An intelligent image segmentation system for lower limb trauma fracture detection

Abstract

The present invention relates to the technical field of local image enhancement, and specifically to an intelligent image segmentation system for detecting lower limb traumatic fractures. The present invention obtains the degree of change of each pixel point through the change regularity of the pixels in the rows and columns in the lower limb CT image; obtains the abnormality of each pixel point based on the fluctuation influence of the degree of change of each pixel point in the row where each pixel point is located; obtains the adjacent areas corresponding to the preset neighborhood area of ​​each pixel point, obtains the possibility of trauma of each pixel point based on the difference in abnormality between the preset neighborhood area corresponding to each pixel point and each adjacent area, and the symmetry of the adjacent areas, obtains the lower limb enhanced CT image, and segments the lower limb enhanced CT image. The present invention obtains more accurate and clear image segmentation results by enhancing the contrast of pixel points.

Description

Technical Field

[0001] The present invention relates to the technical field of local image enhancement, and in particular to an intelligent image segmentation system for lower limb trauma fracture detection. Background Art

[0002] Medical CT imaging is a commonly used imaging technology used to obtain high-resolution three-dimensional images of the human body. The main purpose of medical CT image segmentation is to extract tissue structures or specific regions in CT images and segment them into different regions or objects. Therefore, image segmentation is a relatively important image processing technology.

[0003] In the existing segmentation of lower limb CT images, the situation that the pixel values ​​of the pixels located in the traumatic fracture area and the normal area of ​​the image may not be significantly different is not taken into account, resulting in poor accuracy of the segmentation results obtained by directly segmenting the lower limb CT image after filtering. Summary of the Invention

[0004] In order to solve the technical problem of inaccurate lower limb CT image segmentation results in the prior art, the present invention aims to provide an intelligent image segmentation system for lower limb trauma fracture detection. The technical solutions adopted are as follows:

[0005] The present invention provides an intelligent image segmentation system for lower limb trauma fracture detection, the system comprising:

[0006] An image acquisition module, used for acquiring CT images of lower limbs;

[0007] The pixel abnormality acquisition module is used to obtain the degree of change of each pixel point based on the difference in the change rules of the pixel value of each pixel point in the row and column of the lower limb CT image; and obtain the abnormality of each pixel point based on the fluctuation influence of the change degree of each pixel point in the row where each pixel point is located;

[0008] A local anomaly analysis module is used to obtain the adjacent areas of each pixel point within the local range of the preset neighborhood area corresponding to each pixel point; based on the abnormality difference between the preset neighborhood area corresponding to each pixel point and each adjacent area, as well as the symmetry of the adjacent areas, the possibility of trauma at each pixel point is obtained;

[0009] The image segmentation module is used to obtain the lower limb enhanced CT image according to the trauma possibility of each pixel point and segment the lower limb enhanced CT image.

[0010] Furthermore, the method for obtaining the degree of change includes:

[0011] In each row of pixels in the lower limb CT image, the position index value of each pixel is obtained in order from left to right; the pixel sequence corresponding to each row in the lower limb CT image is obtained by combining the position index values ​​of each pixel; the pixels in the pixel sequence are arranged in ascending order according to pixel value, and pixels with the same pixel value are arranged in ascending order according to position index value;

[0012] Obtain the pixel index value of each pixel point according to the arrangement order of the pixels in the pixel sequence;

[0013] Each pixel in the lower limb CT image is sequentially used as a reference pixel, and the row where the reference pixel is located is used as the target row. In each row adjacent to the target row, the pixel index value of the pixel with the same position index value as the reference pixel is obtained as the adjacent index value of the reference pixel. The difference between the pixel index value of the reference pixel and the adjacent index value is calculated to obtain the adjacent difference of the reference pixel.

[0014] When the number of adjacent differences of the reference pixel points is 1, the preset change degree is used as the change degree of the reference pixel points; when the number of adjacent differences of the reference pixel points is greater than 1, the difference between the adjacent differences of the reference pixel points is used as the change degree of the reference pixel points.

[0015] Furthermore, the method for obtaining the abnormality includes:

[0016] Each row in the lower limb CT image is used as a reference row. For any pixel in the reference row, the variance of the change of all other pixels in the reference row except the pixel is calculated to obtain the change difference of the pixel.

[0017] Obtain the mean of the change differences corresponding to all pixels in the reference row, and use the difference between the change difference of each pixel in the reference row and the mean of the change differences as the change difference coefficient of each pixel;

[0018] The product of the change difference coefficient and the change degree of each pixel is calculated to obtain the abnormality of each pixel.

[0019] Furthermore, the method for obtaining the adjacent area includes:

[0020] Each pixel point in the lower limb CT image is taken as the central pixel point in turn, and the preset neighborhood area corresponding to the central point is taken as the central area;

[0021] The nearest adjacent region corresponding to the central region is obtained under different preset directions of the central region. The shape and size of the adjacent region are consistent with the central region, and the pixels in the adjacent region do not overlap with the pixels in the central region.

[0022] Furthermore, the method for obtaining the trauma possibility includes:

[0023] For any central region, in the central region and all corresponding adjacent regions, the abnormality of the pixels is sorted according to the preset arrangement order based on the pixel positions to obtain the abnormality sequence corresponding to the central region and each adjacent region;

[0024] Two adjacent areas in each preset symmetry direction of the central area are regarded as a symmetric area pair;

[0025] Taking each adjacent area of ​​the central area as a detection area in turn, and obtaining the abnormal difference between the central area and the detection area according to the difference between the abnormality sequences corresponding to the central area and the detection area;

[0026] In each symmetrical region pair corresponding to the central region, the difference between the abnormal differences of the two adjacent regions is taken as the regional difference of the central region; the cumulative value of the differences of all regions corresponding to the central region is calculated to obtain the trauma possibility of the central pixel point corresponding to the central region.

[0027] Furthermore, the specific expression of the abnormal difference is:

[0028] ;

[0029] Where, Represented as the central area Corresponding to The abnormal differences between adjacent areas, Represented as the central area The first in the abnormal sequence An abnormality, Expressed as The abnormal sequence of adjacent regions An abnormality, Expressed as the total number of anomalies in the anomaly sequence, Expressed as an absolute value extraction function.

[0030] Furthermore, obtaining a lower limb enhanced CT image according to the trauma possibility of each pixel point includes:

[0031] The trauma probability after negative correlation mapping and normalization is used as the pixel adjustment coefficient of each pixel;

[0032] The product of the maximum grayscale of each pixel and the pixel adjustment coefficient of each pixel is calculated to obtain a new pixel value of each pixel and obtain a lower limb enhanced CT image.

[0033] Furthermore, the method for obtaining the nearest neighboring area includes:

[0034] In each preset direction, the distance between the adjacent center points of the adjacent areas and the central pixel point of the central area is calculated, and the adjacent area corresponding to the adjacent center point with the smallest distance is taken as the nearest adjacent area.

[0035] Furthermore, the preset neighborhood area is eight neighborhood areas.

[0036] Furthermore, the Otsu threshold segmentation algorithm was used to segment the lower limb enhanced CT images.

[0037] The present invention has the following beneficial effects:

[0038] The present invention analyzes the variation pattern of pixels in rows and columns in lower limb CT images to obtain the variation degree of each pixel. In the bone area, the variation degree of each row of pixels is similar, but fractures will destroy the similarity of the variation degree. Therefore, based on the fluctuation of the variation degree of each pixel in the row where each pixel is located, the abnormality of each pixel is obtained. The abnormality can reflect the possibility that the pixel is in the traumatic fracture area. However, when the lower limb bones are fractured, there will be overlap. At this time, the abnormality of the pixel is closer to that of normal pixels. Therefore, it is necessary to further analyze the regional abnormality of each pixel in the local area range, that is, to obtain the adjacent areas corresponding to the preset neighborhood area of ​​each pixel. Based on the difference in abnormality between the preset neighborhood area and each adjacent area, as well as the symmetry of the adjacent areas, the possibility of trauma of each pixel is obtained. A more comprehensive analysis of the possibility that the pixel is in the traumatic fracture area makes it more accurate to further enhance the pixel value. The contrast between the pixels in the traumatic area and the normal area of ​​the lower limb enhanced CT image is increased, so that the segmentation result obtained by the lower limb enhanced CT image is more reliable and more accurate and clear image segmentation results are obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. 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 any creative work.

[0040] Figure 1 This is a structural diagram of an intelligent image segmentation system for lower limb trauma fracture detection provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0041] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of an intelligent image segmentation system for detecting lower limb traumatic fractures according to the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0042] Unless defined otherwise, 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 belongs.

[0043] The specific scheme of the intelligent image segmentation system for lower limb trauma fracture detection provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a structural diagram of an intelligent image segmentation system for detecting lower limb traumatic fractures provided by an embodiment of the present invention. The system includes: an image acquisition module 101, a pixel abnormality acquisition module 201, a local abnormality analysis module 301, and an image segmentation module 401.

[0045] The image acquisition module 101 is used to acquire CT images of the lower limbs.

[0046] In an embodiment of the present invention, a CT scanner can be used to acquire an initial lower limb CT image. It should be noted that the initial lower limb CT image is a grayscale image immediately after generation and does not require further grayscale processing. Furthermore, the lower limb bones are typically leg bones, which are arranged vertically in a CT image. Gaussian filtering is then used to denoise the initial lower limb CT image to obtain a lower limb CT image. CT image acquisition and Gaussian filtering denoising are well known techniques to those skilled in the art and are not described in detail here.

[0047] The pixel abnormality acquisition module 201 is used to obtain the degree of change of each pixel point based on the difference in the change rules of the pixel value of each pixel point in the row and column in the lower limb CT image; and obtain the abnormality of each pixel point based on the fluctuation influence of the change degree of each pixel point in the row where each pixel point is located.

[0048] Since the leg bones are arranged vertically in the lower limb CT images, and the pixels in the bone area have continuous and similar pixel value change characteristics, for the pixels in the traumatic fracture area, the fracture will destroy the similarity of the change of the corresponding pixel values ​​of the pixels. Therefore, the change degree of each pixel point is first obtained based on the difference in the change rules of the pixel values ​​of each pixel point in the lower limb CT images in rows and columns.

[0049] Preferably, in each row of pixels in the lower limb CT image, a position index value of each pixel is obtained in order from left to right, and the pixel sequence corresponding to each row in the lower limb CT image is obtained in combination with the position index value of each pixel, wherein the pixel values ​​of the pixels in the pixel sequence are arranged from small to large, and pixels with the same pixel values ​​are arranged from small to large according to the position index value. The pixel sequence is obtained by combining the pixel value sequence of the pixels with the arrangement position information of the pixels in each row. The pixel sequence of each row mainly reflects the distribution and arrangement pattern of the pixel values ​​of the pixels in each row. After combining the position index value of the pixels, the distribution change of the pixel values ​​when the pixel columns are in the same position can be analyzed.

[0050] The pixel index value of each pixel is obtained based on the order of the pixels in the pixel sequence. The pixel index value reflects the pixel value distribution information of each pixel. Each pixel in the lower limb CT image is used as the reference pixel, and the row where the reference pixel is located is used as the target row. In each row adjacent to the target row, the pixel index value of the pixel with the same position index value as the reference pixel is obtained as the neighboring index value of the reference pixel. The neighboring index value is the pixel value distribution information of the neighboring pixels in the same column as the reference pixel.

[0051] Preferably, the difference between the pixel index value of the reference pixel and the adjacent index value is calculated to obtain the adjacent difference of the reference pixel. The adjacent difference is the difference in pixel value distribution information between the reference pixel and an adjacent pixel with the same column number.

[0052] When the number of adjacent differences of the reference pixel point is 1, it means that the reference pixel point is located at the first and last rows in the lower limb CT image, and there is only one row adjacent to the target row. Since the trauma position to be analyzed in actual CT shooting will not be located at the first and last rows of the image, the preset change degree is used as the change degree of the reference pixel point. In the embodiment of the present invention, the preset change degree is 0, and the specific value can be adjusted by the implementer.

[0053] When the number of adjacent differences of the reference pixel is greater than 1, that is, the reference pixel is located in a row other than the first row and the last row, and there are two rows adjacent to the target row, the difference between the adjacent differences of the reference pixel is used as the variation degree of the reference pixel. In this embodiment of the present invention, the specific expression of the variation degree is:

[0054] ;

[0055] Where, Expressed as The degree of change of pixels, Expressed as The pixel is located at The pixel index value of the row, Expressed as The pixel corresponds to The image adjacent index value of the row, Expressed as The pixel corresponds to The image adjacent index value of the row, Expressed as an absolute value extraction function.

[0056] in, and All expressed as The adjacent differences of the pixels are calculated. In the embodiment of the present invention, the differences are calculated as the absolute values ​​of the differences. The degree of change reflects the degree of change in the pixel value distribution arrangement of each pixel and the pixels in the same column in the adjacent rows.

[0057] Furthermore, considering that the bones of the leg are distributed vertically, the degree of change of the pixels in each row is similar. However, when the pixels in the row are the pixels corresponding to the traumatic fracture area, the change of the pixels in the traumatic fracture area is inconsistent with the regularity of the normal pixels. The difference between the degree of change of the pixels in the traumatic fracture area and the overall regularity of the row is large.

[0058] In an embodiment of the present invention, each row in the lower limb CT image is taken as a reference row in turn. For any pixel point in the reference row, the variance of the change degree of all other pixel points in the reference row except the pixel point is calculated to obtain the change difference of the pixel point. The change difference reflects the contribution of the pixel point to the variance of the change degree of the entire row.

[0059] Get the mean of the corresponding change differences of all pixels in the reference row, and use the mean to reflect the average contribution of each pixel to the variance. The difference between the change difference of each pixel in the reference row and the mean of the change difference is used as the change difference coefficient of each pixel. The larger the change difference coefficient, the greater the influence of the pixel on the variance, and the more obvious the abnormality of the pixel.

[0060] The product of the variation difference coefficient and the variation degree of each pixel is calculated to obtain the abnormality of each pixel. In the embodiment of the present invention, the specific expression of abnormality is:

[0061] ;

[0062] Where, Expressed as The abnormality of each pixel, Expressed as The degree of change of pixels, Expressed as The mean value of the change difference of the row where the pixels are located, Expressed as The difference in pixel changes, Expressed as an absolute value extraction function.

[0063] in, Expressed as The coefficient of variation of each pixel point is used to determine the abnormality. The abnormality reflects the possibility that the pixel point is a pixel point in the traumatic fracture area. When the pixel point belongs to the normal area, the degree of variation of the pixel point is not much different from the overall regularity of the row, and the overall regularity of the pixels in each row is relatively consistent. Therefore, the abnormality of the pixel point will be smaller. When the pixel point belongs to the traumatic fracture area, the degree of variation of the pixel point is significantly different from the overall regularity of the row, and the abnormality of the pixel point will also be larger.

[0064] The local abnormality analysis module 301 is used to obtain the adjacent areas of the preset neighborhood area corresponding to each pixel point within the local range of the preset neighborhood area corresponding to each pixel point; based on the abnormality differences between the preset neighborhood area corresponding to each pixel point and each adjacent area, as well as the symmetry of the adjacent areas, the possibility of trauma of each pixel point is obtained.

[0065] By analyzing the regularity of pixel rows and columns in lower limb CT images, we can determine the likelihood that a pixel is located in a traumatic fracture area. However, considering that some traumatic fracture areas may not have obvious pixel abnormalities, for example, when a leg fracture occurs, there may be areas of overlapping bone fragments. When there are overlapping bone fragments, the regularity of the local row and column variation is disrupted and the abnormality tends to be normal. Therefore, regularity alone cannot fully determine that a pixel is located in a traumatic bone area. It is necessary to consider the overall degree of difference of each pixel within the local range.

[0066] Therefore, a preset neighborhood area is set for each pixel point for analysis. In an embodiment of the present invention, the preset neighborhood area is an eight-neighborhood area centered on each pixel point. Each pixel point in the lower limb CT image is taken as the center pixel point in turn, and the preset neighborhood area corresponding to the center point is taken as the center area. The center area contains 9 pixels. The nearest adjacent area corresponding to the center area is obtained under different preset directions of the center area. The shape and size of the adjacent area are consistent with the center area, and the pixels in the adjacent area do not overlap with the pixels in the center area. In an embodiment of the present invention, the preset direction is the 8-chain code direction, that is, the center area obtains an adjacent area with the same shape and size under each direction of the 8-chain code, and each adjacent area contains 9 pixels. In order to ensure that the adjacent area is the nearest adjacent area, under each preset direction, the distance between the adjacent center point of the adjacent area and the center pixel point of the central area is calculated, and the adjacent area corresponding to the adjacent center point with the smallest distance is taken as the nearest adjacent area. It should be noted that the distance can be calculated using Euclidean distance. Euclidean distance is a technical means well known to those skilled in the art and will not be elaborated here.

[0067] By analyzing the overall differences between regions within the local range of the central region, a more accurate judgment can be made on the pixels. Preferably, for any central region, the abnormalities of the pixels in the central region and all corresponding adjacent regions are sorted according to the pixel positions in a preset arrangement order to obtain an abnormality sequence corresponding to the central region and each adjacent region. In an embodiment of the present invention, the preset arrangement order is set to take the pixel point at the upper left corner of each region as the starting point, sort in order from left to right per row, and the pixel point at the lower right corner of the region as the end point. The abnormalities of the pixels are sorted in sequence according to this arrangement order to obtain an abnormality sequence. The implementer of the preset arrangement order can adjust it at will, and only needs to ensure that the abnormalities in each region are arranged in the same pixel position traversal order to ensure that when calculating the difference between the abnormalities, the abnormality is the abnormality corresponding to the pixel position in the two regions.

[0068] Two adjacent regions in each preset symmetric direction of the central region are used as a symmetric region pair. In the embodiment of the present invention, the preset symmetric directions are the horizontal direction, the vertical direction, the 45-degree diagonal direction, and the 135-degree diagonal direction. When the central region can obtain an adjacent region in each direction of the 8-chain code, there are four pairs of symmetric region pairs. In each symmetric direction, two adjacent regions can be obtained to form a symmetric region pair.

[0069] Each adjacent area in the central area is sequentially used as a detection area. Based on the difference between the abnormality sequences corresponding to the central area and the detection area, the abnormality difference between the central area and the detection area is obtained. In this embodiment of the present invention, the expression of the abnormality difference is:

[0070] ;

[0071] Where, Represented as the central area Corresponding to The abnormal differences between adjacent areas, Represented as the central area The first in the abnormal sequence An abnormality, Expressed as The abnormal sequence of adjacent regions An abnormality, Expressed as the total number of anomalies in the anomaly sequence, Expressed as an absolute value extraction function.

[0072] The abnormal difference reflects the deviation change of the abnormality between the central area and each adjacent area. Furthermore, in each symmetrical area pair corresponding to the central area, the difference between the abnormal differences of the two adjacent areas is used as the regional difference of the central area. The regional difference reflects the deviation degree of the difference in each symmetrical direction of the central area. The greater the regional difference, the more likely the local range is to have a traumatic fracture area. The cumulative value of the differences corresponding to all areas of the central area is calculated to obtain the trauma possibility of the central pixel point corresponding to the central area. In the embodiment of the present invention, the specific expression of the trauma possibility is:

[0073] ;

[0074] Where, Expressed as The potential for trauma per pixel, Expressed as The central region of the symmetrical region pair The abnormal difference corresponding to the first adjacent area, Expressed as The central region of the symmetrical region pair The abnormal difference corresponding to the second adjacent area, Expressed as the total number of symmetric region pairs, Expressed as an absolute value extraction function.

[0075] in, Expressed as The corresponding central area of ​​the symmetric region regional differences, Expressed as The larger the regional difference, the greater the fluctuation of abnormality in the local range corresponding to the central region. The greater the overall deviation of abnormality, the more likely the corresponding central pixel is a pixel in the traumatic fracture area.

[0076] The image segmentation module 401 is used to obtain the lower limb enhanced CT image according to the trauma possibility of each pixel point and segment the lower limb enhanced CT image.

[0077] The trauma possibility of the pixel points can be used to further enhance the lower limb CT image, increase the pixel value difference between the pixels in the traumatic fracture area and the normal pixels, so that the traumatic fracture area can be segmented more accurately during subsequent segmentation. Since the pixels in the traumatic fracture area usually have lower pixel values, in an embodiment of the present invention, the trauma possibility of negative correlation mapping and normalization is used as the pixel adjustment coefficient of each pixel point. The greater the trauma possibility of the pixel point, the more likely the pixel point is a pixel point in the traumatic fracture area, and the smaller the adjustment coefficient of the pixel point. Furthermore, the product of the maximum grayscale level of the pixel point and the pixel adjustment coefficient of each pixel point is calculated to obtain a new pixel value for each pixel point. The smaller the adjustment coefficient, the smaller the new pixel value of the pixel point, and the greater the difference between the pixel point in the traumatic fracture area and the normal pixel point, thereby obtaining an enhanced CT image of the lower limb.

[0078] In other embodiments of the present invention, a trauma threshold can be set to filter out pixels that are greater than the trauma threshold and may be in the trauma fracture area, and the pixel value contrast of the pixel points can be increased by reducing the pixel values ​​of the pixel points in the trauma fracture area. This will not be elaborated here.

[0079] The enhanced CT images of the lower extremities are segmented using the Otsu threshold segmentation algorithm. In the embodiment of the present invention, the Otsu threshold segmentation algorithm is used to obtain a segmentation threshold. Pixel values ​​less than the segmentation threshold are set to 0, and pixel values ​​greater than the segmentation threshold are set to 255, resulting in a binary image. Regions with a pixel value of 0 in the binary image represent traumatic fracture areas. It should be noted that the Otsu threshold segmentation algorithm is a well-known technique for those skilled in the art and will not be described in detail here.

[0080] Thus, more accurate image segmentation is performed by increasing the contrast between the pixels in the traumatic fracture area and the normal pixels.

[0081] In summary, the present invention obtains the degree of change of each pixel by analyzing the change pattern of pixels in rows and columns in the lower limb CT image. The degree of change of each pixel in each row of pixels in the bone area is similar, but fractures will destroy the similarity of the degree of change. Therefore, the abnormality of each pixel is obtained based on the fluctuation influence of the degree of change of each pixel in the row where each pixel is located. The abnormality can reflect the possibility that the pixel is in the traumatic fracture area. However, when bone fractures occur in the lower limb bones, there will be overlap. At this time, the abnormality of the pixel is closer to that of normal pixels. Therefore, it is necessary to further analyze the regional abnormality of the local area range of each pixel, that is, to obtain the adjacent areas corresponding to the preset neighborhood area of ​​each pixel, and according to the abnormality difference between the preset neighborhood area corresponding to each pixel and each adjacent area, as well as the symmetry of the adjacent areas, to obtain the possibility of trauma of each pixel. A more comprehensive analysis of the possibility that the pixel is in the traumatic fracture area makes it more accurate to further enhance the pixel value, obtain the contrast between the pixels in the traumatic area and the normal area of ​​the enhanced CT image of the lower limb, and segment the enhanced CT image of the lower limb to obtain a more accurate and clear image segmentation result.

[0082] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0083] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. An intelligent image segmentation system for lower limb trauma fracture detection, characterized in that: The system comprises: An image acquisition module, used for acquiring CT images of lower limbs; The pixel abnormality acquisition module is used to obtain the degree of change of each pixel in the lower limb CT image based on the difference in the change pattern of the pixel value of each pixel in the row and column; and obtain the abnormality of each pixel based on the fluctuation influence of the change degree of each pixel in the row where each pixel is located. The abnormality indicates the possibility that the pixel is in the traumatic fracture area. The abnormality of the pixel in the normal area is smaller than that of the pixel in the traumatic fracture area. A local anomaly analysis module is used to obtain the adjacent areas of each pixel point within the local range of the preset neighborhood area corresponding to each pixel point; based on the abnormality difference between the preset neighborhood area corresponding to each pixel point and each adjacent area, as well as the symmetry of the adjacent areas, the possibility of trauma at each pixel point is obtained; The method for obtaining the adjacent regions includes: sequentially taking each pixel point in the lower limb CT image as the central pixel point, and taking a preset neighborhood area corresponding to the central point as the central region; obtaining the nearest adjacent region corresponding to the central region at different preset directions of the central region, wherein the shape and size of the adjacent region are consistent with the central region, and the pixels in the adjacent region do not overlap with the pixels in the central region; The method for obtaining the possibility of trauma includes: for any central region, in the central region and all corresponding adjacent regions, sorting the abnormalities of the pixels in a preset arrangement order according to the pixel positions to obtain an abnormality sequence corresponding to the central region and each adjacent region; using two adjacent regions in each preset symmetric direction of the central region as a symmetric region pair; using each adjacent region in the central region as a detection region in turn, and obtaining an abnormality difference of the detection region corresponding to the central region based on the difference between the abnormality sequences corresponding to the central region and the detection region; in each symmetric region pair corresponding to the central region, using the difference between the abnormality differences of the two adjacent regions as the regional difference of the central region; and calculating the cumulative value of the differences of all regions corresponding to the central region to obtain the possibility of trauma corresponding to the central pixel of the central region; The image segmentation module is used to obtain the lower limb enhanced CT image according to the trauma possibility of each pixel point and segment the lower limb enhanced CT image.

2. The intelligent image segmentation system for detecting lower limb traumatic fractures according to claim 1, characterized in that: The method for obtaining the degree of change includes: In each row of pixels in the lower limb CT image, the position index value of each pixel is obtained in order from left to right; the pixel sequence corresponding to each row in the lower limb CT image is obtained by combining the position index values ​​of each pixel; the pixels in the pixel sequence are arranged in ascending order according to pixel value, and pixels with the same pixel value are arranged in ascending order according to position index value; Obtain the pixel index value of each pixel point according to the arrangement order of the pixels in the pixel sequence; Each pixel in the lower limb CT image is sequentially used as a reference pixel, and the row where the reference pixel is located is used as the target row. In each row adjacent to the target row, the pixel index value of the pixel with the same position index value as the reference pixel is obtained as the adjacent index value of the reference pixel. The difference between the pixel index value of the reference pixel and the adjacent index value is calculated to obtain the adjacent difference of the reference pixel. When the number of adjacent differences of the reference pixel points is 1, the preset change degree is used as the change degree of the reference pixel points; when the number of adjacent differences of the reference pixel points is greater than 1, the difference between the adjacent differences of the reference pixel points is used as the change degree of the reference pixel points.

3. The intelligent image segmentation system for detecting lower limb traumatic fractures according to claim 1, characterized in that: The method for obtaining the abnormality includes: Each row in the lower limb CT image is used as a reference row. For any pixel in the reference row, the variance of the change of all other pixels in the reference row except the pixel is calculated to obtain the change difference of the pixel. Obtain the mean of the change differences corresponding to all pixels in the reference row, and use the difference between the change difference of each pixel in the reference row and the mean of the change differences as the change difference coefficient of each pixel; The product of the variation coefficient and the degree of variation of each pixel is calculated to obtain the abnormality of each pixel.

4. The intelligent image segmentation system for detecting lower limb traumatic fractures according to claim 1, characterized in that: The specific expression of the abnormal difference is: ； Where, Represented as the central area Corresponding to The abnormal differences between adjacent areas, Represented as the central area The first in the abnormal sequence An abnormality, Expressed as The abnormal sequence of adjacent regions An abnormality, Expressed as the total number of anomalies in the anomaly sequence, Expressed as an absolute value extraction function.

5. The intelligent image segmentation system for detecting lower limb traumatic fractures according to claim 1, characterized in that: Obtaining a lower limb enhanced CT image according to the trauma possibility of each pixel point includes: The trauma probability of negative correlation mapping and normalization is used as the pixel adjustment coefficient of each pixel; The product of the maximum grayscale of each pixel and the pixel adjustment coefficient of each pixel is calculated to obtain a new pixel value of each pixel and obtain a lower limb enhanced CT image.

6. The intelligent image segmentation system for lower limb trauma fracture detection according to claim 1, characterized in that: The method for obtaining the nearest adjacent area includes: In each preset direction, the distance between the adjacent center points of the adjacent areas and the central pixel point of the central area is calculated, and the adjacent area corresponding to the adjacent center point with the smallest distance is taken as the nearest adjacent area.

7. The intelligent image segmentation system for detecting lower limb traumatic fractures according to claim 1, characterized in that: The preset neighborhood area is eight neighborhood areas.

8. The intelligent image segmentation system for lower limb trauma fracture detection according to claim 1, characterized in that: The Otsu threshold segmentation algorithm was used to segment the enhanced CT images of the lower limbs.

Images