JP2006042998A - Region extraction method and apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To create an image of a pixel density for each search target area and determine a density threshold value, so that when the search target area does not include an image of an extraction target such as a blood vessel, the image is not an extraction target Solve the problem of extracting.
After calculating a density threshold value from a histogram, an adjacent slice image when the density threshold value is smaller than the average density value of the entire search target area or an area extracted by the density threshold value is larger than a predetermined value. The density threshold value used in the above is used to check again whether it is smaller than the average density value or whether the extracted area is larger than a predetermined value, and when the condition is satisfied, the search target area is cleared. Only the region of the extraction object can be extracted.
[Selection] Figure 1

Description

  The present invention relates to a region extraction method for extracting a target region such as a blood vessel from a slice image such as a CT image and an improvement of the device in order to generate a three-dimensional model image such as a blood vessel, bone, or organ in the medical field. It is.

  As a prior art of a region extracting method for extracting a target region such as a blood vessel from a slice image obtained by MR (Magnetic Resonance Imaging) or CT (Computed Tomography), there are the following.

Cerebrovascular extraction program and apparatus for MRA images

  In addition, as a method for accurately extracting only the target region, there is a method filed by the present inventor as Japanese Patent Application No. 2004-060117.

  FIG. 5 shows the configuration of a computer-aided diagnosis apparatus in which such a region extraction method is implemented. In FIG. 5, image data of a slice image such as a CT image is stored in the storage device 4. The personal computer 5 searches the image data stored in the storage device 4, performs image processing such as blood vessel extraction, and displays the result on the display device 6.

  FIG. 6 shows a flowchart of an algorithm for extracting a blood vessel region. In FIG. 6, the initial target area is designated in step (C-1). That is, a key slice image that is an image that best represents the feature of the region of interest is used, and an extraction region such as a blood vessel is designated using a mouse or the like while viewing this image.

  Next, it moves to an adjacent slice image at step (C-2), and expands the region at step (C-3) to set a search target region. In this step, the search target area is enlarged by performing an expansion process of adding pixels around the set target area. Next, in step (C-4), a histogram of pixel density distribution in this region is created, and in step (C-5), a density threshold is determined from this histogram.

  FIG. 7 shows an example of a pixel density histogram. The vertical axis represents frequency (number of pixels), and the horizontal axis represents pixel density values. Target areas such as blood vessels are distributed where the density value is high, and non-target areas are distributed where the density value is low. Assuming that both the histograms of the target region and the non-target region have a normal distribution, an approximate distribution map is obtained, and the boundary between these distribution maps is set as a density threshold value.

  Next, in step (C-6), the image is binarized using this threshold value. Then, the images before and after performing these processes are compared in step (C-7), and if there is a difference, the process returns to step (C-3). If they match, the image before and after processing is compared for all slice images in step (C-8). If there is a difference, the process returns to step (C-2), and if all match, the process ends.

  Next, the processing of these steps will be specifically described with reference to FIG. 71 is a key slice image in which the target area is designated in step (C-1). Reference numeral 711 denotes a designated target area. Reference numeral 72 denotes a slice image adjacent to 71, and a target area (shadow portion) is set based on the target area 711 set in the slice image 71. Then, the search target area 721 is set by performing expansion processing in step (C-3). A threshold value is set for the search target area 721 in step (C-5) and binarized in step (C-6) to obtain a target area 722.

  However, in such a region extraction method, since a threshold is determined by creating a histogram for each search target region of each slice image, the search target region does not include an image of an extraction target such as a blood vessel. In some cases, there is a problem that a region that is not an image of the extraction target is extracted.

  This will be described with reference to FIG. 9A is an original image of the head, and FIG. 9B is an image obtained by extracting blood vessels from the original image by the algorithm of FIG. As is apparent from this figure, it can be seen that a portion other than the blood vessel 8 (non-blood vessel portion 9) is also extracted.

  Accordingly, an object of the present invention is to determine a threshold value in consideration of continuity between slice images, and to add a noise removal process to accurately extract a target region such as a blood vessel. And providing an apparatus thereof.

In order to achieve such a problem, the invention according to claim 1 of the present invention is:
Among the plurality of slice images obtained by slicing a three-dimensional object, a search target region is sequentially set in adjacent slice images based on the target region specified in the key slice image, and binarization of the search target region is performed. In the region extraction method for obtaining the target region by repeating,
A first step of calculating a density threshold from a histogram of pixel density in the search target region;
A second step of determining whether the search target area is a target area based on the density threshold;
A third step of resetting the density threshold when it is determined that the target region is not the target region in the second step;
A fourth step of determining whether or not the search target region is a target region based on the density threshold reset by the third step;
A fifth step of clearing the search target region when it is determined in the fourth step that the target region is not;
A sixth step of binarizing the search target area using the density threshold when the search target area is determined to be the target area in the second step or the fourth step; ,
Is provided. Only the region of the object can be extracted accurately.

The invention according to claim 2 is the invention according to claim 1,
In the second step or the fourth step, the density threshold is smaller than the average density value of the entire search target region, or the area extracted by the density threshold is equal to or larger than the average density value. When the value is larger than the predetermined value, it is determined that the target region is not the target region, and the target region is determined otherwise. It is possible to accurately determine whether the region is a target region.

The invention according to claim 3 is the invention according to claim 1 or claim 2,
As the density threshold value to be reset in the third step, the density threshold value of the connected region of adjacent slice images is used. A reasonable density threshold can be set using connectivity.

The invention according to claim 4 is the invention according to claim 1 or claim 3,
In the second step or the fourth step, when the number of holes in the region extracted by the density threshold is larger than a predetermined value, it is determined that the region is not the target region. Whether it is the target area can be determined.

The invention according to claim 5 is the invention according to claim 4,
In the second step or the fourth step, the determination is made based on the difference in the number of pixels or the difference between the search target region and the region extracted by the density threshold value. It is possible to accurately determine whether or not the target area.

The invention according to claim 6 is the invention according to claim 1 or 5,
After completing the extraction of the target region for all of the plurality of slice images, the unconnected pixels are deleted in consideration of the connectivity of the slice images. Noise can be removed.

The invention according to claim 7 is the invention according to claim 1 or 6,
The slice image is an image obtained by MR (Magnetic Resonance Imaging) or CT (Computed Tomography). Wide range of application.

The invention according to claim 8 is the invention according to claim 1 or 7,
The target regions to be extracted are blood vessels, bones, trachea, brain parenchyma, and organs. Wide range of application.

The invention according to claim 9
A storage device for storing slice image data;
It is determined whether or not the search target area is a target area based on a density threshold calculated based on a pixel density histogram in the search target area, and the density threshold is reset when it is determined that the search target area is not the target area. Then, based on the reset density threshold, it is determined whether the search target area is a target area. When it is determined that the search target area is not the target area, the search target area is cleared and the search target area is A computer that obtains the target area by repeating the process of binarizing the search target area using the density threshold at that time when it is determined that the area is an area;
A display device for displaying the target area obtained by the computer;
Is provided. Only the image of the object can be extracted accurately.

As is apparent from the above description, the present invention has the following effects.
According to the first, second, third, fourth, sixth, seventh, eighth, and ninth aspects of the present invention, a density threshold value is calculated from a pixel density histogram of a search target area, and the search target area is calculated from the density threshold value. Is not the target region, the density threshold is reset to determine whether it is the target region, and if it is not the target region, the target region is cleared. In addition, the image is binarized using the density threshold value determined to be the target region.

  In a slice image of a three-dimensional object such as a human body obtained by CT or MR, a gray region having an intermediate density occupies a lot, and therefore, the boundary for dividing the region is often ambiguous. According to the present invention, it is possible to accurately extract a target region even in an ambiguous image with many gray regions by determining whether the search target region is a target region twice and clearing the region that is not the target region. There is.

  In addition, by using the property that the density threshold values of the physically connected portions take similar values, when it is first determined that the area is not the target region, the density threshold values of the surrounding physical threshold values are physically connected. It was reset and it was determined whether it is a target area. Therefore, there is an effect that the target region can be extracted more accurately.

  Furthermore, since the non-connected pixels are removed by utilizing the connectivity of slice images, unnecessary noise can be effectively removed, and a clearer image can be extracted. There is also.

  Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a flowchart showing an embodiment of a region extracting method according to the present invention. The program created based on this flowchart is executed by the personal computer 5 shown in FIG. 5, and the result is displayed on the display device 6. A slice image taken by CT or MR angiography is stored in the storage device 4.

  In FIG. 1, an initial target area is specified in step (A-1), the target area is moved to an adjacent slice image in step (A-2), and the moved target area is displayed in step (A-3). Is expanded to generate a search target region. In step (A-4), a histogram of pixel density as shown in FIG. 7 is created for this search target region, and discriminant analysis is performed in step (A-5) to calculate a threshold value. These steps are the same as steps (C-1) to (C-5) in the flowchart of FIG. The discriminant analysis in step (A-5) corresponds to (C-5) in the flowchart of the conventional example in FIG. 6, and calculates the density threshold value.

  Next, in step (A-6), target area determination for determining whether the search target area is the target area is performed. Details of this step (A-6) will be described later. If the result of step (A-6) is True (the target area), the process proceeds to threshold determination in step (A-7), and the already obtained density threshold is set as the correct density threshold.

  Next, in step (A-8), the slice image is binarized using the determined density threshold value, and the images before and after the binarization are compared in step (A-9). Return to A-3). If the two images match, the image before and after processing is compared for all slice images in step (A-10). If they do not match, the process returns to step (A-2), and all slice images are If they match, the noise removal process in step (A-11) is performed. Steps (A-8) to (A-10) are the same processes as steps (C-6) to (C-8) of the conventional example of FIG.

  If it is determined that the target region is determined to be False (not the target region) in Step (A-6), the density threshold is reset in Step (A-12). As the threshold value to be reset, for example, a density threshold value used in a connected region of adjacent slice images, that is, a physically connected region is employed. In step (A-13), the target area is extracted using the reset density threshold value, and the target area is determined in step (A-14). This target area determination is the same processing as the target area determination in step (A-6). If the target area determination is true, the process proceeds to step (A-7) to determine the density threshold value, and the processing after step (A-8) is performed. If the result of the target area determination is False, the target area obtained in step (A-15) is cleared, and the process returns to step (A-2).

  Next, the noise removal process in step (A-11) will be described. Since a slice image is an image obtained by cutting one object such as a human body, the connected portion between adjacent pixels in one slice image or between adjacent slice images rarely takes a random value. There is connectivity unique to objects. Using this property, noise can be removed by deleting pixels that are not connected.

  FIG. 2 shows a flowchart of target area determination in steps (A-6) and (A-14). First, in step (B-1), the average density value of the entire search target area is calculated, and the density threshold value is compared with this average density value.

  If the density threshold value is equal to or greater than the average density value, then it is checked whether the area extracted by the density threshold value is greater than a predetermined value, and if it is greater, False (not the target area) is output. If it is smaller, True (the target area) is output and the process ends.

  FIG. 3 shows an example of the effect of this embodiment. FIG. 3A is an image obtained by extracting cerebral blood vessels from blood vessel image data of the head by MR angiography using the algorithm of the conventional example of FIG. 1 is blood vessel, 2 is brain parenchyma, 3 is intraorbital fat. Since the concentration of the brain parenchyma and blood vessel region is ambiguous, not only the blood vessel 1 to be extracted but also the brain parenchyma 2 and intraorbital fat 3 are extracted. FIG. 5B shows an example in which blood vessels are extracted by the algorithm shown in FIG. As can be seen from comparison with (A), it can be seen that only blood vessels have been extracted.

  FIG. 4 shows an example of extracting blood vessels. (A) is an example in which only blood vessels are extracted, and is a single region. FIG. 5B shows an example in which noise is extracted. In the case of noise that is not a blood vessel, there are cases where pixels are missing from the inside of the region as shown on the right side of (B). In determining the target area, not only the algorithm of FIG. 2 but also the number of holes in the extracted area can be counted to determine whether or not the extracted area is the target area.

  In addition, for the area determined to be the target area based on the number of holes, the number of pixels or areas when the area is expanded and the number of areas or areas when the area is contracted by binarization It can also be determined by the difference from the number. That is, the determination can also be made based on the difference in the number of pixels or the difference in the number of areas between the search target area and the area extracted based on the density setting value.

  In this embodiment, the extraction of blood vessels has been described. However, the present invention can be applied to extraction of not only blood vessels but also trachea, bone, brain parenchyma, and organs.

It is a flowchart which shows one Example of this invention. It is a flowchart which shows one Example of this invention. It is a characteristic view which shows the effect of this invention. It is a characteristic view which shows the effect of this invention. It is a block diagram of a computer-aided diagnosis apparatus. It is a flowchart which shows the conventional area | region extraction method. It is an example of a histogram of density values. It is a figure for demonstrating the area | region extraction method. It is an example of the blood vessel image extracted by the conventional area | region extraction method.

Explanation of symbols

1 Blood vessel 2 Brain parenchyma 3 Orbital fat 4 Storage device 5 Personal computer 6 Display device

Claims (9)

Among the plurality of slice images obtained by slicing a three-dimensional object, a search target region is sequentially set in adjacent slice images based on the target region specified as the key slice image, and the search target region is binarized. In the region extraction method for obtaining the target region by repeating,
A first step of calculating a density threshold from a histogram of pixel density in the search target region;
A second step of determining whether the search target area is a target area based on the density threshold;
A third step of resetting the density threshold when it is determined that the target region is not the target region in the second step;
A fourth step of determining whether the search target region is a target region based on the density threshold value reset by the third step;
A fifth step of clearing the search target region when it is determined in the fourth step that the target region is not;
A sixth step of binarizing the search target region using the density threshold at that time when the search target region is determined to be the target region in the second step or the fourth step; ,
A region extraction method comprising:   In the second step or the fourth step, the density threshold is smaller than the average density value of the entire search target region, or the area extracted by the density threshold is equal to or larger than the average density value. 2. The region extracting method according to claim 1, wherein the region is determined not to be a target region when larger than a predetermined value but to be a target region at other times.   3. The area extraction method according to claim 1, wherein the density threshold value reset in the third step is a density threshold value of a connected area of adjacent slice images.   2. The method according to claim 1, wherein the second step or the fourth step determines that the target region is not the target region when the number of holes in the region extracted by the density threshold is larger than a predetermined value. Alternatively, the region extraction method according to claim 3.   5. The determination according to claim 4, wherein in the second step or the fourth step, the determination is made based on a difference between pixels of a search target region and a region extracted by a density threshold or a difference in the number of regions. Region extraction method.   6. The non-connected pixels are deleted in consideration of the connectivity of the slice images after the extraction of the target region is completed for all of the plurality of slice images. The region extraction method according to any one of the above.   The region extraction method according to claim 1, wherein the slice image is an image obtained by MR (Magnetic Resonance Imaging) or CT (Computed Tomography).   The region extraction method according to claim 1, wherein the target region to be extracted is a blood vessel, a bone, a trachea, a brain parenchyma, or an organ. A storage device for storing slice image data;
It is determined whether or not the search target area is a target area based on a density threshold calculated based on a pixel density histogram in the search target area, and the density threshold is reset when it is determined that the search target area is not the target area. Then, based on the reset density threshold, it is determined whether the search target area is a target area. When it is determined that the search target area is not the target area, the search target area is cleared and the search target area is A computer that obtains a target area by repeating the process of binarizing the search target area using a density threshold at that time when it is determined to be an area;
A display device for displaying the target area obtained by the computer;
An area extracting apparatus comprising:

Images