A pancreatic CT image segmentation method, system and storage medium

Abstract

The present invention relates to the field of image segmentation technology, and in particular to a pancreatic CT image segmentation method, system and storage medium, which analyzes the gray value characteristics of the pancreas in the pancreatic threshold segmentation image, removes the edge interference area of ​​the pancreas, and obtains the pancreatic interference-free image; analyzes the fluctuation characteristics of the edge pixel direction of the pancreatic interference-free image to obtain the inflection point; based on the inflection point, the pancreatic interference-free image is linearly cut to obtain paired cutting areas, and the grayscale difference and shape feature difference of the two cutting areas are analyzed to determine the adhesion area inflection point and the adhesion area; based on the adhesion area inflection point, the distribution characteristics of the pixel points in the adhesion area are analyzed to obtain the possibility that the pixel point is the pancreatic edge pixel point; the pixel points are classified according to the possibility size to obtain the pancreatic edge pixel point sequence. The present invention improves the accuracy of pancreatic CT segmentation and makes the segmentation result more authentic.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a pancreatic CT image segmentation method, system and storage medium. Background Art

[0002] Pancreatic CT (Computed Tomography) image segmentation technology has important application value in the medical field, especially in the diagnosis and treatment of diseases such as pancreatic cancer, diabetes, and pancreatic neuroendocrine tumors. However, due to the complex and highly variable anatomical structure of the pancreas, its automatic segmentation has always been a challenging task. At present, pancreatic CT image segmentation technology based on deep learning has made significant progress. Another innovative method is to use convolutional neural networks with attention modules. However, the existing segmentation technology still has some shortcomings.

[0003] When taking CT images of the pancreas, the pancreas is roughly located outside the body. Since there are many types of organs in the human body and their shapes are complex, the resulting pancreatic CT image is not just a single pancreas, but also contains other organs that are very close to the pancreas, resulting in adhesions in the image, which increases the complexity of segmentation. In addition, the pancreas accounts for a small proportion of the CT image (<0.5%), and the shape, size, and position of different human bodies vary greatly, further increasing the difficulty of the segmentation task.

[0004] Therefore, there is an urgent need for a method to solve the problem of adhesion between the pancreas and other organs in the result image after threshold segmentation. Summary of the invention

[0005] In order to solve the above technical problems, the object of the present invention is to provide a pancreatic CT image segmentation method, system and storage medium.

[0006] According to a first aspect of an embodiment of the present invention, a pancreatic CT image segmentation method is provided, and the technical solution adopted is specifically as follows:

[0007] Acquire a pancreatic CT image, and perform threshold segmentation on the pancreatic CT image to obtain a pancreatic threshold segmentation image;

[0008] Analyzing the grayscale value features of the pancreas in the pancreas threshold segmentation image, removing the edge interference region of the pancreas in the pancreas threshold segmentation image, and obtaining a pancreas interference-free image;

[0009] Extracting edge pixel points of the pancreas interference-free image and obtaining the direction of the edge pixel points, analyzing the fluctuation characteristics of the direction of the edge pixel points, and obtaining the inflection point;

[0010] Based on the inflection point, the pancreatic interference-free image is linearly cut to obtain a pair of cut regions, and the grayscale difference and shape feature difference of the two cut regions in a pair are analyzed to determine the adhesion region inflection point and the adhesion region;

[0011] Based on the inflection point of the adhesion area, analyzing the distribution characteristics of the pixel points in the adhesion area, and obtaining the possibility that the pixel point is a pixel point at the edge of the pancreas;

[0012] Classifying the pixel points according to the likelihood to obtain a pancreatic edge pixel point sequence;

[0013] The pancreatic image region is improved according to the pancreatic edge pixel sequence and combined with the pancreatic interference-removed image.

[0014] In some embodiments of the present invention, analyzing the gray value features of the pancreas in the pancreas threshold segmentation image, removing the edge interference region of the pancreas in the pancreas threshold segmentation image, and obtaining the pancreas interference-free image includes:

[0015] Extracting pixels of the foreground area in the pancreas threshold segmentation image, and obtaining the mean grayscale value of all pixels of the foreground area;

[0016] Performing edge detection on the pancreas threshold segmentation image to obtain a plurality of closed edge contours, and obtaining the gray value contour mean of all pixel points within the closed edge contours;

[0017] Establishing a minimum circumscribed rectangle for each closed edge contour, and calculating the aspect ratio of the rectangle;

[0018] Analyze the difference between the mean value of the gray value contour and the mean value of the gray value of the pixels in the foreground area, and combine the aspect ratio of the rectangle to obtain the possibility of the pancreatic area;

[0019] A pancreatic region possibility threshold is set, and a pancreatic interference-free image is obtained according to the pancreatic region possibility.

[0020] In some embodiments of the present invention, extracting edge pixels of the pancreas interference-free image and obtaining the direction of the edge pixels, analyzing the fluctuation characteristics of the direction of the edge pixels, and obtaining the inflection points include:

[0021] Extracting edge pixel points of the pancreas de-interference image;

[0022] Generate eight-direction chain codes for all edge pixels, and obtain the directions of the edge pixels according to the eight-direction chain codes;

[0023] Arrange the values ​​of the directions of the edge pixels in the order of extension of the edge pixels to obtain a direction value sequence;

[0024] Pixel points whose values ​​change in the direction value sequence are extracted to obtain inflection points.

[0025] In some embodiments of the present invention, based on the inflection point, the pancreas interference-free image is linearly cut, and the method further includes:

[0026] The inflection points are combined in pairs and connected to obtain an inflection point connection line;

[0027] Calculating the number of pixels on the inflection point line as the ratio of the number of pixels in the pancreas interference-free image to the total number of pixels on the inflection point line;

[0028] A ratio threshold is set, and the inflection point connection line whose ratio is greater than the ratio threshold is retained to obtain the suspected inflection point of the pancreatic region.

[0029] In some embodiments of the present invention, based on the inflection point, the pancreatic interference-free image is linearly cut to obtain a pair of cut regions, and the grayscale difference and shape feature difference of the two pairs of cut regions are analyzed to determine the adhesion region inflection point and the adhesion region, including:

[0030] Connecting the suspected pancreatic region inflection points in pairs to obtain a connecting line of the suspected pancreatic region inflection points;

[0031] Performing linear cutting on the pancreatic interference-free image through the connecting line of the inflection points of the suspected pancreatic regions to obtain paired cutting regions;

[0032] Calculating the difference between the grayscale value mean of the pixel points in the two paired cutting areas and the grayscale value mean of all the pixel points in the pancreatic interference-free image, and then analyzing the difference between the two differences to obtain the grayscale difference of the two paired cutting areas;

[0033] Establishing minimum circumscribed rectangles for the two cutting areas in a pair, respectively, and obtaining the aspect ratio of the rectangles, analyzing the difference between the two aspect ratios, and obtaining the difference in shape characteristics of the two cutting areas in a pair;

[0034] According to the grayscale difference and the shape feature difference, the adhesion region inflection point and the adhesion region are determined.

[0035] In some embodiments of the present invention, based on the inflection point of the adhesion area, analyzing the distribution characteristics of the pixel points in the adhesion area to obtain the possibility that the pixel point is a pixel point at the edge of the pancreas includes:

[0036] In the edge pixel points of the pancreas interference-free image, edge pixel points located between the two adhesion region inflection points and adjacent to the adhesion region inflection points are used as adjacent pixel points;

[0037] The two inflection points of the adhesion area and the two adjacent pixel points are respectively formed into vectors, and an expression of a straight line where the two vectors are located is obtained;

[0038] Based on the two expressions, the distances from the pixel points in the adhesion area to the straight lines where the two vectors are located are calculated respectively;

[0039] According to the distance, the possibility that the pixel point is a pancreatic edge pixel point is obtained.

[0040] In some embodiments of the present invention, the pixel points are classified according to the likelihood to obtain a pancreatic edge pixel point sequence, including:

[0041] Setting a pancreatic edge pixel possibility threshold, and classifying the pixel into pancreatic edge area pixel and other area pixel according to the possibility;

[0042] Merging the pancreatic edge region pixel points with known pancreatic edge region pixel points in other regions of the pancreas to obtain a pancreatic edge pixel point sequence;

[0043] It is determined whether the pixel points in the other regions belong to the pixel points in the inner region of the pancreatic edge pixel point sequence, and the pixel points in the other regions that belong to the inner region of the pancreatic edge pixel point sequence are retained.

[0044] In some embodiments of the present invention, the pancreatic image region is improved according to the pancreatic edge pixel sequence and combined with the pancreatic interference-free image, including:

[0045] The pancreatic edge pixel sequence and the contour area in the pancreatic interference-free image are compared, and the pixels outside the pancreatic edge pixel sequence are removed, and the pixels within the pancreatic edge pixel sequence are completed to improve the pancreatic image area.

[0046] According to a second aspect of an embodiment of the present invention, a pancreatic CT image segmentation system is provided, the system comprising:

[0047] A pancreatic CT image acquisition module is used to acquire a pancreatic CT image, perform threshold segmentation on the pancreatic CT image, and obtain a pancreatic threshold segmentation image;

[0048] A pancreatic image interference removal module is used to analyze the gray value characteristics of the pancreas in the pancreatic threshold segmentation image, remove the edge interference area of ​​the pancreas in the pancreatic threshold segmentation image, and obtain a pancreatic interference-free image;

[0049] An adhesion region determination module is used to extract edge pixel points of the pancreatic interference-free image and obtain the direction of the edge pixel points, analyze the fluctuation characteristics of the direction of the edge pixel points, and obtain the inflection point; and based on the inflection point, perform linear cutting on the pancreatic interference-free image to obtain paired cutting regions, and analyze the grayscale difference and shape feature difference of the two paired cutting regions to determine the adhesion region inflection point and the adhesion region;

[0050] A pancreatic edge pixel determination module is used to analyze the distribution characteristics of the pixels in the adhesion area based on the inflection point of the adhesion area to obtain the probability that the pixel is a pancreatic edge pixel; and then classify the pixel according to the probability to obtain a pancreatic edge pixel sequence;

[0051] The pancreas contour improvement module is used to improve the pancreas image area according to the pancreas edge pixel sequence in combination with the pancreas interference-free image.

[0052] According to a third aspect of the embodiments of the present invention, a computer-readable storage medium is provided, on which a computer program is stored; the computer program is executed by a processor to implement a pancreatic CT image segmentation method as described in the first aspect of the embodiments of the present invention.

[0053] Compared with the prior art, the pancreatic CT image segmentation method, system and storage medium provided by the present invention have the following beneficial effects:

[0054] The present invention analyzes the gray value characteristics of the pancreas in the pancreas threshold segmentation image, removes the edge interference area of ​​the pancreas, obtains the pancreas interference-free image, and performs preliminary separation of the pancreas as a basis for subsequent analysis; analyzes the fluctuation characteristics of the edge pixel direction of the pancreas interference-free image to obtain the inflection point, and the inflection point is used as a reference for identifying the adhesion area formed by the pancreas and other organs; based on the inflection point, the pancreas interference-free image is linearly cut to obtain paired cutting areas, and the grayscale difference and shape feature difference of the two cutting areas are analyzed to determine the adhesion area inflection point and the adhesion area, and the determination of the adhesion area inflection point and the adhesion area limits a reliable range for determining the pancreatic edge pixel point; based on the adhesion area inflection point, the distribution characteristics of the pixel points in the adhesion area are analyzed to obtain the possibility that the pixel point is the pancreatic edge pixel point; the pixel points are classified according to the possibility to obtain a pancreatic edge pixel point sequence. The present invention determines the possibility of pixels suspected to be at the edge of the pancreas according to the trend of the pancreatic contour, and then classifies the pixels of the adhesion area according to the size of the possibility, further realizing the segmentation of the pancreatic edge and gap area and other organ pixels, which is beneficial to improving the accuracy of pancreatic CT segmentation and making the segmentation result more authentic. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of the basic flow of a pancreatic CT image segmentation method provided by one embodiment of the present invention;

[0057] Figure 2 A pancreatic CT image directly output by a computer through CT scanning is provided in one embodiment of the present invention;

[0058] Figure 3 A pancreas threshold segmentation image obtained by performing Ojin threshold segmentation on a pancreas CT image provided by an embodiment of the present invention;

[0059] Figure 4 A pancreas interference-free image obtained by removing the edge interference area of ​​the pancreas from a pancreas threshold segmentation image provided by an embodiment of the present invention;

[0060] Figure 5 A schematic diagram of the basic flow of a method for acquiring a pancreatic interference-free image provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a pancreatic CT image segmentation method, system and storage medium proposed by the present invention, its specific implementation, structure, features and effects in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. Terms such as "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such article or device. In the absence of further restrictions, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the article or device including the element.

[0063] The pancreatic CT image is taken externally at the approximate location of the pancreas. Since the organs in the human body are of many types and complex shapes, the pancreatic CT image obtained is not a single pancreas, but also contains other organs. The distance between these organs and the pancreas is very close, resulting in adhesion between the pancreas and other organs in the pancreatic CT image. This situation is not conducive to the doctor's judgment of the patient's condition.

[0064] The main purpose of the present invention is to solve the problem of adhesion between the pancreatic organ and other organs in the result image after the threshold segmentation of the pancreatic CT image, and to judge the possibility of the pixel points suspected to be the edge of the pancreas by the trend of the pancreatic contour, which is conducive to improving the accuracy of the pancreatic CT image segmentation and making the segmentation result more authentic.

[0065] A specific scheme of a pancreatic CT image segmentation method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0066] See also Figure 1 , which shows the basic process of a pancreatic CT image segmentation method provided by an embodiment of the present invention.

[0067] like Figure 1 As shown, an embodiment of the present invention provides a pancreatic CT image segmentation method, which specifically includes:

[0068] S100: Acquire a pancreatic CT image, perform threshold segmentation on the pancreatic CT image, and obtain a pancreatic threshold segmentation image.

[0069] In order to study the pancreatic CT image segmentation method, we first need to obtain pancreatic CT images. Hospitals often use X-rays in CT scans to perform multi-layer scans of the human pancreas, with each layer of scan being several millimeters thick. The detector receives the X-rays that pass through the human body and converts them into visible light, and then converts the visible light into electrical signals. These electrical signals are converted into digital signals through a digital simulator and input into a computer, which then outputs them in the form of images to obtain the original pancreatic CT images.

[0070] In reality, it is impossible to perform a CT scan of a single pancreas. The actual CT scan images are taken from a certain part of the human body, and the organs and tissues in various parts of the human body are relatively complex and closely arranged, which greatly hinders doctors from analyzing and judging the patient's condition based on the image of the pancreas alone. Figure 2 , which shows a pancreatic CT image directly output by a computer through CT scanning provided by an embodiment of the present invention, from Figure 2The approximate position of each organ can be distinguished in the image, but the objects presented in the image are complex and numerous, and the grayscale values ​​between the background and the object are similar, making it difficult to distinguish them. In order to distinguish the background and the object of the pancreatic CT image, this embodiment uses the computer image threshold segmentation technology to perform Otsu threshold segmentation on the pancreatic CT image to obtain a pancreatic threshold segmentation image, such as Figure 3 shown.

[0071] S200: Analyze the gray value features of the pancreas in the pancreas threshold segmentation image, remove the edge interference area of ​​the pancreas in the pancreas threshold segmentation image, and obtain a pancreas interference-free image.

[0072] There are large differences in the overall pixel grayscale values ​​of different organ tissues in the pancreatic threshold segmentation image, and the pancreatic organ has a different morphology from other organs, that is, the pancreatic organ presents a long strip shape. In order to facilitate subsequent research, it is necessary to further process the pancreatic threshold segmentation image, that is, to remove the edge interference part. Therefore, this embodiment analyzes the grayscale value characteristics of the pancreas in the pancreatic threshold segmentation image, removes the edge interference area of ​​the pancreas in the pancreatic threshold segmentation image, and obtains a pancreatic interference-free image, such as Figure 4 shown.

[0073] See also Figure 5 , which shows the basic process of a method for acquiring a pancreatic interference-free image provided by an embodiment of the present invention. Figure 5 As shown, the grayscale value features of the pancreas in the pancreas threshold segmentation image are analyzed, and the edge interference area of ​​the pancreas in the pancreas threshold segmentation image is removed to obtain the pancreas interference-free image, which specifically includes the following steps:

[0074] By observing the pancreatic threshold segmentation image, it is found that the overall grayscale value of the pancreatic region is relatively stable and has a large grayscale value. This feature can be used to screen out points and areas with abnormal grayscale value changes in the pancreatic threshold segmentation image. The specific operations are as follows:

[0075] S201: Extracting pixels in the foreground area of ​​the pancreas threshold segmentation image, and obtaining the mean grayscale value of all pixels in the foreground area.

[0076] Extract all the foreground pixels in the pancreas threshold segmentation image, obtain the grayscale value of each foreground pixel based on the pancreas threshold segmentation image, and then calculate the mean grayscale value of all foreground pixels .

[0077] S202: Perform edge detection on the pancreas threshold segmentation image to obtain a plurality of closed edge contours, and obtain the gray value contour mean of all pixel points within the closed edge contours.

[0078] Edge detection is performed on the pancreas threshold segmentation image to obtain multiple irregular closed edge contours, and the area corresponding to each closed edge contour is marked. It should be noted that the obtained closed edge contour cannot determine which is the pancreatic area, but subsequent judgments are based on the closed edge contour.

[0079] Then, based on the pancreatic threshold segmentation image, the gray value mean of all pixels in the area contained in each closed edge contour is calculated to obtain the gray value contour mean. The difference between the mean gray value contour and the mean gray value of the pixels in the foreground area is calculated as a reference for removing the interference area.

[0080] By observing the pancreas threshold segmentation image, we can find that there are some other interference areas, and the grayscale values ​​of the pixels in the interference area are similar to those in the pancreatic area. The interference area cannot be completely removed according to the above steps. Here, we need to make a judgment based on the difference between the morphological characteristics of the pancreatic area and the morphological characteristics of the surrounding interference areas. The most obvious external feature of the pancreatic area and other interference areas is that the pancreatic area is distributed in long strips. Here, the aspect ratio of the pixels in the closed contour area can be used as a possible judgment indicator for the pancreatic area. The specific steps are as follows:

[0081] S203: Establish a minimum circumscribed rectangle for each closed edge contour, and calculate the aspect ratio of the rectangle.

[0082] Create a minimum enclosing rectangle for each closed edge contour, calculate the aspect ratio of all minimum enclosing rectangles, and get the maximum aspect ratio. Compare the aspect ratio of all minimum enclosing rectangles with the maximum aspect ratio to get the aspect ratio of the rectangle. The calculation formula for constructing the aspect ratio of the rectangle is:

[0083] ;

[0084] In the formula, Indicates The aspect ratio of the rectangle; Indicates The aspect ratio of the minimum circumscribed rectangle corresponding to the closed edge contour; Represents the maximum value of the aspect ratio of all minimum enclosing rectangles. The larger the value is, the greater the possibility that the closed edge contour area is the pancreatic area. The closed edge contour area is composed of all pixel points within the closed edge contour.

[0085] S204: Analyze the difference between the mean gray value contour value and the mean gray value of the pixels in the foreground area, and combine the aspect ratio of the rectangle to obtain the possibility of the pancreatic area.

[0086] Analyze the difference between the gray value contour mean and the gray value mean of the foreground area pixels, specifically calculate the absolute value of the difference between the gray value contour mean and the gray value mean of the foreground area pixels, and then combine the aspect ratio of the rectangle to obtain the pancreatic region possibility. The calculation formula for constructing the pancreatic region possibility is:

[0087] ;

[0088] In the formula, Indicates The possibility that the closed edge contour area is the pancreatic area; Indicates The mean gray value of all pixels in the closed edge contour area; Represents the mean gray value of pixels in the foreground area; Indicates The aspect ratio of the smallest enclosing rectangle; Indicates the maximum value among the aspect ratios of all minimum bounding rectangles.

[0089] Indicates The difference between the mean grayscale value of the pixels in the closed edge contour area and the foreground area, The larger the value, the greater the difference in grayscale value between the closed edge contour area and the foreground pixel area. It can be used to preliminarily remove areas in the pancreas threshold segmentation image where the grayscale value is significantly different from the overall value. The larger the value, the smaller the difference in grayscale value between the closed edge contour area and the foreground pixel area, and the greater the possibility that the closed edge contour area is the pancreatic area; 0.01 is set to avoid the denominator of the fraction being zero. As the aspect ratio, the larger the value is, the greater the probability that the closed edge contour area corresponding to the minimum circumscribed rectangle is the pancreatic area.

[0090] S205: Setting a pancreatic region possibility threshold, and obtaining a pancreatic interference-free image according to the pancreatic region possibility.

[0091] Set the pancreatic region possibility threshold. According to experience, the pancreatic region possibility threshold can be 0.8. According to the pancreatic region possibility, if the pancreatic region possibility If it is smaller than the pancreatic region possibility threshold, it is removed to obtain the pancreatic interference-free image.

[0092] S300: extract edge pixels of the pancreas de-interference image and obtain the direction of the edge pixels, analyze the fluctuation characteristics of the direction of the edge pixels, and obtain the inflection point.

[0093] By observing the pancreas de-interference image, it is found that the pixels in the adhesion area between the pancreas and other organs have certain regular characteristics compared with other normal areas, including: the pancreatic organ has fixed morphological characteristics that are different from other organs; there is a pair of corresponding inflection points in the adhesion area; the line connecting the pair of inflection points passes through the inside of the pancreatic area, and the line can be roughly used as the cutting line between the pancreatic area and other areas; the pixels at the edge of the pancreas located in the adhesion area are related to the trend of the pair of inflection points.

[0094] Because the shape of the human pancreas is a relatively fixed smooth curve contour, its edge pixels have a relatively fixed trend. Therefore, the fluctuation of the direction of its edge pixels is relatively fixed, that is, the number and position of the inflection points where the direction of the pancreatic edge pixels changes are relatively fixed. However, the occurrence of adhesion will cause the inflection points in the adhesion area to cause large fluctuations in the direction of the edge pixels. Due to the flaky characteristics of the adhesion area, such inflection points appear in a pair.

[0095] At the position of this pair of inflection points, the direction of the pancreatic edge pixel changes significantly, and the inflection point is located at the edge of the adhesion area, so the area passed by the line connecting the two inflection points is basically inside the target area extracted from the pancreatic interference-free image, that is, inside the pancreatic area with adhesion. In addition, since the direction of the pixel between the two inflection points is abnormal, the line between the two inflection points can also be used as a dividing line between the pancreatic area and other areas.

[0096] Therefore, based on the above situation, in order to solve the problem of adhesion in the pancreatic area, it is first necessary to locate the location of adhesion in the pancreatic area. This location is determined by the pair of inflection points where the direction of the pixels at the intersection of the adhesion area and the pancreatic edge area changes significantly.

[0097] Based on the above analysis, this embodiment first extracts edge pixels of the pancreas interference-free image and obtains the direction of the edge pixels, analyzes the fluctuation characteristics of the direction of the edge pixels, and obtains inflection points. Based on these inflection points, the inflection points are matched according to the characteristic rules of the inflection points in the adhesion area.

[0098] The specific steps for obtaining the inflection point are as follows: first, extract the edge pixels of the pancreatic interference-free image; then generate the eight-direction chain codes of all edge pixels, and obtain the direction of the edge pixels according to the eight-direction chain codes; then arrange the values ​​of the direction of the edge pixels in the order of the extension of the edge pixels to obtain a direction value sequence; finally, extract the pixels whose values ​​change in the direction value sequence to obtain the inflection point.

[0099] There is a pair of inflection points in the pancreatic edge pixels at the adhesion area, where the direction of the pixels fluctuates greatly. This pair of inflection points is generally located at the edge of the adhesion area, that is, the area where adhesion begins and ends. Then the line connecting this pair of inflection points exists in the adhesion area. This feature is used to screen and match the inflection points. To determine whether the line connecting the inflection points is located in the adhesion area, the proportion of the target pixel points in the pixels passed by the line to all the pixels of the corresponding straight line can be analyzed by connecting the two inflection points.

[0100] The specific steps of obtaining the inflection point of the suspected pancreatic area are as follows: first, the inflection points are combined in pairs to obtain multiple inflection point combinations, and the inflection points are connected in pairs to obtain multiple inflection point lines; then, the pixels on the inflection point lines are extracted as the pixels in the pancreatic interference-free image, and the pixels on the inflection point lines are obtained as the number of pixels in the pancreatic interference-free image, and the ratio of the number of pixels on the inflection point lines to the total number of pixels on the inflection point lines is calculated; finally, a ratio threshold is set. According to experience, the ratio threshold value can be 0.8. If the ratio is greater than 0.8, it is considered that most of the pixels on the inflection point line are pixels of the pancreatic area in the pancreatic interference-free image, that is, the inflection point line is most likely located inside the pancreatic area in the pancreatic interference-free image, and the inflection point lines with a ratio greater than the ratio threshold are retained, and the inflection points of the inflection point lines are extracted to obtain the inflection points of the suspected pancreatic area.

[0101] S400: Based on the inflection point, straight-line cutting is performed on the pancreatic interference-free image to obtain a pair of cutting regions, and grayscale differences and shape feature differences between the two pairs of cutting regions are analyzed to determine the inflection point of the adhesion region and the adhesion region.

[0102] Based on the operation of step S300, the inflection point corresponding to the line passing through the pancreatic area is obtained, but there are individual pixels in the middle of the adhesion area in the pancreatic interference-free image that are not identified during the threshold segmentation, making the adhesion area incompletely continuous. Therefore, the possibility of the inflection point cannot be judged only by the proportion of the pancreatic area pixels in the area where the inflection point pair line passes. In this case, it can be determined whether the pair of inflection points is the inflection point of the adhesion area by judging the degree of fit between the cut area features and the pancreatic shape features. Since the inflection point pair of the adhesion area can cut the pancreas and the adhesion area in the image, if a pair of inflection points is combined into an inflection point pair of the adhesion area, then the image it cuts should conform to the shape-related laws of the pancreas.

[0103] Because one of the two regions segmented by the inflection point of the adhesion region is a pancreatic region and the other is not, the probability value of the pancreatic region is the largest, and the probability value of the non-pancreatic region is small. In this case, the difference in the probability values ​​of the two cut regions being pancreatic regions is the largest. Therefore, this embodiment performs linear cutting on the pancreatic interference-free image based on the inflection point to obtain paired cut regions, and analyzes the grayscale difference and shape feature difference of the two paired cut regions to determine the inflection point of the adhesion region and the adhesion region.

[0104] The specific steps for determining the inflection point and adhesion area of ​​the adhesion area are as follows: connect the inflection points of the suspected pancreatic area in pairs to obtain the connecting line of the inflection points of the suspected pancreatic area; perform linear cutting on the pancreatic interference-free image through the connecting line of the inflection points of the suspected pancreatic area to obtain paired cutting areas; calculate the difference between the mean grayscale value of the pixel points in the two paired cutting areas and the mean grayscale value of all the pixels in the pancreatic interference-free image, and then analyze the difference between the two differences to obtain the grayscale difference of the two paired cutting areas; establish the minimum circumscribed rectangle for the two paired cutting areas respectively, and obtain the aspect ratio of the rectangle, analyze the difference between the two aspect ratios, and obtain the shape feature difference of the two paired cutting areas; obtain the possibility of the inflection point of the adhesion area based on the grayscale difference and the shape feature difference, and determine the inflection point and adhesion area of ​​the adhesion area based on the possibility of the inflection point of the adhesion area.

[0105] The calculation formula for the possibility of constructing the inflection point of the adhesion area is:

[0106] ;

[0107] In the formula, Indicates The possibility of the suspected pancreatic region inflection point being the adhesion region inflection point; Indicates A (first The mean gray value of the pixels in the cutting area; Indicates Another (first) The mean gray value of the pixels in the cutting area; Represents the mean gray value of all pixels in the pancreas interference-free image; Indicates A (first The aspect ratio of the rectangle corresponding to the cutting area; Indicates Another (first) The aspect ratio of the rectangle corresponding to the cutting area; Represents the min-max normalized function.

[0108] and They represent the absolute value of the difference between the mean grayscale value of the pixels in the two cut regions and the mean grayscale value of all pixels in the pancreas interference-free image. It indicates the grayscale difference between two paired cut areas. The larger the grayscale difference is, the greater the possibility that the corresponding suspected pancreatic region inflection point is the adhesion region inflection point. The difference in the aspect ratio of the rectangles corresponding to the two paired cutting areas indicates the difference in the shape characteristics of the two paired cutting areas. The greater the difference in the shape characteristics, the greater the possibility that the corresponding suspected pancreatic region inflection point is the adhesion region inflection point. Normalization is performed and their results are placed in the interval [0,1]. The size of the difference is reflected according to the normalization result, and the size of the difference is used to reflect the possibility that the pair of inflection points are inflection points of the adhesion area.

[0109] Through all the pairs of suspected pancreatic region inflection points, the possibility of all suspected pancreatic region inflection points being adhesion region inflection points can be obtained. If the value is close to 1, it means that it is larger before normalization, that is, the larger the normalized result of the difference in the possibility that the two regions segmented by the pair of suspected pancreatic region inflection points are pancreatic regions, the greater the possibility that the pair of suspected pancreatic region inflection points are adhesion region inflection points. Therefore, the suspected pancreatic region inflection point corresponding to the maximum possibility is determined as the adhesion region inflection point, and the area where the adhesion region inflection point is located is the adhesion region.

[0110] S500: Based on the inflection point of the adhesion area, the distribution characteristics of the pixel points in the adhesion area are analyzed to obtain the possibility that the pixel point is a pixel point at the edge of the pancreas.

[0111] Through step S400, the location of the adhesion area is roughly found, and the possibility of the pancreatic edge pixel point is judged by the distribution law characteristics of the adhesion area pixel points. The adhesion area is not a pixel point, but a part of the area. There are two inflection points on both sides of the adhesion area and the suspected edge of the pancreatic area, and the direction trends of the two inflection points are related to the adhesion area. Then, through this relationship, the possibility of the adhesion area pixel point being the pancreatic edge area pixel point can be further analyzed.

[0112] Here, it cannot be determined that the original edge of the pancreatic region in the adhesion area is a straight line, so the possibility that the pixel point in the adhesion area is a pixel point on the edge of the pancreas cannot be determined simply by connecting the two inflection points and based on the relationship between the connected straight line and other pixel points.

[0113] So, to solve this problem, we can use the vectors formed by the two inflection points and the edge pixels around them respectively (the vector formed by the two inflection points and the vector formed by the edge pixels around them respectively), calculate the distance from the pixel point in the adhesion area to the straight line where the vector is located, and use this distance to reflect the size of the possibility. In this way, a pixel point will get two different sizes of possibilities, and extract the maximum possibility as the possibility size of the pixel point.

[0114] The specific operation is as follows: First, in the edge pixel points of the pancreatic interference-free image, the edge pixel points located between the two adhesion area inflection points and adjacent to the adhesion area inflection points are taken as adjacent pixel points. It should be noted that what is studied here is the regional characteristics between the adhesion area inflection points, so the adjacent pixel points at the located adhesion area inflection points should be located in the edge area between the two adhesion area inflection points. Then form a vector with the two adhesion area inflection points and a vector with the two adjacent pixel points, and obtain the expression of the straight line where the two vectors are located. Based on the two expressions, calculate the distance from the pixel points in the adhesion area to the straight lines where the two vectors are located. According to the distance, the possibility of the pixel point being a pancreatic edge pixel point is obtained. The calculation formula for the possibility of constructing a pixel point being a pancreatic edge pixel point is as follows:

[0115] ;

[0116] In the formula, Indicates the first The probability size of the pixel point being the edge pixel point of the pancreas, where the probability size is based on the Pixels to The distance of the vector from the straight line; Indicates the first Pixels to The distance of the line on which the vector lies; express Vectors include vectors formed by two inflection points of the adhesion area and vectors formed by two adjacent pixels.

[0117] The larger the value is, the smaller the distance between the pixel and the straight line is, and the greater the possibility that the pixel in the adhesion area will become a pixel in the pancreatic edge area.

[0118] Finally, each pixel in the adhesion area has two The maximum value is taken as the probability that the pixel is the edge pixel of the pancreas. If the values ​​are equal, then the value is taken as the probability that the pixel is a pancreatic edge pixel.

[0119] S600: Classify the pixels according to the likelihood to obtain a pancreatic edge pixel sequence.

[0120] There is a certain relationship between the pixels in the adhesion area and the pixels on the edge of the pancreas. Based on this relationship, a probability formula is established between them, and the obtained probability size is used as the basis for judging whether these pixels are the edge of the pancreas. The obtained probability size can not only be used to judge whether the pixels in the adhesion area are the edge of the pancreas, but also as a reference for judging the pixels in other areas.

[0121] Specifically, after the probability of the pixel points in the adhesion area is obtained in step S500, the specific position of the pixel points in the pancreatic edge area can be determined according to the probability. The pixel points can be divided into the pancreatic edge area pixel points and the pancreatic edge outer pixel points according to their different positional features. The pancreatic outer pixel points can be further divided into the middle area pixel points of the two organs and the other organ pixel points.

[0122] Therefore, this embodiment classifies the pixels according to the likelihood to obtain a sequence of pancreatic edge pixels. The specific operation steps are as follows: first, set the pancreatic edge pixel likelihood threshold, which can be set to 0.9, and then classify the pixels into pancreatic edge area pixels and other area pixels according to the likelihood. That is, if the likelihood is If the probability is greater than 0.9, the pixel is considered to be a pixel in the pancreatic edge area. If the value of the pixel is less than or equal to 0.9, the pixel is considered to be a pixel in other regions. Then, the pixel points judged as the edge region of the pancreas are extracted, and these pixel points are connected, so that the contour line formed is the contour composed of the edge pixel points of the pancreas in the adhesion region, that is, the contour of a part of the pixel points blocked due to adhesion; the contour composed of the edge pixel points of the pancreas and the known edge pixel points of the pancreas in other regions of the pancreas are merged to obtain a sequence of edge pixel points of the pancreas; wherein, the edge pixel points of the pancreas in other regions of the pancreas are the edge pixel points of the non-adhesion region of the pancreas divided by a pair of adhesion region inflection points in step S400. In addition, it is determined whether the pixel points of other regions belong to the pixel points of the inner region of the pancreas edge pixel point sequence. If so, the pixel points of other regions are considered to be the pixel points of the pancreas region, and the pixel points of other regions belonging to the inner region of the pancreas edge pixel point sequence are retained; if not, the pixel points of other regions are considered to be the outer pixel points of the pancreas edge. The pixels outside the edge of the pancreas can be further distinguished, that is, according to the probability size, the smaller the probability value, the farther the pixel is from the edge of the pancreas contour. Here, a threshold [0.5, 0.85] (empirical value) is set. If the probability is between [0.5, 0.85], it is considered that the pixel is a pixel in the gap between the pancreas and other organs, that is, a pixel in the middle area between the two organs. If it is less than 0.5, the pixel is considered to be a pixel on other organs. The inner region of the pancreatic edge pixel sequence refers to the region composed of all pixels within the closed edge formed by all pancreatic edge pixels in the pancreatic edge pixel sequence.

[0123] S700: According to the pancreatic edge pixel sequence, the pancreatic interference removal image is combined to improve the pancreatic image area.

[0124] In the above steps, the positions of the pixels in the pancreatic edge region are determined based on the trend of the pixels in the adhesion region of the pancreatic image, thereby obtaining a complete sequence of pixels in the pancreatic edge region.

[0125] For the final imaging effect of the pancreatic CT image, in addition to the complete pancreatic edge area pixel points, the completeness of the internal pixel points of the pancreatic area is also required. Therefore, this embodiment improves the pancreatic image area based on the pancreatic edge pixel point sequence and the pancreatic interference-free image, that is, further adjusts the inaccurately divided area in the pancreatic image under the Otsu threshold segmentation background. Further, the pancreatic edge pixel point sequence and the contour area in the pancreatic interference-free image are compared, the pixel points outside the pancreatic edge pixel point sequence are removed, and the pixel points within the pancreatic edge pixel point sequence are completed to improve the pancreatic image area.

[0126] Based on the same inventive concept as the above method, this embodiment further provides a pancreatic CT image segmentation system, which includes a pancreatic CT image acquisition module, a pancreatic image interference removal module, an adhesion area determination module, a pancreatic edge pixel point determination module, and a pancreatic contour improvement module, wherein:

[0127] A pancreatic CT image acquisition module is used to acquire a pancreatic CT image, perform threshold segmentation on the pancreatic CT image, and obtain a pancreatic threshold segmentation image;

[0128] The pancreatic image interference removal module is used to analyze the gray value characteristics of the pancreas in the pancreatic threshold segmentation image, remove the edge interference area of ​​the pancreas in the pancreatic threshold segmentation image, and obtain the pancreatic interference-free image;

[0129] The adhesion region determination module is used to extract edge pixel points of the pancreatic interference-free image and obtain the direction of the edge pixel points, analyze the fluctuation characteristics of the direction of the edge pixel points, and obtain the inflection point; and based on the inflection point, perform linear cutting on the pancreatic interference-free image to obtain paired cutting regions, and analyze the grayscale difference and shape feature difference of the two paired cutting regions to determine the adhesion region inflection point and the adhesion region;

[0130] The pancreatic edge pixel determination module is used to analyze the distribution characteristics of the pixels in the adhesion area based on the inflection point of the adhesion area, and obtain the probability that the pixel is a pancreatic edge pixel; then classify the pixels according to the probability to obtain a pancreatic edge pixel sequence;

[0131] The pancreatic contour improvement module is used to improve the pancreatic image area based on the pancreatic edge pixel sequence combined with the pancreatic interference removal image.

[0132] Based on the same inventive concept as the above method, the present embodiment further provides a computer-readable storage medium on which a computer program is stored; the computer program is executed by a processor to obtain a pancreatic CT image, perform threshold segmentation on the pancreatic CT image, and obtain a pancreatic threshold segmentation image; analyze the gray value characteristics of the pancreas in the pancreatic threshold segmentation image, remove the edge interference area of ​​the pancreas in the pancreatic threshold segmentation image, and obtain a pancreatic interference-free image; extract the edge pixel points of the pancreatic interference-free image and obtain the direction of the edge pixel points, analyze the fluctuation characteristics of the direction of the edge pixel points, and obtain the inflection point; and based on the inflection point, perform linear cutting on the pancreatic interference-free image to obtain paired cutting areas, and analyze the grayscale difference and shape feature difference of the two paired cutting areas to determine the adhesion area inflection point and the adhesion area; based on the adhesion area inflection point, analyze the distribution characteristics of the pixel points in the adhesion area to obtain the possibility that the pixel point is a pancreatic edge pixel point; then classify the pixel points according to the possibility to obtain a pancreatic edge pixel point sequence; according to the pancreatic edge pixel point sequence, combined with the pancreatic interference-free image, improve the pancreatic image area.

[0133] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0134] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A pancreatic CT image segmentation method, characterized in that: The method comprises: Acquire a pancreatic CT image, and perform threshold segmentation on the pancreatic CT image to obtain a pancreatic threshold segmentation image; Analyzing the grayscale value features of the pancreas in the pancreas threshold segmentation image, removing the edge interference region of the pancreas in the pancreas threshold segmentation image, and obtaining a pancreas interference-free image; Extracting edge pixel points of the pancreas interference-free image and obtaining the direction of the edge pixel points, analyzing the fluctuation characteristics of the direction of the edge pixel points, and obtaining the inflection point; Based on the inflection point, the pancreatic interference-free image is linearly cut to obtain a pair of cut regions, and the grayscale difference and shape feature difference of the two cut regions in a pair are analyzed to determine the adhesion region inflection point and the adhesion region; Based on the inflection point of the adhesion area, analyzing the distribution characteristics of the pixel points in the adhesion area, and obtaining the possibility that the pixel point is a pixel point at the edge of the pancreas; Classifying the pixel points according to the likelihood to obtain a pancreatic edge pixel point sequence; According to the pancreatic edge pixel sequence, combined with the pancreatic interference-free image, the pancreatic image region is improved; Based on the inflection point of the adhesion area, the distribution characteristics of the pixel points in the adhesion area are analyzed to obtain the possibility that the pixel point is a pixel point at the edge of the pancreas, including: In the edge pixel points of the pancreas interference-free image, edge pixel points located between the two adhesion region inflection points and adjacent to the adhesion region inflection points are used as adjacent pixel points; The two inflection points of the adhesion area and the two adjacent pixel points are respectively formed into vectors, and an expression of a straight line where the two vectors are located is obtained; Based on the two expressions, the distances from the pixel points in the adhesion area to the straight lines where the two vectors are located are calculated respectively; According to the distance, the possibility that the pixel point is a pancreatic edge pixel point is obtained.

2. The pancreatic CT image segmentation method according to claim 1, characterized in that: Analyzing the grayscale value features of the pancreas in the pancreas threshold segmentation image, removing the edge interference region of the pancreas in the pancreas threshold segmentation image, and obtaining a pancreas interference-free image, including: Extracting pixels of the foreground area in the pancreas threshold segmentation image, and obtaining the mean grayscale value of all pixels of the foreground area; Performing edge detection on the pancreas threshold segmentation image to obtain a plurality of closed edge contours, and obtaining the gray value contour mean of all pixel points within the closed edge contours; Establishing a minimum circumscribed rectangle for each closed edge contour, and calculating the aspect ratio of the rectangle; Analyze the difference between the mean value of the gray value contour and the mean value of the gray value of the pixels in the foreground area, and combine the aspect ratio of the rectangle to obtain the possibility of the pancreatic area; A pancreatic region possibility threshold is set, and a pancreatic interference-free image is obtained according to the pancreatic region possibility.

3. The pancreatic CT image segmentation method according to claim 1, characterized in that: Extracting edge pixel points of the pancreas interference-free image and obtaining the direction of the edge pixel points, analyzing the fluctuation characteristics of the direction of the edge pixel points, and obtaining the inflection points, including: Extracting edge pixel points of the pancreas de-interference image; Generate eight-direction chain codes for all edge pixels, and obtain the directions of the edge pixels according to the eight-direction chain codes; Arrange the values ​​of the directions of the edge pixels in the order of extension of the edge pixels to obtain a direction value sequence; Pixel points whose values ​​change in the direction value sequence are extracted to obtain inflection points.

4. The pancreatic CT image segmentation method according to claim 1, characterized in that: Based on the inflection point, the pancreas interference-free image is linearly cut, which also includes: The inflection points are combined in pairs and connected to obtain an inflection point connection line; Calculating the number of pixels on the inflection point line as the ratio of the number of pixels in the pancreas interference-free image to the total number of pixels on the inflection point line; A ratio threshold is set, and the inflection point connection line whose ratio is greater than the ratio threshold is retained to obtain the suspected inflection point of the pancreatic region.

5. The pancreatic CT image segmentation method according to claim 4, characterized in that: Based on the inflection point, the pancreatic interference-free image is linearly cut to obtain a pair of cut regions, and the grayscale difference and shape feature difference of the two cut regions in a pair are analyzed to determine the adhesion region inflection point and the adhesion region, including: Connecting the suspected pancreatic region inflection points in pairs to obtain a connecting line of the suspected pancreatic region inflection points; Performing linear cutting on the pancreatic interference-free image through the connecting line of the inflection points of the suspected pancreatic regions to obtain paired cutting regions; Calculating the difference between the grayscale value mean of the pixel points in the two paired cutting areas and the grayscale value mean of all the pixel points in the pancreatic interference-free image, and then analyzing the difference between the two differences to obtain the grayscale difference of the two paired cutting areas; Establishing minimum circumscribed rectangles for the two cutting areas in a pair, respectively, and obtaining the aspect ratio of the rectangles, analyzing the difference between the two aspect ratios, and obtaining the difference in shape characteristics of the two cutting areas in a pair; According to the grayscale difference and the shape feature difference, the adhesion region inflection point and the adhesion region are determined.

6. The pancreatic CT image segmentation method according to claim 1, characterized in that: According to the likelihood, the pixel points are classified to obtain a pancreatic edge pixel point sequence, including: Setting a pancreatic edge pixel possibility threshold, and classifying the pixel into pancreatic edge area pixel and other area pixel according to the possibility; Merging the pancreatic edge region pixel points with known pancreatic edge region pixel points in other regions of the pancreas to obtain a pancreatic edge pixel point sequence; It is determined whether the pixel points in the other regions belong to the pixel points in the inner region of the pancreatic edge pixel point sequence, and the pixel points in the other regions that belong to the inner region of the pancreatic edge pixel point sequence are retained.

7. The pancreatic CT image segmentation method according to claim 1, characterized in that: According to the pancreatic edge pixel sequence, combined with the pancreatic interference-free image, the pancreatic image region is improved, including: The pancreatic edge pixel sequence and the contour area in the pancreatic interference-free image are compared, and the pixels outside the pancreatic edge pixel sequence are removed, and the pixels within the pancreatic edge pixel sequence are completed to improve the pancreatic image area.

8. A pancreatic CT image segmentation system, characterized in that: The system comprises: A pancreatic CT image acquisition module is used to acquire a pancreatic CT image, perform threshold segmentation on the pancreatic CT image, and obtain a pancreatic threshold segmentation image; A pancreatic image interference removal module is used to analyze the gray value characteristics of the pancreas in the pancreatic threshold segmentation image, remove the edge interference area of ​​the pancreas in the pancreatic threshold segmentation image, and obtain a pancreatic interference-free image; An adhesion region determination module is used to extract edge pixel points of the pancreatic interference-free image and obtain the direction of the edge pixel points, analyze the fluctuation characteristics of the direction of the edge pixel points, and obtain the inflection point; and based on the inflection point, perform linear cutting on the pancreatic interference-free image to obtain paired cutting regions, and analyze the grayscale difference and shape feature difference of the two paired cutting regions to determine the adhesion region inflection point and the adhesion region; A pancreatic edge pixel determination module is used to analyze the distribution characteristics of the pixels in the adhesion area based on the inflection point of the adhesion area to obtain the probability that the pixel is a pancreatic edge pixel; and then classify the pixel according to the probability to obtain a pancreatic edge pixel sequence; A pancreatic contour improvement module, used to improve the pancreatic image area according to the pancreatic edge pixel sequence in combination with the pancreatic interference-free image; Based on the inflection point of the adhesion area, the distribution characteristics of the pixel points in the adhesion area are analyzed to obtain the possibility that the pixel point is a pixel point at the edge of the pancreas, including: In the edge pixel points of the pancreas interference-free image, edge pixel points located between the two adhesion region inflection points and adjacent to the adhesion region inflection points are used as adjacent pixel points; The two inflection points of the adhesion area and the two adjacent pixel points are respectively formed into vectors, and an expression of a straight line where the two vectors are located is obtained; Based on the two expressions, the distances from the pixel points in the adhesion area to the straight lines where the two vectors are located are calculated respectively; According to the distance, the possibility that the pixel point is a pancreatic edge pixel point is obtained.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon; the computer program is executed to implement a pancreatic CT image segmentation method as claimed in any one of claims 1 to 7.

Images