Three-dimensional reconstruction system for visceral fat of obese patients based on CT image

Abstract

This invention relates to the field of image segmentation technology, specifically to a three-dimensional reconstruction system for visceral fat in obese patients based on CT images. The system includes a memory and a processor. The processor executes a computer program stored in the memory to perform the following steps: acquiring CT images of the visceral region of an obese patient; filtering normal and abnormal frame images based on differences in fat area between adjacent images; determining candidate regions for the presence of visceral fat in each abnormal frame image based on the distribution of edge lines of the fat region in the nearest normal frame image to each abnormal frame image; obtaining edge segments of visceral fat based on the positions of pixels in the neighborhood of the edge lines within the candidate regions; filling in the edge segments to obtain the target edge lines of the visceral fat, thereby performing three-dimensional reconstruction of the visceral fat in the obese patient. This invention improves the accuracy of CT image segmentation results and can provide more reliable data for three-dimensional reconstruction of visceral fat.

Description

Technical Field

[0001] This invention relates to the field of image segmentation technology, and more specifically to a three-dimensional reconstruction system for visceral fat in obese patients based on CT images. Background Technology

[0002] Traditional CT images can only observe two-dimensional information in a plane, while three-dimensional reconstruction technology can display the structure in multiple CT images in the form of 3D modeling, making the information obtained more three-dimensional. As body fat increases, obese patients will have a much higher probability of developing cardiovascular disease, type 2 diabetes, metabolic syndrome and other diseases. Three-dimensional reconstruction of visceral fat in obese patients can obtain the distribution and size of body fat, which is an important indicator for evaluating the effect of weight loss and also a guide for surgical procedures.

[0003] In the abdominal cavity, the anterior abdominal wall muscles, posterior abdominal wall muscles, and lumbar muscles form a closed ring-shaped dividing line to distinguish between visceral fat and subcutaneous fat. However, in the area below the arcuate line and above the pubic symphysis, the rectus abdominis muscle, which was originally in front, is replaced by the abdominal wall boundary. The abdominal wall boundary is extremely thin, especially in obese patients. Due to the compression of abdominal fat, the abdominal wall boundary becomes a blurred boundary in the CT images of the visceral region of obese patients. This causes subcutaneous fat and visceral fat to be classified as the same type during image segmentation, affecting the accuracy of three-dimensional reconstruction of visceral fat. Summary of the Invention

[0004] To address the problem that existing methods suffer from biases when segmenting CT images of visceral regions in obese patients, affecting the accuracy of three-dimensional reconstruction results of visceral fat, the present invention aims to provide a three-dimensional reconstruction system for visceral fat in obese patients based on CT images. The specific technical solution adopted is as follows: This invention provides a three-dimensional reconstruction system for visceral fat in obese patients based on CT images. The system includes a memory and a processor, the processor executing a computer program stored in the memory to achieve the following steps: Acquire multi-frame CT images of the visceral regions of obese patients; Based on the difference in fat area between two adjacent frames, normal and abnormal frames are screened; based on the edge distribution of fat regions in the normal frames closest to each abnormal frame, candidate regions for the presence of visceral fat in each abnormal frame are determined. Based on the location of the pixels in the neighborhood of the edge line within the candidate region, the edge segments of visceral fat are obtained; the edge segments are filled using the minimum cost path method to obtain the target edge line of visceral fat. Three-dimensional reconstruction of visceral fat in obese patients is performed based on the location of the target edge line of visceral fat.

[0005] Preferably, the step of filtering normal and abnormal frame images based on the difference in fat area between two adjacent frames includes: All acquired images are arranged to obtain an image sequence; in the image sequence, an image segmentation error factor is obtained based on the difference in fat area between each frame and its neighboring frames. If the image segmentation error factor is less than the preset segmentation error threshold, then the fat area in the image sequence is curve-fitted, and based on the difference between the fat area in the image sequence and the curve fitting result, normal frame images and abnormal frame images are obtained.

[0006] Preferably, determining the candidate regions for the presence of visceral fat in each abnormal frame image based on the edge line distribution of the fat region in the nearest normal frame image includes: For any anomalous frame image: In the image sequence, the nearest normal frame image before and the nearest normal frame image after any abnormal frame image are found respectively, and used as two reference frame images for any abnormal frame image. The outermost edge lines of two reference frame images are obtained by using the watershed algorithm, and the outermost edge lines of the two reference frame images are mapped to any abnormal frame image to obtain the first edge line and the second edge line. The region between the first edge line and the second edge line is considered as a candidate region for the presence of visceral fat in any of the abnormal frame images.

[0007] Preferably, obtaining the edge segment of visceral fat based on the location of pixels in the neighborhood of the edge line within the candidate region includes: A sliding window of a preset size is slid along the edge line of the candidate area. When there is a pixel on the first edge line or a pixel on the second edge line in the area corresponding to the sliding window, the center point of the area corresponding to the sliding window is marked, and the difference weight of the marked pixel is determined according to the relative position of the marked pixel with the first edge line and the second edge line. Based on the difference in the number of pixels on each edge line within the candidate region and the marked pixels, the shortest distance between the marked pixels and the first and second edge lines, and the difference weight of the marked pixels, the difference index corresponding to each edge line is obtained. By comparing the magnitude of the difference index corresponding to different edge lines within the candidate region, edge segments of visceral fat are screened.

[0008] Preferably, determining the difference weight of the marked pixel based on its relative position to the first edge line and the second edge line includes: If the marked pixel does not intersect with the first edge line and the second edge line, then the difference weight of the marked pixel is set to a preset first value; If a marked pixel intersects with one of the first and second edge lines, then the difference weight of the marked pixel is set to a preset second value. If a marked pixel intersects both the first edge line and the second edge line, then the difference weight of the marked pixel is set to a preset third value. Among them, the first preset value is greater than the second preset value, and the second preset threshold is greater than the third preset threshold.

[0009] Preferably, the step of obtaining the difference index corresponding to each edge line based on the difference in the number of pixels on each edge line within the candidate region and the marked pixels, the shortest distance between the marked pixels and the first edge line and the second edge line, and the difference weight of the marked pixels includes: For any edge line within the candidate region: Calculate the first product of the difference weight of each marked pixel on any edge line and the corresponding shortest distance; By combining the first product corresponding to each marked pixel on any edge line, the corresponding quantity difference, and the average number of pixels on the first edge line and the second edge line, the difference index corresponding to any edge line is determined.

[0010] Preferably, the step of comparing the magnitudes of the difference indices corresponding to different edge lines within the candidate region to screen edge segments of visceral fat includes: Among all edge lines in the candidate region whose length is greater than a preset length threshold, the edge line with the smallest difference index is taken as the edge segment of visceral fat.

[0011] Preferably, the step of filling in the edge segments using the minimum cost path method to obtain the target edge line of visceral fat includes: A cost function is constructed for the pixels in the candidate region. The two endpoints of the edge segment of visceral fat are taken as the start and end points respectively. The minimum cost path method is used to find the line connecting the start and end points to obtain the target edge line of visceral fat.

[0012] Preferably, the step of using the minimum cost path method to find the line connecting the starting point and the ending point to obtain the target edge line of visceral fat includes: If there is only one path that satisfies the minimum cost, then the edge segment of visceral fat is filled in using the path that satisfies the minimum cost to obtain the target edge line of visceral fat. If there is more than one path that satisfies the minimum cost, the path that satisfies the minimum cost is recorded as the initial path. The deviation value of each initial path is evaluated based on the first difference between the total number of pixels on the edge line of the visceral fat target and the average number of pixels on the first edge line and the second edge line, and the second difference between the area of ​​the closed region enclosed by each initial path and the edge line of the visceral fat target and the average area of ​​the closed region enclosed by the first edge line and the second edge line. The edge segments of visceral fat are filled in using the initial path with the smallest deviation value to obtain the target edge line of visceral fat.

[0013] Preferably, the evaluation of the deviation value for each initial path includes: The weighted sum of the normalized result of the first difference and the normalized result of the second difference corresponding to each initial path is used as the deviation value of each initial path.

[0014] The present invention has at least the following beneficial effects: This invention effectively filters out normally segmented and abnormally segmented images based on the difference in fat area between two adjacent CT images of the visceral region of obese patients. Further, for each abnormal frame image, the edge distribution of the fat region in the nearest normal frame image is used to accurately determine candidate regions where visceral fat exists in the abnormal frame. Edge segments of visceral fat are extracted based on the neighboring pixel positions of the edge lines within the candidate regions, and these edge segments are filled using the minimum cost path method, thereby obtaining accurate and complete target edge lines of visceral fat. Finally, based on these target edge lines, high-precision three-dimensional reconstruction of visceral fat in obese patients is achieved. The method provided in this embodiment improves the ability to distinguish between visceral fat and subcutaneous fat in areas with blurred abdominal wall boundaries, overcoming the missegmentation problem caused by unclear boundaries in traditional segmentation methods. This improves the accuracy and reliability of three-dimensional reconstruction of visceral fat, providing a more reliable data foundation for clinical assessment of fat distribution, volume measurement, and surgical navigation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the method performed by a CT image-based three-dimensional reconstruction system for visceral fat in obese patients, as provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of a three-dimensional reconstruction system for visceral fat in obese patients based on CT images, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a three-dimensional reconstruction system of visceral fat in obese patients based on CT images, provided by the present invention.

[0020] Example of a 3D reconstruction system for visceral fat in obese patients based on CT images: The specific scenario addressed in this embodiment is as follows: When performing three-dimensional reconstruction of visceral fat in obese patients, multiple frames of CT images of the visceral region of obese patients are first acquired. By combining the accurate division results of different regions in the CT images, three-dimensional reconstruction of visceral fat is achieved.

[0021] This embodiment proposes a three-dimensional reconstruction system for visceral fat in obese patients based on CT images, such as... Figure 1 As shown, the method performed by the CT image-based visceral fat three-dimensional reconstruction system for obese patients in this embodiment includes the following steps: Step S1: Acquire multiple CT images of the visceral regions of obese patients.

[0022] First, lateral X-ray localization images of the abdominal region were acquired in obese patients. Axial CT images of the abdomen were then acquired using a step-through CT scanning system with a slice thickness of 1 mm. The images from the axial CT images and the lateral X-ray localization images of the abdominal region were registered based on the pubic symphysis region. CT images of the region from the arcuate line to the lower edge of the fifth lumbar vertebra were used for subsequent analysis. Considering the potential presence of background in the acquired CT images, a semantic segmentation network was used to segment each acquired CT image, extracting images of the visceral region, resulting in multi-frame CT images of the visceral region of the obese patient. These images will be further analyzed and processed later. The semantic segmentation network used is an existing neural network, which will not be elaborated upon here.

[0023] Step S2: Based on the difference in fat area between two adjacent frames, normal and abnormal frames are filtered out; based on the edge distribution of the fat region in the normal frame closest to each abnormal frame, candidate regions for the presence of visceral fat in each abnormal frame are determined.

[0024] Considering that obese patients have abdominal fat deposits that cause visceral fat and subcutaneous fat to be inaccurately distinguished during 3D reconstruction of the area below the arcuate line in some CT images due to blurred abdominal wall boundaries, it is important to understand that these factors can lead to the problem of abdominal fat being indistinguishable from subcutaneous fat.

[0025] In the abdominal region, the closed curve formed by the posterior wall of the muscle group is the natural boundary between visceral fat and subcutaneous fat. Once this curve is determined, the area inside the curve is the location of visceral fat. The problem of the region is transformed into the problem of calculating the edge. Similarly, in the local blurry area of ​​the muscle group, as long as the edge of this area can be calculated, the incorrectly segmented fat area can be repaired.

[0026] Three-dimensional reconstruction of visceral fat typically selects the area from the upper edge of the first lumbar vertebra to the lower edge of the fifth lumbar vertebra, which is the abdominal region. This is the main area of ​​fat deposition. The first step in three-dimensional reconstruction of visceral fat is to accurately segment the visceral fat region in the CT images. This can be achieved through anatomy. The abdominal muscle groups form continuous closed curves; the area within the abdominal muscle groups is the visceral region, and the area outside the abdominal muscle groups is subcutaneous fat. However, the posterior abdominal sheath disappears below the arcuate line to the lower edge of the fifth lumbar vertebra. Especially in obese patients, excessive accumulation of abdominal fat and muscle stretching can lead to blurred local boundaries. When segmenting visceral fat using medical software, subcutaneous fat may also be included due to these blurred boundaries, resulting in a significantly larger pixel area for the segmented region. In this embodiment, the anatomical structure of the human body is continuous and smooth, and the slice thickness of the CT scan is 1 mm; therefore, structural abrupt changes between adjacent slices can be ignored.

[0027] All acquired CT images are arranged in the order from the arcuate line to the lower edge of the fifth lumbar vertebra to obtain an image sequence; in the image sequence, the image segmentation error factor is obtained based on the difference in fat area between each frame and the adjacent frames.

[0028] As a concrete example, the specific formula for calculating the image segmentation error factor is given. The image segmentation error factor can be expressed as: in, Indicates the image segmentation error factor. This indicates the total number of frames of the acquired CT images. This represents the fat area in the i-th frame of the image sequence. This represents the fat area in the (i+1)th frame of the image sequence. This represents the fat area in the (i-1)th frame of the image sequence. This represents the normalization function, which maps function values ​​to the range [0, 1].

[0029] Used to represent the difference in fat area between the i-th frame and the (i+1)-th frame. This value is used to represent the difference in fat area between the i-th frame and the (i-1)-th frame. The larger the value, the greater the difference in fat area between the two frames.

[0030] It should be noted that since there are no images before the first frame of the image sequence and no images after the last frame of the image sequence, for the first frame, the fat area in the next frame (i.e., the second frame) is used as the area of ​​the fat region in the frame preceding the first frame, and is included in the calculation of the image segmentation error factor. For the last frame, the fat area in the previous frame (i.e., the (M-1)th frame) is used as the area of ​​the fat region in the frame following the last frame, and is included in the calculation of the image segmentation error factor.

[0031] The image segmentation error factor was determined using the above method. A larger image segmentation error factor indicates a higher probability of incorrect image segmentation. Therefore, if the image segmentation error factor is greater than or equal to a preset segmentation error threshold, it indicates an excessively high error rate in image segmentation, requiring manual intervention. In this embodiment, the preset segmentation error threshold is 0.6. In specific applications, the implementer can set this threshold according to the specific circumstances. If the image segmentation error factor is less than the preset segmentation error threshold, curve fitting is performed on the fat area within the images in the image sequence to obtain a fitting curve. The horizontal axis of the fitting curve represents the order of the images in the image sequence, and the vertical axis represents the fat area within the images. The vertical axis corresponding to the order of each image in the fitting curve is used as the area fitting value. The absolute value of the difference between the fat area within each image in the image sequence and its corresponding area fitting value is used as the area residual for each image. In this embodiment, subsequent processing only applies to images where the image segmentation error factor is less than the preset segmentation error threshold. It should be noted that the area in this embodiment refers to the number of pixels.

[0032] After determining the area residual corresponding to each image, the images in the image sequence are classified based on the area residual, thereby identifying the images that are accurately segmented and those that are not. The segmentation results of the inaccurately segmented images are then corrected to obtain accurate segmentation results.

[0033] As a concrete example, the area residuals of all frames are arranged according to their order in the image sequence to obtain an area residual sequence. The K-means clustering algorithm is then used to cluster the elements in the area residual sequence. When using the K-means algorithm, the value of k is set to 2, forming two clusters. Images in the cluster with the smallest average area residual are designated as normal frames, and images in the cluster with the smallest average area residual are designated as abnormal frames. The K-means clustering algorithm is existing technology and will not be elaborated further here.

[0034] Using the above method, all images are divided into two categories: normal frame images and abnormal frame images. Next, the segmentation results of abnormal frame images are corrected by combining the segmentation results of normal frame images to improve the accuracy of the segmentation results.

[0035] In an image sequence, the change in the area of ​​visceral fat in the image should be continuous. If the area of ​​visceral fat in an image is significantly larger than the adjacent visceral fat areas above and below, then it can be considered that the image has made an error in segmentation, including subcutaneous fat. Once the images with segmentation errors appear consecutively, it is difficult to repair them. Therefore, the amount of change between adjacent elements in the sequence can be used to determine whether the images with segmentation errors appear consecutively.

[0036] The Canny edge detection algorithm was used to detect the edge lines in the candidate regions where visceral fat exists in each frame of the image. The Canny edge detection algorithm is an existing technology and will not be described in detail here.

[0037] The following explanation uses an abnormal frame image as an example. Other abnormal frame images can be processed using the method provided in this embodiment.

[0038] For any anomalous frame image: In the image sequence, find the nearest normal frame image before and after the anomalous frame image as two reference frames for the anomalous frame image. Use the watershed algorithm to obtain the outermost edge lines of these two reference frames. Map these outermost edge lines onto the anomalous frame image, and denote the two lines projected onto it as the first edge line and the second edge line. The first and second edge lines represent the boundary between visceral fat and abdominal muscle, with cutouts inside the mask caused by the separation of internal organs. Normally, the first and second edge lines should be closed curves. If either the first or second edge line is not a closed edge line, connect the nearest endpoints of the two broken curve segments with a straight line to form a closed edge line. The first and second edge lines mentioned below are all closed edge lines. The segmented region of this abnormal frame image includes visceral fat and subcutaneous fat. It differs from the normally segmented region in that the visceral fat is surrounded by a continuous, closed ring of subcutaneous fat. In areas with blurred boundaries, the two regions are connected. Edge detection yields edges representing various structures including subcutaneous fat, visceral fat, internal organs, and bones. The region between the first and second edge lines is considered a candidate region for the presence of visceral fat in this abnormal frame image.

[0039] Step S3: Based on the location of the pixels in the neighborhood of the edge line within the candidate region, obtain the edge segment of visceral fat; fill the edge segment using the minimum cost path method to obtain the target edge line of visceral fat.

[0040] The visceral fat region in a CT image is a cross-section of the three-dimensional visceral fat region. The surface of the three-dimensional visceral fat region is smooth, so the cross-sectional changes in the thickness of adjacent layers are also smooth. The size of the visceral fat region in adjacent consecutive layers will show an overall increase or decrease. Therefore, the size of the visceral fat region in the i-th image should be the area between the first edge and the second edge.

[0041] In the edge connecting visceral fat and subcutaneous fat, the pixels are judged. The edge pixels belonging to visceral fat are in the candidate area where visceral fat exists. The human body structure between adjacent axial images is continuous. If a pixel in the area belongs to the edge of visceral fat, the distance between it and the first edge line and the second edge line should be relatively short.

[0042] The segmented image is incorrectly segmented in the blurred boundary area, but the edge detection results in other areas are accurate. Moreover, the morphological changes of muscles between adjacent images are continuous. In the vertically adjacent CT images, the edges formed by the abdominal muscle groups and visceral fat should also be similar in length and highly similar in position. Therefore, the pixel edges with high accuracy can be screened out by the first edge line and the second edge line. The higher the accuracy of the pixel edge, the closer it is to the first edge line and the second edge line in length, and the higher the similarity of the pixel position distribution with the first edge line and the second edge line.

[0043] The above-mentioned edge detection method obtains the edge lines within the candidate region where visceral fat is present. These edge lines are not necessarily closed. For each edge line within the candidate region where visceral fat is present, a sliding window of a preset size is used to slide along the edge line within the candidate region. During the sliding, the center point of the sliding window is always a pixel on the edge line within the candidate region where visceral fat is present. In this embodiment, the size of the sliding window is [missing information]. In specific applications, implementers can configure the settings according to specific circumstances. When there are pixels on either the first or second edge line within the area corresponding to the sliding window, the center point of the area corresponding to the sliding window is marked to obtain multiple marked pixels, which are pixels on the edge line. The difference weight of the marked pixels is determined based on their relative positions to the first and second edge lines. Specifically, if the marked pixel does not intersect with either the first or second edge line, the difference weight is a preset first value; if the marked pixel intersects with either the first or second edge line, the difference weight is a preset second value; if the marked pixel intersects with both the first and second edge lines, the difference weight is a preset third value. The preset first value is greater than the preset second value, and the preset second value is greater than the preset third value. In this embodiment, the preset first value is 0.6, the preset second value is 0.3, and the preset third value is 0.1.

[0044] Using the above method, multiple marked pixels can be obtained on each edge line of the candidate region where visceral fat exists. In the following embodiment, we will take one edge line in the candidate region as an example for explanation. Other edge lines in the candidate region can be processed using the method provided in this embodiment.

[0045] Specifically, for any edge line within the candidate region: The shortest distance between each marked pixel on the edge line and the first and second edge lines is obtained and denoted as the shortest distance for each marked pixel. The absolute value of the difference between the total number of pixels on the edge line in the candidate region and the number of marked pixels on the edge line is denoted as the quantity difference for the edge line. The product of the difference weight of each marked pixel on the edge line and its corresponding shortest distance is calculated and denoted as the first product. The difference index for the edge line is determined by combining the first products for all marked pixels on the edge line, the quantity difference for the edge line, and the average number of pixels on the first and second edge lines.

[0046] As an example, a specific formula for calculating the difference index is given. The difference index corresponding to the edge line can be expressed as: in, This indicates the difference index corresponding to the edge line. This represents the average number of pixels on the first edge line and the second edge line. This indicates the number of marked pixels on the edge line. This represents the difference weight of the i-th marked pixel on the edge line. This represents the shortest distance corresponding to the i-th marked pixel on the edge line. This indicates the quantity difference corresponding to the edge line.

[0047] This represents the first product corresponding to the i-th marked pixel on the edge line, and the distance is adjusted using difference weights. When the marked pixel on the edge line does not intersect with either the first or second edge line, the similarity between the marked pixel's location and the visceral fat edge is low. However, when the marked pixel intersects with either the first or second edge line, or is located between the two edges, the distance from the marked pixel to either edge is the shortest distance through that point, and these two distances are always the same. Points intersecting with the first and second edge lines obviously have higher positional similarity, so difference weights are needed to adjust the distance cost at intermediate points. The larger the first product corresponding to the marked pixels on the edge line, and the greater the difference in the number of points corresponding to the edge line, the less likely the edge line is to be the edge line of visceral fat; that is, the larger the difference index corresponding to the edge line.

[0048] The above method allows us to determine the difference index corresponding to each edge line within the candidate region where visceral fat exists. Next, by comparing the magnitude of the difference index corresponding to different edge lines within the candidate region, edge segments of visceral fat are selected. Specifically, among all edge lines within the candidate region whose length is greater than a preset length threshold, the edge line with the smallest difference index is selected as the edge segment of visceral fat. In this embodiment, the preset length threshold is the length of the first edge line. In specific applications, implementers can configure it according to the specific circumstances.

[0049] In the above steps, the edge fragments of visceral fat were screened. The screened edge fragments of visceral fat are more likely to be non-closed edge lines, so they need to be filled in to obtain complete edge lines.

[0050] In adjacent CT images, in addition to the continuity of the position of human structures, the cross-sectional area of ​​organs is also continuous. The slice thickness interval between adjacent images is 0.1 mm. The organs contained are the same. Since the cross-sectional area of ​​organs is continuous between adjacent images, the area of ​​visceral fat filling between organs should also be continuous. The specific shape of the discontinuous region lines can be determined by limiting the area of ​​the completed region in the i-th image.

[0051] In the area from below the arcuate line to the pubic symphysis, the muscle groups in the lower abdomen show a general tendency to contract inward. The edge and size of visceral fat in CT images should be between two adjacent images. Therefore, for discontinuous areas, the true edge should be in the area between the first and second edge lines.

[0052] The two endpoints of the edge segment of visceral fat are the locations where the curve discontinuities. Using these two endpoints as the start and end points, a cost function is constructed for the pixels within the candidate region. The minimum cost path method is then used to find the line connecting the start and end points, thus obtaining the target edge line of the visceral fat. The cost function can be expressed as: in, The value of the cost function. This represents the gradient magnitude after pixel normalization. This is the minimum distance between the pixel and the first and second edge lines. This represents the balance coefficient.

[0053] When normalizing the gradient magnitude of pixels, the maximum-minimum normalization method is used. This method is existing technology and will not be elaborated upon further here. In this embodiment, the balance coefficient is set to 0.1.

[0054] When selecting pixels, they should be points with edge marks, that is, points with large gradient magnitudes. The closer the normalized gradient magnitude is to 1, the lower the cost of selection. Therefore, the calculation... The selected path represents the edge generated by the abdominal wall boundary, so it should be the closest to the edge generated by the abdominal wall boundary in the reference frame image. The larger the value, the greater the cost of selecting that pixel.

[0055] Using the above method, a path that satisfies the minimum cost can be obtained. However, considering that there may be more than one path that satisfies the minimum cost, the closed curve formed by the posterior wall of the abdominal muscle group is the edge of visceral fat. This curve is elliptical in general, but its curvature is irregular in detail. Therefore, when filling in the edge of visceral fat, it cannot be treated as a regular circle. The filled edge should be a curved curve. The edge obtained by using the first edge line and the second edge line as boundaries is similar in shape to the adjacent layers, but there are deviations in detail. These deviations will disrupt the continuity of the filled curve in terms of length and area. In adjacent CT images, because the changes in the length and cross-sectional area of ​​the outer edge of the same structure are continuous, the more realistic the filled curve is, the smaller the deviation in length and area of ​​the closed curve formed by it and the edge segment of visceral fat from the first edge line and the second edge line will be.

[0056] Based on the above characteristics, if there is only one path that satisfies the minimum cost, then the edge segment of visceral fat is filled using the path that satisfies the minimum cost, thus obtaining a complete closed line segment, which is recorded as the target edge line of visceral fat. If there is more than one path that satisfies the minimum cost, then the path that satisfies the minimum cost is recorded as the initial path. The average number of pixels on the first edge line and the second edge line is calculated, and the absolute value of the difference between the number of pixels on each initial path and the average value is recorded as the first difference of each initial path. The average area of ​​the closed region enclosed by the first edge line and the closed region enclosed by the second edge line is calculated, and this average value is recorded as the average area. The absolute value of the difference between the area of ​​the closed region enclosed by each initial path and the target edge line of visceral fat and the average area is recorded as the second difference of each initial path. The weighted sum of the normalized result of the first difference and the normalized result of the second difference corresponding to each initial path is used as the deviation value of each initial path. In this embodiment, when performing a weighted summation of the normalized results of the first difference and the second difference, the weights of both the normalized results of the first and second differences are 0.5. As a specific example, the first and second differences can be normalized as follows: the average number of pixels on the first and second edge lines is recorded as the average pixel count; the ratio of the first difference to the average pixel count is used as the normalized result of the first difference; and the ratio of the second difference to the aforementioned average area is used as the normalized result of the second difference. Alternatively, other existing data normalization methods can be used, such as the maximum-minimum value normalization method, which will not be elaborated upon here.

[0057] Using the above method, the deviation value of each initial path can be obtained. The edge segment of visceral fat is filled in using the initial path with the smallest deviation value, which is to obtain a complete closed line segment. This line segment is recorded as the target edge line of visceral fat.

[0058] It should be noted that if the edge segment of visceral fat is a closed curve, then it is not necessary to fill it using the above filling method; the edge segment can be directly used as the target edge line of visceral fat.

[0059] Step S4: Perform three-dimensional reconstruction of visceral fat in obese patients based on the location of the target edge line of visceral fat.

[0060] In the above steps, the target edge line of visceral fat was obtained. The target edge line of visceral fat is the accurate edge line of visceral fat in the abnormal frame image, which means that the segmentation result of the abnormal frame image is corrected and the accurate segmentation result of each CT image is obtained.

[0061] Next, based on the segmentation results of each CT image, a three-dimensional reconstruction of the visceral fat of obese patients is performed. The specific process of three-dimensional reconstruction is as follows: (1) Confirm that the visceral fat region segmentation results of all thicknesses are obtained, and the segmentation results have only two values: 1 represents the white visceral fat part and 0 represents the black background. (2) Ensure that all two-dimensional images are correctly aligned in three-dimensional space, and transfer the aligned data to the reconstruction software in DICOM format; (3) Generate a surface model of the visceral fat region and use the Laplacian algorithm to smooth the surface without changing the overall shape of the model, so that the model is more in line with the real shape of the visceral fat region. (4) Render the surface model to generate an observable 3D model, calculate the fat volume in the model, or analyze the fat thickness distribution. Among these, 3D reconstruction is an existing technology and will not be elaborated on further here.

[0062] Thus, by using the above method, three-dimensional reconstruction of visceral fat was achieved, making the final three-dimensional reconstruction results more realistic and the data more accurate when analyzing the volume of visceral fat and the ratio of visceral fat to subcutaneous fat volume.

[0063] This embodiment effectively filters out normally segmented and abnormally segmented images based on the difference in fat area between two adjacent CT images of the visceral region of obese patients. Further, for each abnormal frame image, the distribution of the edge lines of the fat region in the nearest normal frame image is used to accurately determine the candidate regions where visceral fat exists in the abnormal frame. Edge segments of visceral fat are extracted based on the neighboring pixel positions of the edge lines within the candidate regions, and these edge segments are filled using the minimum cost path method, thereby obtaining accurate and complete target edge lines of visceral fat. Finally, based on these target edge lines, high-precision three-dimensional reconstruction of visceral fat in obese patients is achieved. The method provided in this embodiment improves the ability to distinguish between visceral fat and subcutaneous fat in areas with blurred abdominal wall boundaries, overcoming the missegmentation problem caused by unclear boundaries in traditional segmentation methods. This improves the accuracy and reliability of three-dimensional reconstruction of visceral fat, providing a more reliable data foundation for clinical assessment of fat distribution, volume measurement, and surgical navigation.

[0064] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A CT image-based visceral fat three-dimensional reconstruction system for obese patients, characterized by, The system includes a memory and a processor, the processor executing a computer program stored in the memory to perform the following steps: Acquire multi-frame CT images of the visceral regions of obese patients; Based on the difference in fat area between two adjacent frames, normal and abnormal frames are screened; based on the edge distribution of fat regions in the normal frames closest to each abnormal frame, candidate regions for the presence of visceral fat in each abnormal frame are determined. Based on the location of the pixels in the neighborhood of the edge line within the candidate region, the edge segments of visceral fat are obtained; the edge segments are filled using the minimum cost path method to obtain the target edge line of visceral fat. Three-dimensional reconstruction of visceral fat in obese patients is performed based on the location of the target edge line of visceral fat.

2. The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 1, wherein, The process of filtering normal and abnormal frames based on the difference in fat area between two adjacent frames includes: All acquired images are arranged to obtain an image sequence; in the image sequence, an image segmentation error factor is obtained based on the difference in fat area between each frame and its neighboring frames. If the image segmentation error factor is less than the preset segmentation error threshold, then the fat area in the image sequence is curve-fitted, and based on the difference between the fat area in the image sequence and the curve fitting result, normal frame images and abnormal frame images are obtained. 3.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 2, characterized in that, The step of determining candidate regions for the presence of visceral fat in each abnormal frame image based on the edge line distribution of the fat region in the nearest normal frame image includes: For any anomalous frame image: In the image sequence, the nearest normal frame image before and the nearest normal frame image after any abnormal frame image are found respectively, and used as two reference frame images for any abnormal frame image. The outermost edge lines of two reference frame images are obtained by using the watershed algorithm, and the outermost edge lines of the two reference frame images are mapped to any abnormal frame image to obtain the first edge line and the second edge line. The region between the first edge line and the second edge line is considered as a candidate region for the presence of visceral fat in any of the abnormal frame images. 4.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 3, characterized in that, The step of obtaining the edge segment of visceral fat based on the location of pixels in the neighborhood of the edge line within the candidate region includes: A sliding window of a preset size is slid along the edge line of the candidate area. When there is a pixel on the first edge line or a pixel on the second edge line in the area corresponding to the sliding window, the center point of the area corresponding to the sliding window is marked, and the difference weight of the marked pixel is determined according to the relative position of the marked pixel with the first edge line and the second edge line. Based on the difference in the number of pixels on each edge line within the candidate region and the marked pixels, the shortest distance between the marked pixels and the first and second edge lines, and the difference weight of the marked pixels, the difference index corresponding to each edge line is obtained. By comparing the magnitude of the difference index corresponding to different edge lines within the candidate region, edge segments of visceral fat are screened.

5. The three-dimensional reconstruction system for visceral fat in obese patients based on CT images according to claim 4, characterized in that, The step of determining the difference weight of the marked pixel based on its relative position to the first edge line and the second edge line includes: If the marked pixel does not intersect with the first edge line and the second edge line, then the difference weight of the marked pixel is set to a preset first value; If a marked pixel intersects with one of the first and second edge lines, then the difference weight of the marked pixel is set to a preset second value. If a marked pixel intersects both the first edge line and the second edge line, then the difference weight of the marked pixel is set to a preset third value. Among them, the first preset value is greater than the second preset value, and the second preset threshold is greater than the third preset threshold. 6.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 4, wherein, The difference index corresponding to each edge line is obtained based on the difference in the number of pixels on each edge line within the candidate region and the marked pixels, the shortest distance between the marked pixels and the first and second edge lines, and the difference weight of the marked pixels. This includes: For any edge line within the candidate region: Calculate the first product of the difference weight of each marked pixel on any edge line and the corresponding shortest distance; By combining the first product corresponding to each marked pixel on any edge line, the corresponding quantity difference, and the average number of pixels on the first edge line and the second edge line, the difference index corresponding to any edge line is determined. 7.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 4, characterized in that, The comparison of the magnitude of the difference index corresponding to different edge lines within the candidate region to screen edge segments of visceral fat includes: Among all edge lines in the candidate region whose length is greater than a preset length threshold, the edge line with the smallest difference index is taken as the edge segment of visceral fat. 8.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 1, wherein, The method of combining minimum cost path to fill in edge segments and obtain the target edge line of visceral fat includes: A cost function is constructed for the pixels in the candidate region. The two endpoints of the edge segment of visceral fat are taken as the start and end points respectively. The minimum cost path method is used to find the line connecting the start and end points to obtain the target edge line of visceral fat.

9. The three-dimensional reconstruction system for visceral fat in obese patients based on CT images according to claim 8, characterized in that, The step of using the minimum cost path method to find the line connecting the starting point and the ending point to obtain the target edge line of visceral fat includes: If there is only one path that satisfies the minimum cost, then the edge segment of visceral fat is filled in using the path that satisfies the minimum cost to obtain the target edge line of visceral fat. If there is more than one path that satisfies the minimum cost, the path that satisfies the minimum cost is recorded as the initial path. The deviation value of each initial path is evaluated based on the first difference between the total number of pixels on the edge line of the visceral fat target and the average number of pixels on the first edge line and the second edge line, and the second difference between the area of ​​the closed region enclosed by each initial path and the edge line of the visceral fat target and the average area of ​​the closed region enclosed by the first edge line and the second edge line. The edge segments of visceral fat are filled in using the initial path with the smallest deviation value to obtain the target edge line of visceral fat. 10.The CT image-based visceral fat three-dimensional reconstruction system for obese patients according to claim 9, wherein, The evaluation of the deviation value for each initial path includes: The weighted sum of the normalized result of the first difference and the normalized result of the second difference corresponding to each initial path is used as the deviation value of each initial path.

Images