 2000;

The first lung determining module 405 is configured to determine that the connected region with the region number 2 is the left lung region when the third judging module judges yes. Observer-referenced here, not consistent with anatomy.

In Embodiment 1, it is judged whether the connected region labeled 2 is the left lung, that is, it is judged whether the connected region satisfies condition 1 (judging whether the column of the leftmost pixel of region 2 is greater than 20 (the corresponding index is greater than 10000)), condition 2 ( Judging whether the column of the rightmost pixel in area 2 is less than 290 columns (the corresponding index is less than 150000)), condition 3 (judging whether the area of area 2 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that region 2 is the left lung region.

Fig. 14 is a structural block diagram of the second embodiment of the lung extraction device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 14 that in the second embodiment, the threshold segmentation device Specifically include:

The connected region extraction module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The first judging module 402 is used to judge whether the column where the leftmost pixel of the connected region whose region label is 2 is located is greater than 20, that is, whether the corresponding index is greater than 10000;

The second judging module 403 is used to determine whether the column where the rightmost pixel of the connected region whose region label is 2 is located is less than 290 columns, that is, whether the corresponding index is less than 150,000 when the judgment of the first judging module is yes;

The third judging module 404, for when the second judging module is judged to be yes, whether the area (the number of pixels included) of the connected region whose judgment region label is 2 is greater than 2000;

The first lung determining module 405 is configured to determine that the connected region with the region number 2 is the left lung region when the third judging module judges yes. Observer-referenced here, not consistent with anatomy.

The second lung determination module 406 is configured to select a connected region with an area larger than 2000 from the connected regions whose region numbers are not 1 and 2, that is, the right lung region.

In the second embodiment, it is firstly judged whether the connected region marked with 2 is the left lung, that is, it is judged whether the connected region satisfies condition 1, condition 2, and condition 3. If and only if the three conditions are met, it can be judged that the area marked 2 is the left lung area, and at the same time, the area marked 1 is the gap area between the human body and CT, and then search from the area marked 3 to find the area larger than 2000 If there is a connected area, it is regarded as the right lung, otherwise the tomographic image only finds one lung area.

Fig. 15 is a structural block diagram of Embodiment 3 of a lung extraction device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 15 that in Embodiment 3, the threshold segmentation device Specifically include:

The connected region extraction module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The first judging module 402 is used for judging whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000. This step judges whether the connected region labeled 2 satisfies condition 1.

The fourth judging module 407, for when the first judging module is judged as yes, whether the area (the number of pixels contained) of the connected region whose judgment area label is 2 is greater than 2000, this step judges the connected region whose label is 2 Whether condition 3 is satisfied.

The fifth judging module 408 is used to determine whether the column where the rightmost pixel of the connected region whose region label is 2 is located is less than 290 columns, that is, whether the corresponding index is less than 150000, when the fourth judging module judges as yes, this step judges Whether the connected region labeled 2 satisfies condition 2.

The third lung determination module 409 is configured to determine that the connected area with the area number 2 is the left and right lung adhesion lung area when the fourth determination module determines no.

In Embodiment 3, it is first judged whether the connected region labeled 2 satisfies condition 1, condition 3, and condition 2. If and only if the condition 2 is not satisfied, that is, the column of the rightmost pixel of the 2 area is not less than 290, it can be judged that the area labeled 2 is the lung area with the adhesion of the left and right lungs.

Fig. 16 is a structural block diagram of Embodiment 4 of a lung extraction device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 16 that in Embodiment 4, lung region extraction The device specifically includes:

The connected region extraction module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The first judging module 402 is used for judging whether the column of the leftmost pixel of the connected region whose region number is 2 is greater than 20, that is, whether the corresponding index is greater than 10000. This step judges whether the connected region labeled 2 satisfies condition 1.

The second judging module 403 is used to determine whether the column where the rightmost pixel of the connected region whose region label is 2 is located is less than 290 columns, that is, whether the corresponding index is less than 150000, when the first judging module is judged to be yes. Determine whether the connected region labeled 2 satisfies condition 2.

When the judgment of the second judging module is no, the lung region extracting device further includes: a fourth lung determining module 410, configured to determine whether the connected region labeled 3 according to the CT image is left lung area.

In Embodiment 4, it is first judged whether the connected region labeled 2 satisfies condition 1 and condition 2. If and only if condition 1 is satisfied and condition 2 is not satisfied, determine whether the connected region labeled 3 is the left lung region according to the CT image.

Fig. 17 is a structural block diagram of Embodiment 5 of a lung extraction device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 17 that in Embodiment 5, lung region extraction The device specifically includes:

The connected region extraction module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The sixth judging module 411 is used for judging whether the column of the rightmost pixel of the connected region labeled 2 is less than 290 columns, that is, whether the corresponding index is less than 150000. This step judges whether the connected region labeled 2 satisfies condition 2.

The seventh judging module 412 is used for judging whether the column of the leftmost pixel of the connected region labeled 2 is greater than 20, that is, whether the corresponding index is greater than 10000, when the sixth judging module judges yes. This step judges whether the connected region labeled 2 satisfies condition 1.

When the judgment of the seventh judging module is no, the lung region extracting device further includes: a fourth lung determining module 410, configured to determine whether the connected region labeled 3 according to the CT image is left lung area.

In Embodiment 5, it is first judged whether the connected region labeled 2 satisfies condition 1 and condition 2. If and only if condition 1 is not satisfied and condition 2 is satisfied, it is determined whether the connected region labeled 3 is the left lung region according to the CT image.

Fig. 18 is a structural block diagram of Embodiment 1 of the fourth lung determination module in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 18 that in Embodiment 1, the first The four-lung determination module 410 specifically includes:

The first judging unit 4101 is used to judge whether the column of the leftmost pixel of the connected region with the region number 2 is less than 2, that is, whether the corresponding index is less than 1000;

The second judging unit 4102 is configured to judge whether the column of the rightmost pixel of the connected region whose region number is 3 is less than 290 columns, that is, whether the corresponding index is less than 150000, when the first judging unit judges yes;

The third judging unit 4103 is used to determine whether the area (the number of pixels included) of the connected region with the region number 3 is greater than 2000 when the second judging unit judges yes;

The first lung determination unit 4104 is configured to determine that the connected area with the area number 3 is the left lung area when the third determination unit determines yes.

In this embodiment, judging whether the connected region labeled 3 is the left lung, that is, judging whether the connected region satisfies whether the column of the leftmost pixel is less than 2, and judging whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns (the corresponding index is less than 150000), and judge whether the area of the connected region labeled 3 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that the area 3 is the left lung area.

Fig. 19 is a structural block diagram of Embodiment 2 of the fourth lung determination module in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 19 that in this embodiment, the first The four-lung determination module 410 specifically includes:

The fourth lung determination module 4101 is used to determine whether the column of the leftmost pixel of the connected region whose region number is 2 is less than 2, that is, whether the corresponding index is less than 1000;

The second judging unit 4102 is configured to judge whether the column of the rightmost pixel of the connected region whose region number is 3 is less than 290 columns, that is, whether the corresponding index is less than 150000, when the first judging unit judges yes;

The third judging unit 4103 is used to determine whether the area (the number of pixels included) of the connected region with the region number 3 is greater than 2000 when the second judging unit judges yes;

The first lung determination unit 4104 is configured to determine that the connected area with the area number 3 is the left lung area when the third determination unit determines yes.

The second determination unit 4105 is configured to select a connected region with an area larger than 2000 from the connected regions whose region labels are not 1 and 2, which is the right lung region.

In this embodiment, judging whether the connected region labeled 3 is the left lung, that is, judging whether the connected region satisfies whether the column of the leftmost pixel is less than 2, and judging whether the column of the rightmost pixel of the connected region labeled 3 is less than 290 columns (the corresponding index is less than 150000), and judge whether the area of the connected region labeled 3 (the number of pixels included) is greater than 2000). If and only if the three conditions are met, it can be judged that the area marked 3 is the left lung area, the area marked 1 and the area marked 2 are the gap area between the human body and CT, and then start from the area marked 4 Search to find a connected region with an area greater than 2000, if it exists, it will be regarded as the right lung, otherwise, only one lung region will be found in the tomographic image.

Fig. 20 is a structural block diagram of Embodiment 6 of a lung extraction device in a lung segmentation and extraction system based on chest cross-sectional CT images provided by an embodiment of the present invention. It can be seen from Fig. 20 that in Embodiment 6, the above-mentioned embodiment None of the conditions from the first to fifth embodiments are met, and the lung region extraction device 400 specifically includes:

The starting region determination module 401 is configured to extract connected regions with region labels from the CT image after threshold segmentation. Taking Fig. 22 as an example, the extracted connected regions are the connected region with region number 1 and the connected region with region number 2. Taking FIG. 21 as an example, the extracted connected regions are connected regions whose region numbers are 1 to 5 in sequence.

The region prediction and determination module 413 is configured to determine a region set from the connected regions, the region set includes multiple regions, and each region satisfies that the column of the leftmost pixel is less than 2 and the area is greater than 100,000.

In a specific embodiment, it is judged sequentially whether the column where the leftmost pixel of each connected region is located is less than 2, and when the judgment is yes, continue to judge whether the area of the connected region is greater than 100,000, when the judgment is yes, the connected region is Available as a region in a region collection.

A start area determination module 414, configured to determine the area with the largest area number from the set of areas, which is called the start area;

The suspected area determination module 415 is configured to screen out suspected areas from the connected areas, and the area number of the suspected area is greater than the area number of the starting area.

The target area determination module 416 is configured to determine a target area from the suspected area, the area of the target area is larger than 2000, and the target area is the lung area.

In Embodiment 6, first judge the label of the gap area between the human body and the CT, and then traverse all the connected areas sequentially, and search for the last one that satisfies the fact that the leftmost pixel of the area is located in a column less than 2 (the corresponding index is less than 1000), and the area area is greater than 100,000 area, the start area is the gap area between the human body and the CT, and this area is marked as start. Start from the start+1 region to traverse all connected regions, arrange their areas from large to small, and select all block combinations with an area greater than 2000 as lung regions.

As mentioned above, it is a lung segmentation and extraction system based on chest cross-sectional CT images proposed by the present invention. Preprocessing is performed on the chest cross-sectional CT images, and then threshold segmentation is performed, and finally according to a preset template. The lung area extraction can realize the precise segmentation of the lung area and ensure the integrity of the lung parenchymal area segmentation.

The technical solution of the present invention will be described in detail below in conjunction with specific embodiments. In this particular example, the regimen includes the following:

1. Collecting data: acquiring lung CT image data, and setting the image as I.

2. Pretreatment.

The joint denoising method of median filtering and wavelet denoising is used to denoise the image I.

3. Threshold segmentation.

(1), the pixel value that Hu value is less than 0 among the image I is classified as 0, obtains I_;

(2), adopt automatic threshold value segmentation method (Ostu algorithm) to carry out preliminary binary segmentation to image I_, then reverse binary image, obtain B_I_;

(3) Open operation is performed on B_I_, and the structural element is a circle with a radius of 2. The purpose is to remove the isolated "island" to obtain a binary image _B_I_;

(4), carry out regional labeling to _B_I_, labeling sequence is from left to right, from top to bottom;

4. Extract the lung area through the set template. Fig. 23 is a schematic flow chart of lung area extraction in this specific embodiment, in this figure, block (2,1) represents the index of the first point in block 2, block (t) represents block t, len(block( t)) represents the area of the block t (the area is represented by the number of points in the area), and L==t means to search for the area labeled t.

The main idea of extracting the lung area is: firstly, judge the area label of the left lung (including the left and right lung adhesion); secondly, determine whether the left and right lungs are adhered, and then determine the number of the left and right lung area; finally, when the above steps cannot determine the lung area: first The first step is to determine the number of the non-voxel block (the remaining area is the voxel area), and the second step is to sort the area (the number of pixels) of the block in the voxel area in descending order, and select the first two digits according to the sorting. The block larger than 2000 is regarded as the lung area, otherwise it is judged that the CT image of this layer does not contain the lung area, and its specific description is as follows:

A. Determine whether the connected area labeled 2 is the left lung (here, the observer is used as a reference, which is inconsistent with the anatomy). Condition 1. Determine whether the column of the leftmost pixel in area 2 is greater than 20 (the corresponding index is greater than 10000); condition 2. Determine whether the column of the rightmost pixel in area 2 is less than 290 columns (the corresponding index is less than 150000); Whether the area (the number of pixels included) is greater than 2000. If the three conditions are met, it can be judged that area 2 is the lung area, and area 1 is the gap area between the human body and CT, and then search from area 3 to find a connected area with an area greater than 2000. If it exists, it will be regarded as the right lung , otherwise the tomographic image only finds one lung region.

B. If in condition 1, condition 2, and condition 3 in A, only condition 2 is not satisfied, that is, the column where the rightmost pixel in area 2 is greater than 290. At this time, it can be judged that area 2 is the lung area where the left and right lungs adhere.

C. If any of condition 1 and condition 2 in A is not satisfied, judge whether area 3 is the left lung. Set the judgment conditions: condition 1, determine whether the column of the leftmost pixel in area 2 is less than 2 (the corresponding index is less than 1000); condition 2, determine whether the column of the rightmost pixel in area 3 is less than 290 (corresponding index is less than 150000); condition 3 , Judging whether the area of the 3 area is greater than 2000. If the three conditions are met, it can be judged that area 3 is the lung area, and area 1 and area 2 are the gap area between the human body and CT, and then search from area 4 to find a connected area with an area greater than 2000. If it exists, it will be regarded as Right lung, otherwise the tomographic image only finds one lung region.

D. If A, B, and C are not satisfied, first judge the label of the gap area between the human body and CT: traverse all connected areas in turn, and search for the column where the leftmost pixel of the last satisfied area is less than 2 (the corresponding index is less than 1000), and If the area is greater than 100,000, if it is satisfied that the area is regarded as the gap area between the human body and the CT, the area is marked as start. Start from the start+1 region to traverse all connected regions, arrange their areas from large to small, and select all block combinations with an area greater than 2000 as lung regions.

E. If none of the above conditions are met, it is considered that the tomographic image does not contain the lung area.

5. Lung region segmentation results.

Figure 24 is a schematic diagram of the lung region extracted in Example 1, Figure 25 is a schematic diagram of the lung parenchyma region in Example 1, Figure 26 is a schematic diagram of the lung region extracted in Example 2, and Figure 27 is a schematic diagram of the lung parenchyma in Example 2 The schematic diagram of the region, FIG. 28 is the schematic diagram of the lung region extracted in the third embodiment, and FIG. 29 is the schematic diagram of the lung parenchyma region in the third embodiment. From the comparison of Fig. 24 to Fig. 29, it can be seen that in the LIDC database, compared with the gold standard calibrated by experts, the accuracy rate of the lung region extracted by the present invention is greater than 96%.

In summary, the lung segmentation and extraction method and system based on chest cross-sectional CT images proposed by the present invention can achieve precise segmentation of lung regions, ensure the integrity of lung parenchymal region segmentation, and avoid Deletion and the lack of regions lead to missed diagnosis in the follow-up diagnosis process.

The main protection point of this patent is the lung area extraction process, including five steps A, B, C, D, and E in the specific examples. The main idea of the lung region extraction process is: first determine the region label of the left lung (including the left and right lung adhesions); secondly determine whether the left and right lungs are adherent, and then determine the number of the left and right lung regions; finally, when the above steps cannot determine the lung region: the first step First determine the number of the non-voxel block (the remaining area is the voxel area), and in the second step, sort the area (number of pixels) of the block in the voxel area in descending order, and select the first two digits whose area is greater than The block of 2000 is regarded as the lung area, otherwise it is judged that the CT image of this layer does not contain the lung area.

The beneficial effects of the present invention are:

1. Make full use of the diagnostic value of CT imaging technology in the early stage of lung cancer, assist doctors to improve the correct diagnosis rate of pulmonary nodules, improve the efficiency of doctors' diagnosis, and reduce labor fatigue;

2. If the small nodules in the early stage of lung cancer can be treated in time, the patient can obtain a longer survival rate. This patent can effectively avoid missed detection of pulmonary nodules in computer-aided diagnosis, and timely diagnosis can reduce the patient's pain and medical cost while reducing the patient's medical treatment cost.

3. It can realize the precise segmentation of the lung area and ensure the integrity of the segmentation of the lung parenchyma.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware to complete, and the programs can be stored in general computer-readable storage media. During execution, it may include the processes of the embodiments of the above-mentioned methods. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) or a random access memory (Random Access Memory, RAM) and the like.

Those skilled in the art can also understand that whether various functions listed in the embodiments of the present invention are implemented by hardware or software depends on specific applications and design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the protection scope of the embodiments of the present invention.

In the present invention, specific examples have been applied to explain the principles and implementation methods of the present invention. The description of the above examples is only used to help understand the method of the present invention and its core idea; meanwhile, for those of ordinary skill in the art, according to this The idea of the invention will have changes in the specific implementation and scope of application. To sum up, the contents of this specification should not be construed as limiting the present invention.

Claims (22)

1. based on a lung segmentation extracting method for chest transversal section CT image, it is characterized in that, described method comprises:

Obtain the CT image of chest transversal section;

Pre-service is carried out to described CT image;

Threshold segmentation is carried out to pretreated CT image;

Lung extracted region is carried out to the CT image after Threshold segmentation.

2. method according to claim 1, is characterized in that, adopts medium filtering to carry out pre-service with the denoising method of combining of Wavelet Denoising Method to described CT image.

3. method according to claim 1, is characterized in that, describedly carries out Threshold segmentation to pretreated CT image and comprises:

The pixel that gray-scale value is less than 0 is determined from pretreated CT image;

Pixel gray-scale value being less than 0 is set to 0, obtains the first image;

Determine the grey level histogram of described first image;

Determine the trough between two crests in described grey level histogram, be considered as segmentation threshold;

According to described segmentation threshold, preliminary binary segmentation is carried out to the first described image, obtain the second image;

Negate is carried out to the second described image, obtains the 3rd image;

Adopt morphology opening operation to carry out Transformatin to described 3rd image, obtain bianry image;

According to strategy from left to right, from top to bottom, region labeling is carried out to the connected region in described bianry image, obtains the CT image after Threshold segmentation.

4. method according to claim 3, is characterized in that, carries out lung extracted region comprise the CT image after Threshold segmentation:

The connected region after region labeling is extracted from the CT image after Threshold segmentation;

Judging area label is whether the high order end pixel column of the connected region of 2 is greater than 20;

When being judged as YES, judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

When being judged as YES, judging area label is whether the area of the connected region of 2 is greater than 2000;

When being judged as YES, described region labeling be 2 connected region be lung region, the left side.

5. method according to claim 4, is characterized in that, carries out lung extracted region also comprise the CT image after Threshold segmentation:

Be not the connected region of 1 and 2, select the connected region that area is greater than 2000 from region labeling, be lung region, the right.

6. method according to claim 3, is characterized in that, carries out lung extracted region comprise the CT image after Threshold segmentation:

The connected region after region labeling is extracted from the CT image after Threshold segmentation;

Judging area label is whether the high order end pixel column of the connected region of 2 is greater than 20;

When being judged as YES, judging area label is whether the area of the connected region of 2 is greater than 2000;

When being judged as YES, judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

When being judged as NO, described region labeling be 2 connected region be pulmo adhesion lung region.

7. method according to claim 3, is characterized in that, carries out lung extracted region comprise the CT image after Threshold segmentation:

The connected region after region labeling is extracted from the CT image after Threshold segmentation;

Judging area label is whether the high order end pixel column of the connected region of 2 is greater than 20;

When being judged as YES, judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

When being judged as NO, according to described CT image determine described region labeling be 3 connected region whether be lung region, the left side.

8. method according to claim 3, is characterized in that, carries out lung extracted region comprise the CT image after Threshold segmentation:

The connected region after region labeling is extracted from the CT image after Threshold segmentation;

Judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

When being judged as YES, judging area label is whether the high order end pixel column of the connected region of 2 is greater than 20;

When being judged as NO, according to described CT image determine described region labeling be 3 connected region whether be lung region, the left side.

9. the method according to claim 7 or 8, is characterized in that, according to described CT image determine described region labeling be 3 connected region whether be that lung region, the left side comprises:

Judging area label be 2 connected region high order end pixel column whether be less than 2;

When being judged as YES, judging area label is whether the low order end pixel column of the connected region of 3 is less than 290;

When being judged as YES, judging area label is whether the area of the connected region of 3 is greater than 2000;

When being judged as YES, described region labeling be 3 connected region be lung region, the left side.

10. method according to claim 9, is characterized in that, according to described CT image determine described region labeling be 3 connected region whether be that left side lung also comprises:

Be not the connected region of 1,2,3 and 4, select the connected region that area is greater than 2000 from region labeling, be lung region, the right.

11. methods according to claim 3, is characterized in that, carry out lung extracted region comprise the CT image after Threshold segmentation:

The connected region after region labeling is extracted from the CT image after Threshold segmentation;

From described connected region, determine regional ensemble, described regional ensemble includes multiple region, and each region all meets high order end pixel column and is less than 2 and area is greater than 100000;

From described regional ensemble, determine the region that region labeling is maximum, be called and start region;

From described connected region, filter out suspicious region, the region labeling of described suspicious region is greater than the region labeling in described beginning region;

From described suspicious region, determine target area, the area of described target area is greater than 2000, and described target area is lung region.

12. 1 kinds, based on the lung segmentation extraction system of chest transversal section CT image, is characterized in that, described system comprises:

CT image acquiring device, for obtaining the CT image of chest transversal section;

Pretreatment unit, for carrying out pre-service to described CT image;

Threshold segmentation device, for carrying out Threshold segmentation to pretreated CT image;

Lung region extracting device, for carrying out lung extracted region to the CT image after Threshold segmentation.

13. systems according to claim 12, is characterized in that, adopt medium filtering to carry out pre-service with the denoising method of combining of Wavelet Denoising Method to described CT image.

14. systems according to claim 12, is characterized in that, described Threshold segmentation device comprises:

Determination module, for determining the pixel that gray-scale value is less than 0 from pretreated CT image;

First image determination module, is set to 0 for pixel gray-scale value being less than 0, obtains the first image;

Histogram determination module, for determining the grey level histogram of described first image;

Segmentation threshold determination module, for determining the trough in described grey level histogram between two crests, is considered as segmentation threshold;

Second image determination module, for carrying out preliminary binary segmentation according to described segmentation threshold to the first described image, obtains the second image;

3rd image determination module, for carrying out negate to the second described image, obtains the 3rd image;

Bianry image determination module, for adopting morphology opening operation to carry out Transformatin to described 3rd image, obtains bianry image;

Region labeling module, for carrying out region labeling to the connected region in described bianry image according to strategy from left to right, from top to bottom, obtains the CT image after Threshold segmentation.

15. systems according to claim 14, is characterized in that, described lung region extracting device comprises:

Connected region extraction module, for extracting the connected region after region labeling from the CT image after Threshold segmentation;

First judge module is whether the high order end pixel column of the connected region of 2 is greater than 20 for judging area label;

Second judge module, for when the first described judge module is judged as YES, judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

3rd judge module, for when the second described judge module is judged as YES, judging area label is whether the area of the connected region of 2 is greater than 2000;

First lung's determination module, for when described 3rd judge module is judged as YES, determine described region labeling be 2 connected region be lung region, the left side.

16. systems according to claim 15, is characterized in that, described Threshold segmentation device also comprises:

Second lung's determination module, for from region labeling be not 1 and 2 connected region in select the connected region that area is greater than 2000, be lung region, the right.

17. systems according to claim 14, is characterized in that, described Threshold segmentation device also comprises:

4th judge module, for when the first described judge module is judged as YES, judging area label is whether the area of the connected region of 2 is greater than 2000;

5th judge module, for when described 4th judge module is judged as YES, judging area label is whether the low order end pixel column of the connected region of 2 is less than 290 row;

3rd lung's determination module, for when described 4th judge module is judged as NO, determine described region labeling be 2 connected region be pulmo adhesion lung region.

18. systems according to claim 14, is characterized in that, described lung region extracting device comprises:

6th judge module is whether the low order end pixel column of the connected region of 2 is less than 290 row for judging area label;

7th judge module, for when described 6th judge module is judged as YES, judging area label is whether the high order end pixel column of the connected region of 2 is greater than 20;

When described 7th judge module is judged as NO, described lung region extracting device also comprises: the 4th lung's determination module, for determine according to described CT image described region labeling be 3 connected region whether be lung region, the left side.

19. systems according to claim 14, is characterized in that, when described second judge module is judged as NO, described lung region extracting device also comprises:

4th lung's determination module, for determine according to described CT image described region labeling be 3 connected region whether be lung region, the left side.

20. systems according to claim 18 or 19, it is characterized in that, the 4th described lung's determination module comprises:

First judging unit, whether the connected region high order end pixel column being 2 for judging area label is less than 2;

Second judging unit, for when described first judging unit is judged as YES, judging area label is whether the low order end pixel column of the connected region of 3 is less than 290;

3rd judging unit, for when described second judging unit is judged as YES, judging area label is whether the area of the connected region of 3 is greater than 2000;

First lung's determining unit, for when described 3rd judging unit is judged as YES, determine described region labeling be 3 connected region be lung region, the left side.

21. systems according to claim 20, is characterized in that, the 4th described lung's determination module also comprises:

Second determining unit, for from region labeling be not 1,2,3 and 4 connected region in select the connected region that area is greater than 2000, be lung region, the right.

22. systems according to claim 14, is characterized in that, described lung region extracting device comprises:

District estimates and determination module, and for determining regional ensemble from described connected region, described regional ensemble includes multiple region, and each region all meets high order end pixel column and is less than 2 and area is greater than 100000;

Start area determination module, for determining the region that region labeling is maximum from described regional ensemble, being called and starting region;

Suspicious region determination module, for filtering out suspicious region from described connected region, the region labeling of described suspicious region is greater than the region labeling in described beginning region;

Target area determination module, for determining target area from described suspicious region, the area of described target area is greater than 2000, and described target area is lung region.