Method for segmenting brain tumor lesion region based on CT image analysis

Abstract

The application relates to the technical field of image processing, in particular to a brain tumor lesion region segmentation method based on CT image analysis. The method comprises the following steps: performing segmentation on a meningeal CT image to obtain an initial segmentation region, and performing pretreatment on the initial segmentation region to obtain an enhanced image; acquiring the gray gradient direction vectors of each pixel point in the enhanced image; acquiring the skeleton lines in each foreground region and the tangent lines of each end point of the skeleton lines, and then acquiring the suspected truncated end points of each end point; screening the pixel points in the region of each suspected truncated end point according to the gray gradient direction vectors of the pixel points in the foreground region of each suspected truncated end point of one end point to acquire the chaotic pixel points of the end point, and then acquiring each tumor partial region of the end point; and taking the center points of each tumor partial region of each end point as seed points of region growth to perform region growth in each tumor partial region to obtain a tumor region. The application can effectively segment the brain tumor region.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a brain tumor lesion region segmentation method based on CT image analysis. BACKGROUND

[0002] Precise segmentation of brain tumors is one of the core challenges in the field of medical image analysis, and its segmentation accuracy is directly related to the accuracy of clinical diagnosis, the reliability of surgical planning, and the effectiveness of treatment evaluation. Currently, brain tumor segmentation technology based on multi-slice CT images mainly relies on traditional methods such as image gray scale features, texture analysis, and region growing. For tumors located inside the brain and with obvious gray scale contrast with the surrounding normal brain tissue, these methods can achieve relatively accurate positioning and segmentation by analyzing local gray scale differences. In particular, methods based on threshold segmentation and traditional region growing can utilize the gray scale difference between lesions and normal tissue to achieve accurate positioning and segmentation of tumors when dealing with multi-slice CT images with uniform texture inside the brain, and have achieved certain results in clinical applications.

[0003] Although existing technologies perform well in handling brain tumor segmentation, when tumors are located in the cerebral cortex region (especially meningiomas), traditional segmentation methods face severe challenges. The cerebral cortex surface is not a smooth plane, but is covered with complex and winding sulci structures, which have natural and continuous trends and connectivity characteristics in CT images. Using traditional segmentation methods in the cortical region, if there is overlap between the tumor and the sulci region, the natural existence of sulci will cause unreasonable contraction or expansion of the segmentation boundary at the sulci, making it difficult to accurately determine the actual boundary of the tumor. Although existing preprocessing techniques can obtain the approximate region of the tumor, they cannot achieve fine segmentation of the lesion boundary, resulting in significant deviations between the segmentation results and the actual tumor region, which seriously affects subsequent clinical decision-making. SUMMARY

[0004] To solve the above technical problems, the purpose of the present application is to provide a brain tumor lesion region segmentation method based on CT image analysis, and the technical solution adopted is as follows:

[0005] One embodiment of the present application provides a brain tumor lesion region segmentation method based on CT image analysis, which comprises:

[0006] segmenting the meningeal CT image to obtain an initial segmentation region, and separately mapping the initial segmentation region to an image for preprocessing to obtain an enhanced image; and based on the gray scale change of each pixel point in the enhanced image, obtaining the gray scale gradient direction vector of each pixel point;

[0007] The sulci region in the enhanced image is taken as a foreground region, and other regions are taken as background regions to binarize to obtain a binary image; the foreground region in the binary image is processed by using a Zhang-Suen algorithm to obtain a skeleton line in each foreground region;

[0008] An end point of a skeleton line and coordinates of the end point are used to perform curve fitting to obtain a tangent line of the end point; distances between the end point of a skeleton line and end points of other skeleton lines are used to obtain candidate end points of the end point; angles between the tangent line of an end point and tangent lines of candidate end points of the end point are used to obtain suspected truncated end points of the end point;

[0009] The suspected truncated end points of an end point are used to screen pixel points in regions of the suspected truncated end points to obtain chaotic pixel points of the end point; all the chaotic pixel points of an end point are used to obtain tumor part regions of the end point; center points of the tumor part regions of the end points are taken as seed points for region growing to perform region growing in the tumor part regions to obtain tumor regions.

[0010] Preferably, the gray scale gradient direction vectors of the pixel points are obtained based on gray scale changes of the pixel points in the enhanced image, including:

[0011] Gradient components of a pixel point in an enhanced image in x and y directions are obtained, and the gradient components in the x and y directions form a direction vector of the pixel point.

[0012] Preferably, the tangent line of an end point of a skeleton line is obtained by performing curve fitting based on coordinates of the end point and pixel points adjacent to the end point of the skeleton line, including:

[0013] For an end point of a skeleton line, a preset number of pixel points on the skeleton line from the end point are taken as adjacent pixel points of the end point; coordinates of the end point and the adjacent pixel points are used to perform curve fitting to obtain a fitting curve, and a tangent line of the fitting curve passing through the end point is taken as the tangent line of the end point.

[0014] Preferably, the candidate end points of an end point of a skeleton line are obtained based on distances between the end point and end points of other skeleton lines, including:

[0015] Distances between an end point of a skeleton line and end points of other skeleton lines are calculated, and end points with distances less than a distance threshold are taken as candidate end points of the end point.

[0016] Preferably, the suspected truncated end points of an end point are obtained based on angles between a tangent line of the end point and tangent lines of candidate end points of the end point, including:

[0017] An included angle between a tangent line of an endpoint and a tangent line of a candidate endpoint of the endpoint is obtained as a tangent included angle; a direction consistency index of the candidate endpoint of the endpoint is obtained by using an exponential function with a natural constant as a base to negatively correlate a ratio of a square of an angle of the tangent included angle and an angle attenuation parameter; if the direction consistency index of the candidate endpoint of the endpoint is greater than a direction consistency threshold value, the candidate endpoint is a suspected truncated endpoint of the endpoint.

[0018] Preferably, the pixel points in the region of each suspected truncated endpoint of an endpoint are screened according to the gray scale gradient direction vectors of the pixel points in the foreground region of the suspected truncated endpoint to obtain the chaotic pixel points of the endpoint, comprising:

[0019] For an endpoint, the pixel points in the foreground region of each suspected truncated endpoint of the endpoint are recorded as to-be-analyzed pixel points; a to-be-analyzed pixel point and all the pixel points in the eight-neighborhood of the to-be-analyzed pixel point form a pixel point set corresponding to the to-be-analyzed pixel point; the gray scale gradient direction vectors of all the pixel points in the pixel point set corresponding to the to-be-analyzed pixel point are arithmetically averaged to obtain an average gray scale gradient direction vector; the average value of the cosine values of the included angles between the gray scale gradient direction vectors of each pixel point in the pixel point set corresponding to the to-be-analyzed pixel point and the average gray scale gradient direction vector is recorded as an average included angle cosine value; the chaos degree of the to-be-analyzed pixel point is obtained by subtracting the average included angle cosine value from a first preset value; if the chaos degree of a to-be-analyzed pixel point is greater than a chaos degree threshold value, the to-be-analyzed pixel point is a chaotic pixel point.

[0020] Preferably, all the tumor partial regions of an endpoint are obtained according to all the chaotic pixel points of the endpoint, comprising:

[0021] The morphological processing is performed on all the chaotic pixel points of an obtained endpoint to obtain different mask regions; if a mask region contains a suspected truncated endpoint of the endpoint, the mask region is a tumor partial region.

[0022] The application has at least the following beneficial effects: the application obtains an initial segmentation region by segmenting a meningeal CT image, and separately maps the initial segmentation region to an image to obtain an enhanced image by preprocessing; the gray gradient direction vector of each pixel point in the enhanced image is obtained based on the gray level change of each pixel point, and is used to represent the gradient change of the pixel point; further, the sulcus region in the enhanced image is taken as a foreground region, and other regions are taken as background regions to obtain a binary image by binarization, and the skeleton line in each foreground region is obtained by processing the foreground region in the binary image by using the Zhang-Suen algorithm, so as to represent the sulcus trend of the cerebral cortex surface; further, the suspected truncated end points of each end point are obtained according to the essential difference between the natural continuity of the sulcus trend of the intact cortex region and the connectivity interruption caused by tumor invasion, and then the tumor part region of each end point is obtained by further analyzing the suspected truncated end points of each end point, detecting the region truncated by the tumor in the sulcus trend, and finally the center point of the tumor part region of each end point is taken as a seed point of region growing to perform region growing in the tumor part region to obtain a tumor region, so that a more fine tumor lesion region is segmented, and the boundary of the meningioma is accurately segmented. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0024] Figure 1 A method flowchart of a brain tumor lesion region segmentation method based on CT image analysis provided by the embodiments of the present application. DETAILED DESCRIPTION

[0025] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object, the following will combine the drawings and the preferred embodiments to specifically describe the specific implementation, structure, features and effects of the brain tumor lesion region segmentation method based on CT image analysis according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0026] 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 the present application belongs.

[0027] The application provides a brain tumor lesion region segmentation method based on CT image analysis.

[0028] Embodiments

[0029] The application mainly applies to segmentation of a brain tumor lesion region.

[0030] Please refer to Figure 1 The application provides a brain tumor lesion region segmentation method based on CT image analysis.

[0031] Step S1, the meningeal CT image is segmented to obtain an initial segmentation region, and the initial segmentation region is separately mapped into an image for preprocessing to obtain an enhanced image; and a gray gradient direction vector of each pixel point is obtained based on a gray level change of each pixel point in the enhanced image.

[0032] Firstly, the coarse segmentation region collected needs to be optimized to provide a local region of interest with high quality and high contrast for a subsequent fine segmentation algorithm based on sulcus trend analysis. Through accurate image alignment, structure enhancement and noise suppression on the coarse segmentation region, this step can effectively highlight the detailed features of the sulcus structure in the cortical region, and eliminate the artifacts and registration errors that may be introduced in the previous processing, so as to ensure that the subsequent sulcus connectivity analysis is based on stable and pure local data. This process locks the accurate spatial range for subsequent analysis, and makes the continuous trend features of the sulcus clearly presented in the enhanced local image.

[0033] Firstly, the meningeal CT image is obtained, and then an existing automatic or semi-automatic segmentation method based on CT image (for example, threshold segmentation or lightweight neural network segmentation) is used to coarsely segment the meningeal CT image to obtain an initial segmentation region, and the initial segmentation region includes a tumor, meningeal normal tissue and sulcus.

[0034] Further, the obtained initial segmentation region needs to be separately mapped into an image, and the image is recorded as a to-be-processed image, so as to facilitate subsequent analysis. Then, the to-be-processed image is preprocessed, and each layer of the brain CT image is quickly photographed in the initial segmentation region (to-be-processed image) using multi-spiral CT, the spatial correspondence between the cerebral cortex and its adjacent layers is used to locally finely align the to-be-processed image, and the geometric consistency of the to-be-processed image is ensured, so that the aligned to-be-processed image is obtained.

[0035] Then, in the aligned to-be-processed image, a filter designed for enhancing linear and sulcal structure (such as a blood vessel ridge enhancement filter based on a Hessian matrix) is applied to filter the aligned to-be-processed image to obtain a filtered to-be-processed image, which enhances the continuous trend of the sulcus and suppresses irrelevant tissue texture interference.

[0036] Finally, the filtered to-be-processed image also needs to be denoised to further enhance the quality of the image. Thus, a non-local mean denoising or anisotropic diffusion filter is applied to filter the filtered to-be-processed image that enhances the sulcal trend to obtain a final enhanced image, which can effectively smooth the noise in the uniform region while maximally preserving the integrity of the sulcal edge.

[0037] Thus, an enhanced enhanced image containing the meningeal region, sulcus and tumor can be obtained. It should be noted that the preprocessing of the initial segmentation region mapped to an image is a known technology and will not be described in detail here.

[0038] Further, although the sulcal structure is enhanced in the preprocessed and optimized enhanced image, its trend and other characteristics are still implicit and fuzzy concepts in the image grayscale and cannot be directly processed. Thus, the sulcal trend, an anatomical feature, needs to be converted into a quantitative and structured mathematical model to establish a reliable calculation benchmark for subsequent connectivity analysis. By calculating the direction and angle of the image gradient, the local sulcal trend of each pixel point is accurately quantified, and the "trend" is explicitly defined as the direction and size of the vector, thereby providing indispensable input data and judgment basis for subsequent steps such as region recognition.

[0039] Thus, the gray gradient direction vector of each pixel point is obtained based on the gray level change of each pixel point in the enhanced image.

[0040] Specifically, the gradient components of a pixel point in the enhanced image in the x direction and the y direction are obtained, and the gradient components in the x direction and the y direction form the direction vector of the pixel point.

[0041] The gradient components of a pixel point in the x direction and the y direction are obtained by using a Sobel operator or a more accurate central difference algorithm, which are denoted as G x and G y , respectively. x , G y .

[0042] The angle of the gray gradient direction vector is (the vector direction is perpendicular to the sulcal edge, and points to the direction of the gray level drop): , and the amplitude of the gray gradient direction vector is The prominence of the trend can be characterized. At the position of each pixel point, the direction of the gray scale gradient direction vector is perpendicular to the local sulcus edge, and points to the direction in which the gray scale drops the fastest, wherein the greater the amplitude of the gray scale gradient direction vector, the clearer the sulcus edge at the position and the more explicit the trend. This is a basis for the existing operation, which quantifies the state of the sulcus trend (including normal sulcus trend and abnormal sulcus trend that is truncated), and is a pre-operation for the subsequent steps. Thus, the gray scale gradient direction vector of each pixel point in the enhanced image is obtained.

[0043] In step S2, the sulcus region in the enhanced image is taken as a foreground region, and other regions are taken as background regions for binarization to obtain a binary image; and the Zhang-Suen algorithm is used to process the foreground regions in the binary image respectively to obtain the skeleton lines in each foreground region.

[0044] The above obtains the enhanced image after the sulcus trend is enhanced, wherein the approximate region recognizes the tumor part characteristics, but because the sulcus and gray scale characteristics of the meninges themselves affect the accurate segmentation, it is now necessary to analyze the destruction of the sulcus region characteristics by the tumor to obtain the edge information of the tumor to locate the tumor region.

[0045] In the enhanced image after the sulcus trend is enhanced obtained in the foregoing steps, the traditional idea of relying on gray scale characteristics is abandoned, and the connectivity characteristics of the sulcus trend of the cerebral cortex surface are analyzed in depth. Although the tumor at the sulcus causes difficulty in segmentation, the normal morphological characteristics of the sulcus are destroyed at the destroyed position, and part of the edge of the tumor can be effectively extracted, and the tumor lesion region is effectively located, the coarse segmentation connected domain of the approximate region is greatly reduced, and more accurate lesion region segmentation is realized.

[0046] Thus, the direction vector of each pixel point in the enhanced image is obtained in the foregoing steps, and then skeleton processing is performed on the direction vector to highlight the trend. All connected sulci are obtained and located, and then direction confusion degree analysis is performed to obtain the true tumor truncated edge. The obtained optimized edge can be used as a basis for re-segmentation of the original obtained region, and finally the fine segmentation of the lesion is realized.

[0047] Specifically, first, the enhanced image obtained after preprocessing is skeletonized and extracted to obtain a single-pixel-width sulcus center line as a skeleton line, and then post-analysis is performed.

[0048] First, in the enhanced image, the gray scale difference between the sulcus part and the tumor and normal tissue is large, so a suitable threshold is used for binarization processing, the sulcus region in the enhanced image is taken as a foreground region, and other regions are taken as background regions for binarization to obtain a binary image. In the binary image, the highlighted region represents the sulcus region (foreground region).

[0049] Further, the foreground region in the binary image is processed by using an iterative thinning algorithm, preferably, the Zhang-Suen algorithm is used in the present application to process the foreground region in the binary image, to gradually peel off the boundary pixels, to retain the center line structure, and to obtain a single-pixel-width skeleton as the skeleton line in each foreground region; wherein post-processing is required before obtaining the final skeleton line, including removing short branches (isolated skeleton segments with a length less than a threshold value) and connecting small breaks caused by noise, to obtain the skeleton line in each foreground region (foreground region). After obtaining the skeleton line, the location of the real truncation can be analyzed according to the continuity of the skeleton line. It should be noted that the binarization processing and the skeleton extraction are known technologies and will not be described in detail here.

[0050] In step S3, a curve is fitted according to the coordinates of the pixel points adjacent to one end point of a skeleton line and the end point, and a tangent line of the end point is obtained; candidate end points of the end point are obtained according to the distances between the end point and the end points of other skeleton lines; and a suspected truncated end point of the end point is obtained according to the included angle between the tangent lines of the end point and the candidate end points.

[0051] The skeleton line obtained by the above steps is 1 pixel wide, and by simplifying the sulcus to a single-pixel-width skeleton, the influence of sulcus width variation can be eliminated, and the connectivity can be directly analyzed. For the sulcus truncated by the tumor, the extension lines of the two parts of the skeleton line should naturally connect in an ideal case, but due to the occupation of the tumor, the connection is physically blocked. By analyzing the consistency of the extension line direction of the end point of the skeleton line, regions with highly consistent directions but interrupted in space can be identified, thereby preliminarily locating the suspected truncated region. These regions may contain real tumor truncation, or may include some normal sulcus termination points or noise.

[0052] Further, two end points of each skeleton line are obtained, and for one end point of a skeleton line, a preset number of pixel points on the skeleton line from the end point are taken in sequence, which are recorded as adjacent pixel points of the end point; a fitting curve is obtained by curve fitting using the coordinates of the end point and the adjacent pixel points of the end point, and a tangent line of the fitting curve passing through the end point is obtained as a tangent line of the end point. In this way, the tangent lines of the end points on each skeleton line can be obtained.

[0053] The curve fitting is a known technology and will not be described in detail here. The preset number is 5, and the more the number, the more accurate the fitted curve. According to the balance between calculation efficiency and accuracy, the preset number can be adjusted.

[0054] Further, the distance between an end point on a skeleton line and each end point on other skeleton lines needs to be calculated, and the end points with a distance less than a distance threshold are taken as candidate end points of the end point. The distance threshold is the diameter of the largest tumor in history of the same type of tumor as the patient's tumor and in the early-middle stage, and the distance between two segments of the same sulcus separated by the tumor is usually not more than the diameter of the tumor itself, and the maximum diameter of the tumor in the early-middle stage in history has a common statistical range. In this way, the end points far away from the end point can be excluded to obtain the candidate end points of each end point, and the two end points analyzed with each other are not analyzed again in the subsequent analysis, for example, there are three end points a, b and c, the candidate end points of a are b and c, and when the candidate end points of b and c are obtained, a is not used as the candidate end point of b and c, so as to improve the calculation efficiency.

[0055] Further, the candidate end points need to be analyzed according to the direction change of the skeleton line to determine which end points are the real truncated end points. Thus, the angle between the tangent line of an end point and the tangent line of a candidate end point of the end point is taken as the tangent angle of the end point.

[0056] Specifically, the angle between the tangent line of an end point and the tangent line of a candidate end point of the end point is taken as the tangent angle, and the ratio of the square of the angle of the tangent angle to an angle attenuation parameter is negatively related to the exponential function with a natural constant to obtain the direction consistency index of a candidate end point of the end point. If the direction consistency index of a candidate end point of the end point is greater than a direction consistency threshold, the candidate end point is a suspected truncated end point of the end point.

[0057] The calculation model of the direction consistency index is specifically:

[0058] ,

[0059] Wherein, C represents the direction consistency index between an end point and a candidate end point of the end point; θ represents the angle of the tangent angle between the tangent line of the end point and the tangent line of a candidate end point of the end point, that is, the angle of the tangent angle, which is mapped to to measure the direction consistency between the tangent lines. The denominator is the parameter for controlling the angle attenuation, that is, the angle attenuation parameter to prevent the exponential change from being too fast, take half of the angle of the largest tangent angle, that is, Overall, the smaller the angle is, the more consistent the extension line direction is, indicating that they may originally belong to the same sulcus and the direction remains coherent after being truncated by the tumor, and the larger the value of C is.

[0060] The direction consistency threshold is the 90th percentile of all C values obtained by the above steps in the sulcus region without tumor. Thus, the suspected truncated end point of each end point can be obtained.

[0061] This step first identifies the end points of the skeleton line and analyzes the virtual connection relationship. By calculating the direction consistency of the end points to the nearby tangent lines, the unnatural end points formed due to tumor truncation, i.e., the suspected truncated end points, are preliminarily screened out for subsequent analysis.

[0062] In step S4, the pixel points in the region of each suspected truncated end point of an end point are screened according to the gray scale gradient direction vectors of the pixel points in the foreground region of the suspected truncated end point to obtain the chaotic pixel points of the end point; each tumor partial region of the end point is obtained according to all the chaotic pixel points of the end point; and the center points of the tumor partial regions of each end point are taken as seed points for region growing to obtain the tumor regions in the tumor partial regions.

[0063] The above process obtains the suspected truncated end points of each end point. These suspected truncated end points include the true truncation, but also include the noise of the normal sulcus. Therefore, it is still necessary to exclude the interference of the noise.

[0064] First, the foreground region (sulcus region) in the binary image obtained above needs to be mapped to the enhanced image for further analysis on the enhanced image.

[0065] Next, the pixel points in the region of each suspected truncated end point of an end point are screened according to the gray scale gradient direction vectors of the pixel points in the foreground region of the suspected truncated end point to obtain the chaotic pixel points of the end point.

[0066] Specifically, for an end point, the pixel points in the foreground region of each suspected truncated end point of the end point are recorded as to-be-analyzed pixel points; a to-be-analyzed pixel point and all the pixel points in the eight-neighborhood of the to-be-analyzed pixel point form a pixel point set corresponding to the to-be-analyzed pixel point; the average gray scale gradient direction vector of all the pixel points in the pixel point set corresponding to the to-be-analyzed pixel point is obtained by arithmetic averaging; the average value of the cosine values of the angles between the gray scale gradient direction vectors of each pixel point in the pixel point set corresponding to the to-be-analyzed pixel point and the average gray scale gradient direction vector is recorded as the average angle cosine value; the average angle cosine value is subtracted from the first preset value to obtain the chaos degree of the to-be-analyzed pixel point; if the chaos degree of a to-be-analyzed pixel point is greater than the chaos degree threshold, the to-be-analyzed pixel point is a chaotic pixel point.

[0067] The calculation model of the chaos degree is specifically:

[0068] ,

[0069] wherein, HL represents the confusion degree of one end point corresponding to one pixel point to be analyzed, that is, one pixel point of the foreground region where each suspected truncated end point of the end point is located; N represents the number of pixel points in the pixel point set corresponding to the pixel point to be analyzed, which is composed of the pixel point to be analyzed and all pixel points in the eight-neighborhood of the pixel point to be analyzed, since it is an eight-neighborhood, the value of N is 9; the gray gradient direction vector of the qth pixel point in the pixel point set; represents the average gray gradient direction vector obtained by calculating the arithmetic mean of the gray gradient direction vectors of all pixel points in the pixel point set corresponding to the pixel point to be analyzed, and the denominator is the product of the modulus of the gray gradient direction vector of the qth pixel point and the modulus of the average gray gradient direction vector, is the cosine value of the angle between the gray gradient direction vector of the qth pixel point in the pixel point set and the average gray gradient direction vector, the greater the value, the more consistent the direction, is the average angle cosine value. The greater the value of HL, the more confused the pixel point is, the more likely it is at the truncated end of the tumor, and the more likely it is a pixel point of the tumor region.

[0070] In addition, the confusion degree threshold is obtained by obtaining the HL of each pixel point in the sulcus region (foreground region) without tumor, and then taking the maximum value as the confusion degree threshold.

[0071] Then, morphological processing is performed on all confused pixel points of one end point to obtain different mask regions; if a mask region contains the suspected truncated end point of the end point, the mask region is a tumor partial region.

[0072] It should be noted that the morphological processing of all confused pixel points of one end point includes morphological processing, morphological closing operation, fusion of adjacent regions, filling of cavities, smoothing of boundaries, etc. This technology is known and will not be described in detail here. In addition, the mask region obtained here is a partial region of the tumor region, so subsequent region growing is still needed to obtain the complete tumor region.

[0073] Finally, the above-mentioned real tumor partial region is obtained, which is the region formed by the tumor covering the sulcus. Through these regions, the tumor real region convergence contour can be provided for the final refined segmentation. Using the existing segmentation technology (here, the seed growing segmentation method is used), refined segmentation can be achieved. The general steps are as follows:

[0074] Region merging: align the obtained tumor partial regions and the enhanced image in position, that is, map the tumor partial regions to the enhanced image to obtain an enhanced image with tumor partial regions marked.

[0075] Seed point selection: the above is a real tumor part area, selecting a seed point in it can ensure that the initial growth position is exactly inside the tumor, so the center point of the real tumor part area is selected as the seed point.

[0076] Segmentation: region growing is performed in each tumor part area by using an existing region growing algorithm.

[0077] Boundary processing and result generation: the segmentation boundary after region growing is smoothed to eliminate unreasonable jaggies, morphological opening operation is applied to remove small holes and isolated regions, to ensure that the final boundary is continuous and closed, and finally realize fine segmentation to obtain a tumor area. Region growing and post-processing after region growing are common image processing techniques and will not be described in detail.

[0078] In summary, the application adopts a layered verification strategy, first locates the suspected truncated end point through sulci skeleton morphological analysis, and then verifies and confirms the real tumor part area through gradient features (gray scale gradient direction vector), thereby improving the accuracy of tumor area segmentation.

[0079] It should be noted that: the above sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0080] Each embodiment in the specification is described in a progressive manner, and the same and similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments. The above only describes the preferred embodiments of the application and does not limit the application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A method for segmenting brain tumor lesion regions based on CT image analysis, characterized in that, The method includes: The meningeal CT image is segmented to obtain the initial segmentation region, and the initial segmentation region is mapped to a separate image for preprocessing to obtain the enhanced image; the gray-level gradient direction vector of each pixel is obtained based on the gray-level changes of each pixel in the enhanced image. The brain sulcus region in the enhanced image is used as the foreground region, and other regions are used as the background region. The images are then binarized to obtain a binary image. The Zhang-Suen algorithm is used to process the foreground region in the binary image to obtain the skeleton lines in each foreground region. Get the tangent line of a skeleton line by performing curve fitting based on the pixel point adjacent to an endpoint of the skeleton line and the coordinates of the endpoint; get the candidate endpoints of an endpoint based on the distance between an endpoint of a skeleton line and the endpoints of other skeleton lines; get the suspected truncated endpoints of an endpoint based on the angle between an endpoint and the tangent lines of the candidate endpoints of the endpoint. The pixels in the foreground region of each suspected truncated endpoint are filtered to obtain the disordered pixels of the endpoint. The tumor regions of the endpoint are obtained based on all the disordered pixels of the endpoint. The center point of each tumor region of each endpoint is used as the seed point for region growth. Region growth is performed in each tumor region to obtain the tumor region. The step of filtering pixels in the foreground region where each suspected truncated endpoint is located based on the grayscale gradient direction vector of each pixel in the region where each suspected truncated endpoint is located to obtain the disordered pixels of that endpoint includes: For an endpoint, each pixel in the foreground region where each suspected truncated endpoint is located is recorded as a pixel to be analyzed. A set of pixels corresponding to the pixel to be analyzed and all pixels in its eight-neighborhood are formed. The arithmetic mean of the gray-level gradient direction vectors of all pixels in the set of pixels corresponding to the pixel to be analyzed is calculated to obtain the average gray-level gradient direction vector. The average cosine of the angle between the gray-level gradient direction vector of each pixel in the set of pixels corresponding to the pixel to be analyzed and the average gray-level gradient direction vector is calculated and recorded as the average angle cosine value. The disorder level of the pixel to be analyzed is obtained by subtracting the average angle cosine value from a first preset value. If the disorder level of a pixel to be analyzed is greater than the disorder level threshold, then the pixel to be analyzed is a disordered pixel.

2. The method for segmenting brain tumor lesion regions based on CT image analysis according to claim 1, characterized in that, The step of obtaining the gray-level gradient direction vector of each pixel based on the gray-level changes of each pixel in the enhanced image includes: Obtain the gradient components of a pixel in the enhanced image in the x and y directions, and the gradient components in the x and y directions constitute the direction vector of the pixel.

3. The method for segmenting brain tumor lesion regions based on CT image analysis according to claim 1, characterized in that, The step of obtaining the tangent line at a skeleton line by performing curve fitting based on the pixel point adjacent to one endpoint of the skeleton line and the coordinates of that endpoint includes: For an endpoint of a skeleton line, a preset number of pixels on the skeleton line are sequentially taken from the endpoint and recorded as the neighboring pixels of the endpoint. A curve fitting is performed using the coordinates of the endpoint and each neighboring pixel of the endpoint to obtain a fitted curve. The tangent of the fitted curve is obtained by drawing a tangent line through the endpoint.

4. The method for segmenting brain tumor lesion regions based on CT image analysis according to claim 1, characterized in that, The step of obtaining candidate endpoints for an endpoint based on the distance between an endpoint on a skeleton line and all endpoints on other skeleton lines includes: Calculate the distance between an endpoint on a skeleton line and all endpoints on other skeleton lines, and select endpoints whose distance is less than a distance threshold as candidate endpoints for that endpoint.

5. The method for segmenting brain tumor lesion regions based on CT image analysis according to claim 1, characterized in that, The step of obtaining the suspected truncated endpoint of an endpoint based on the angle between an endpoint and the tangents of each candidate endpoint of that endpoint includes: Obtain the angle between the tangent of an endpoint and the tangent of a candidate endpoint of that endpoint, and use it as the tangent angle; use an exponential function with the natural constant as the base to perform a negative correlation mapping on the ratio of the square of the tangent angle to the angle decay parameter to obtain the directional consistency index of a candidate endpoint of that endpoint; if the directional consistency index of a candidate endpoint of that endpoint is greater than the directional consistency threshold, then the candidate endpoint is a suspected truncated endpoint of that endpoint.

6. The method for segmenting brain tumor lesion regions based on CT image analysis according to claim 1, characterized in that, The step of obtaining each tumor region of an endpoint based on all disordered pixels of that endpoint includes: Morphological processing is performed on all the disordered pixels of an endpoint to obtain different mask regions; if a mask region contains a suspected truncated endpoint of the endpoint, then the mask region is a tumor part region.

Images