A stomach CT image intelligent analysis system

Abstract

The application relates to the technical field of medical image processing, in particular to a stomach CT image intelligent analysis system, which comprises a region segmentation module, a boundary line adjustment module, a boundary line determination module and an image analysis module; the region segmentation module is used for segmenting a CT image to obtain an initial boundary line; the boundary line adjustment module is used for adjusting the initial boundary line based on the feature similarity on both sides of the initial boundary line; the boundary line determination module is used for determining the initial boundary line as a target boundary line in the case that the initial boundary line meets a preset condition; and the image analysis module is used for performing image analysis on the CT image based on the target boundary line. The stomach CT image intelligent analysis system can improve the accuracy of distinguishing the lymphatic region and the stomach wall region.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, and in particular to a stomach CT image intelligent analysis system. BACKGROUND

[0002] Segmenting the stomach wall and perigastric lymphatic tissue in a computed tomography (CT) image can accurately assess the development stage of gastric cancer. It helps doctors determine the depth of tumor invasion into the stomach wall and lymph node metastasis, so as to develop more accurate treatment plans, improve treatment effectiveness, and reduce unnecessary surgical risks.

[0003] In related technologies, the stomach region and lymphatic tissue region can be segmented based on density by threshold segmentation.

[0004] However, in the above method, the stomach wall and lymphatic tissue are both medium soft tissue density in plain CT, and the traditional threshold segmentation method cannot effectively distinguish them, resulting in high boundary segmentation error. SUMMARY

[0005] In order to solve the problem that the stomach wall region and perigastric lymphatic tissue density are similar in the traditional method in the CT image and cannot be effectively distinguished, the purpose of the present application is to provide a stomach CT image intelligent analysis system, and the technical solution adopted is as follows:

[0006] The embodiment of the present application provides a stomach CT image intelligent analysis system, which comprises a region segmentation module, a boundary line adjustment module, a boundary line determination module and an image analysis module; the region segmentation module is used for segmenting a CT image to obtain an initial boundary line, the initial boundary line being a boundary line between an initial stomach wall region and an initial lymphatic region; the boundary line adjustment module is used for adjusting the initial boundary line based on the feature similarity on both sides of the initial boundary line; the boundary line determination module is used for determining the initial boundary line as a target boundary line in the case that the initial boundary line meets a preset condition, the preset condition being that the feature similarity on both sides of the initial boundary line is less than or equal to a feature similarity threshold, or being the minimum feature similarity within a preset adjustment width; and the image analysis module is used for performing image analysis on the CT image based on the target boundary line.

[0007] Optionally, the boundary line adjustment module is specifically used for adjusting the initial boundary line based on a sliding window, so that the feature similarity on both sides of the initial boundary line in each window meets the preset condition.

[0008] Optionally, the boundary line adjustment module comprises a similarity determination submodule and an adjustment submodule; the similarity determination submodule is configured to determine a first feature similarity on both sides of an initial boundary line in the first window, the first window being any one window of the sliding window on the initial boundary line; the similarity determination submodule is further configured to, in a case where the first feature similarity is greater than the feature similarity threshold, determine a second feature similarity of a region in the first window with respect to the initial gastric wall region and a third feature similarity of the region in the first window with respect to the initial lymph node region; and the adjustment submodule is configured to adjust the initial boundary line in the first window based on the second feature similarity and the third feature similarity.

[0009] Optionally, the adjustment submodule is specifically configured to, in a case where the second feature similarity is greater than the third feature similarity, move the initial boundary line in the first window towards the initial lymph node region based on a preset step length; and in a case where the third feature similarity is greater than the second feature similarity, move the initial boundary line in the first window towards the initial gastric wall region based on a preset step length.

[0010] Optionally, the boundary line determination module comprises an adjustment width determination submodule and a target boundary line determination submodule; the adjustment width determination submodule is configured to determine a total adjustment width of the initial boundary line in the first window after each adjustment of the initial boundary line in the first window; and the target boundary line determination submodule is configured to, in a case where the first feature similarity is greater than the feature similarity threshold and the total adjustment width is greater than or equal to a preset adjustment width, determine, as the target boundary line in the first window, an initial boundary line corresponding to a minimum value of the first feature similarity in the historical adjustment times.

[0011] Optionally, the similarity determination submodule is specifically configured to determine a roughness and a CT value on both sides of the initial boundary line in the first window, respectively; and determine the first feature similarity based on the roughness and the CT value.

[0012] Optionally, the similarity determination submodule is specifically configured to determine a roughness difference value and a CT value difference value on both sides of the initial boundary line in the first window; and determine the first feature similarity based on the roughness difference value and the CT value difference value.

[0013] Optionally, the similarity determination submodule is specifically configured to perform grayscale processing on the CT image to obtain a grayscale value of each pixel point in the CT image; and determine a roughness of any one side region in the first window based on a difference value of grayscale values of each pair of adjacent pixel points in the any one side region.

[0014] Optionally, the region segmentation module is specifically configured to: perform threshold segmentation on the CT image to obtain a stomach wall candidate region; and perform a morphological closing operation on the stomach wall candidate region to obtain the initial boundary line.

[0015] Optionally, the image analysis module is specifically configured to: determine a distance between each lymphatic tissue and the target boundary line; determine lymphatic tissues with a distance less than a distance threshold from the target boundary line; determine a risk coefficient based on a number and a size of the lymphatic tissues; and determine an image analysis result of the CT image based on the risk coefficient.

[0016] The present application has the following beneficial effects:

[0017] In the embodiments of the present application, the stomach wall region and the lymphatic region are initially segmented to obtain an initial boundary line which is not accurate enough, and then the feature similarity of the stomach wall region and the lymphatic region on both sides of the initial boundary line is determined, and the initial boundary line is dynamically adjusted according to the size of the feature similarity, so that the feature similarity on both sides of the initial boundary line is small, and at this time, the initial boundary line can accurately distinguish the lymphatic region and the stomach wall region, and the result obtained by image analysis based on the adjusted boundary line is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings.

[0019] Figure 1 FIG. 1 is a structural schematic diagram of a stomach CT image intelligent analysis system provided by an embodiment of the present application;

[0020] Figure 2 FIG. 2 is a structural schematic diagram of another stomach CT image intelligent analysis system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined purposes, the specific embodiments, structures, features and effects of the stomach CT image intelligent analysis system according to the present application are described in detail below in combination with the drawings and preferred embodiments. 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.

[0022] 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.

[0023] Perigastric lymphoid tissue is an important component of the immune system. It is widely distributed throughout all layers of the stomach wall, particularly abundant in the mucosa and submucosa, and extends into the perigastric region. Composed of immune cells, lymphatic vessels, and lymphatic structures, lymphoid tissue is not a continuous whole but rather consists of scattered tissues and discrete lymphatic structures, and is extensively present in the perigastric region.

[0024] Accurate segmentation of the gastric wall and perigastric lymphoid tissue regions is of great significance in patient treatment. Currently, gastric CT image analysis mainly relies on traditional image processing techniques, including threshold-based region segmentation, edge detection algorithms (such as the Canny operator), and region growing methods. Some studies attempt to combine morphological operations to optimize gastric wall contour extraction or utilize machine learning models (such as support vector machines) to classify pre-segmented regions. In clinical practice, doctors typically manually annotate the gastric wall and lymphoid tissue regions in CT images and use contrast-enhanced scans to observe the degree of tissue enhancement to aid in judgment.

[0025] However, existing techniques for segmenting the stomach wall and lymphatic tissue have the following shortcomings: First, both the stomach wall and lymphatic tissue exhibit moderate soft tissue density (CT value 30-90 Hounsfield units, HU) on plain CT scans, resulting in insufficient contrast within the stomach wall and perigastric lymphatic tissue regions, making traditional thresholding methods ineffective in differentiation. Second, lymphatic tissue is dispersed and widely distributed around the stomach, closely adjacent to the anatomical location of the stomach wall, leading to overlapping of edge contours and making effective differentiation impossible. Doctors rely on experience to manually delineate the boundaries, which is time-consuming and highly subjective. Especially in obese patients or cases of gastric wall edema, tissue contrast is further reduced, leading to a significant increase in segmentation errors using traditional algorithms.

[0026] The following description, in conjunction with the accompanying drawings, details the intelligent analysis system for gastric CT images provided in this application.

[0027] like Figure 1 As shown, the intelligent analysis system 10 for gastric CT images includes a region segmentation module 101, a boundary adjustment module 102, a boundary determination module 103, and an image analysis module 104.

[0028] Specifically, the region segmentation module 101 is used to segment the CT image to obtain the initial boundary line.

[0029] The initial boundary line is the boundary between the initial gastric wall region and the initial lymphatic region. This initial boundary line is a somewhat inaccurate boundary line obtained based on the segmentation process.

[0030] It should be understood that the CT image is a stomach wall CT image.

[0031] In an optional implementation, in order to better segment the stomach wall region and the lymph region, the CT image can be pre-processed to a certain extent.

[0032] Optionally, the stomach wall CT image can be subjected to non-local mean denoising processing. Based on the processing, the texture details of the CT image can be retained while reducing the influence of noise on the stomach wall CT image.

[0033] Optionally, adaptive histogram equalization can also be used for image processing. Based on the processing, the edge contrast of the stomach wall and the lymph tissue can be enhanced.

[0034] In an implementation of the embodiment of the present application, the region segmentation module 101 is specifically configured to perform threshold segmentation on the CT image to obtain a stomach wall candidate region, and perform morphological closing operation on the stomach wall candidate region to obtain the initial boundary line.

[0035] Optionally, based on the tissue density of the stomach wall, a threshold method can be used to extract a region with a CT value of -100-100HU, which is determined as the stomach wall candidate region.

[0036] It should be understood that the morphological closing operation can bridge narrow discontinuities and long gullies, eliminate small holes, fill in the broken contour lines, and based on the morphological closing operation, the stomach wall candidate region boundary can be smoothed to determine the boundary as the initial boundary line.

[0037] It can be understood that based on the initial boundary line, the initial stomach wall region and the initial lymph region can also be obtained.

[0038] In an optional implementation, based on the initial boundary line, the lymph region can be expanded outward by 4cm to obtain a perigastric region, internal organs such as liver and pancreas based on prior data segmentation are excluded from the perigastric region to obtain a perigastric region excluding internal organs. A pixel point set with a CT value of 30-90HU is extracted to obtain a suspected pixel point set of perigastric lymph tissue. The pixel point set is subjected to a density-based noise application spatial clustering algorithm (DBSCAN) to determine a region where a connected domain with a diameter of 0.5-2cm is located as the initial lymph region.

[0039] It should be understood that since the CT value of fat is usually less than -30HU and the CT value of calcified blood vessels is usually greater than 150HU, the pixel point set extracted based on the CT value of 30-150HU can exclude the interference of fat and calcification / blood vessels, and the obtained initial lymph region is more accurate.

[0040] Optionally, while a 150 HU threshold can exclude calcification, it may unnecessarily include high-density soft tissue, reducing the specificity of the extraction. Since the density characteristics of lymphoid tissue are 30~90 HU, the pixel set can also be extracted based on the CT value of 30~90 HU, which can more accurately match the density characteristics of lymphoid tissue while minimizing the interference of non-lymphoid soft tissue and blood vessels.

[0041] The boundary adjustment module 102 is used to adjust the initial boundary line based on the feature similarity on both sides of the initial boundary line.

[0042] It should be understood that since the tissue density of the stomach wall and the perigastric lymphatic tissue are similar and they have a close anatomical relationship, the initial boundary line obtained based on the above-mentioned regional segmentation is not very accurate. Therefore, the initial boundary line can be adjusted based on the feature similarity on both sides of the initial boundary line to obtain a more accurate boundary line.

[0043] It is understandable that lymphatic tissue and the stomach wall are different structures, and their characteristic parameters are also different. This characteristic similarity is the similarity between the characteristic parameters of the initial lymphatic region and the initial stomach wall region on both sides of the boundary line.

[0044] It should be understood that when the feature similarity is high, it indicates that the feature parameters of the regions on both sides of the initial boundary are similar, and the regions on both sides of the initial boundary may be of the same structure. When the feature similarity is low, it indicates that the feature parameters of the regions on both sides of the initial boundary are significantly different, and the regions on both sides of the initial boundary may be of different structures.

[0045] It is understandable that if the feature similarity on both sides of the initial boundary line is greater than or equal to the similarity threshold, the initial boundary line can be determined to be relatively accurate. In this case, the initial boundary line can be determined as the target boundary line.

[0046] In one alternative implementation, the initial boundary line can be moved to the left or right to adjust it.

[0047] The boundary line determination module 103 is used to determine the initial boundary line as the target boundary line when the initial boundary line meets the preset conditions.

[0048] The preset condition is that the feature similarity on both sides of the initial boundary line is less than or equal to the feature similarity threshold, or is the minimum feature similarity within the preset adjustment width.

[0049] It should be understood that when the feature similarity is less than or equal to the similarity threshold, it indicates that the feature parameters of the regions on both sides of the initial boundary line are significantly different, and the regions on both sides of the initial boundary line are different structures. In other words, the boundary line accurately distinguishes the lymphatic region and the gastric wall region. At this time, the initial boundary line can be determined as the target boundary line.

[0050] Understandably, in the process of adjusting the initial boundary line, in order to avoid the boundary line deviating too much and leaving the stomach wall area, a preset adjustment width can be set, and the most suitable boundary line can be found within the preset adjustment width, that is, the initial boundary line corresponding to the minimum feature similarity. At this time, the initial boundary line can be determined as the target boundary line.

[0051] The image analysis module 104 is used to perform image analysis on the CT image based on the target boundary line.

[0052] It should be understood that this target boundary line can accurately distinguish between the lymphatic region and the gastric wall region. Based on this, imaging analysis can be performed on either the lymphatic region or the gastric wall region.

[0053] It is understandable that CT imaging is a key means of assessing tumor invasion depth and lymph node metastasis in the diagnosis of gastric cancer. In one implementation of this application, the image analysis module 104 is specifically used to: determine the distance between each lymph node tissue and the target boundary line; identify lymph nodes tissues whose distance from the target boundary line is less than a distance threshold; determine a risk coefficient based on the number and size of the lymph nodes tissues; and determine the image analysis result of the CT image based on the risk coefficient.

[0054] It should be understood that, based on the final target boundary line, the distance between each lymphatic tissue and the stomach wall can be clearly determined, thereby determining the number and size of the lymphatic tissue closely attached to the stomach wall.

[0055] Under normal circumstances, there are approximately 150-200 lymph nodes distributed around the stomach, closely attached to the stomach wall. However, in pathological conditions, if these lymph nodes are close to the stomach wall, the time for cancer cell spread is significantly shortened, greatly increasing the risk of metastasis compared to lymph nodes farther from the stomach wall. Therefore, the number and size of lymph nodes close to the stomach wall can be used to assess the risk factor for cancer cells in the stomach wall.

[0056] It should be understood that this risk factor is used to characterize the magnitude of the risk of cancer cell spread.

[0057] Optionally, the risk coefficient satisfies the following formula:

[0058]

[0059] in, Indicates the risk coefficient. Indicates the size of lymphoid tissue. This indicates the amount of lymphoid tissue.

[0060] Optionally, the size and number of lymphoid tissues can be standardized separately, and the risk coefficient T of gastric wall cancer cells can be normalized based on prior clinical data. Then, the risk can be registered and determined based on the risk coefficient.

[0061] It should be understood that this risk level is included in the image analysis structure.

[0062] Optionally, when the risk level is greater than or equal to 0.7, it can be assessed as a severe spread risk; when the risk level is less than 0.7 but greater than or equal to 0.5, it can be assessed as a general spread risk; and when the risk level is less than 0.5, it can be assessed as having a slight spread risk.

[0063] Understandably, different levels of risk of spread can guide doctors to adopt different treatment strategies and reduce unnecessary surgical risks.

[0064] In this embodiment, the stomach wall region and the lymphatic region are first initially segmented to obtain an initial boundary line that is not precise enough. Then, the feature similarity between the stomach wall region and the lymphatic region on both sides of the initial boundary line is determined. The initial boundary line is dynamically adjusted based on the magnitude of the feature similarity so that the feature similarity on both sides of the initial boundary line is small. At this time, the initial boundary line can accurately distinguish the lymphatic region and the stomach wall region. The results obtained by image analysis based on the adjusted boundary line are more accurate.

[0065] In conjunction with the description of the above embodiments, when the boundary line adjustment module 102 is used to adjust the initial boundary line based on the feature similarity on both sides of the initial boundary line, it is specifically used to: adjust the initial boundary line based on a sliding window so that the feature similarity on both sides of the initial boundary line in each window meets the preset condition.

[0066] Specifically, the initial boundary line within each sliding window is adjusted by sliding the sliding window along the initial boundary line.

[0067] It is understandable that the initial boundary line may be an arc. Therefore, we can first determine the tangent of the initial boundary line and the center width line, align the center width line of the sliding window with the tangent, and then slide it horizontally along the initial boundary line.

[0068] For example, the size of the sliding window can be set to a width of 5 pixels and a length of 20 pixels.

[0069] It is understandable that when the initial boundary line within each window meets the preset conditions, the entire initial boundary line also meets the preset conditions.

[0070] It should be understood that by adjusting the initial boundary line based on the sliding window, the initial boundary line can be divided into multiple segments. Adjusting based on these smaller segments of the initial boundary line allows for smaller adjustment ranges, higher accuracy, and more efficient acquisition of initial boundary lines that meet preset conditions.

[0071] The following example, using the adjustment of the first window, illustrates the process by which the boundary adjustment module 102 adjusts the initial boundary line. (Combined with...) Figure 1 ,like Figure 2 As shown, the boundary adjustment module 102 includes a similarity determination submodule 1021 and an adjustment submodule 1022.

[0072] The similarity determination submodule 1021 is used to determine the first feature similarity on both sides of the initial boundary line within the first window, wherein the first window is any window of the sliding window on the initial boundary line.

[0073] It should be understood that the initial boundary line divides the first window into two regions (the initial lymphoid subregion and the initial gastric wall subregion), and the first feature similarity is the similarity of feature parameters between the initial lymphoid subregion and the initial gastric wall subregion.

[0074] In this embodiment of the application, the feature parameter can be roughness and CT value.

[0075] Understandably, the stomach wall and perigastric lymphatic tissue have different tissue compositions. The stomach wall consists of multiple layers, including the mucosa, submucosa, muscularis propria, and serosa. Under normal conditions, these layers are tightly fused together, forming a relatively uniform and smooth surface. On CT images, the stomach wall typically has a uniform thickness and relatively consistent density, resulting in a smooth texture and minimal roughness in the image. In contrast, the perigastric lymphatic tissue mainly includes mucosa-associated lymphoid tissue and local lymph nodes. These tissues are composed of numerous small lymphoid follicles and cell clusters, with a discrete, granular structure. In CT images, this structure leads to an uneven surface, exhibiting greater roughness.

[0076] Therefore, for the initially delineated initial gastric wall region and initial lymphatic region, the roughness of the two regions can be used to distinguish between the gastric wall and lymphatic tissue.

[0077] It should be understood that because the gastric wall is dense while the lymphatic tissue has a loose reticular structure, and because the gastric wall is richly supplied with blood while the lymphatic tissue is relatively poorly supplied with blood, the affinity of contrast agents for the two regions also differs to some extent. These characteristics lead to a certain difference in the average CT values ​​of the two regions. Therefore, the average CT value between the two regions can be calculated to distinguish the characteristics of that region to a certain extent.

[0078] In one alternative implementation, the similarity determination submodule 1021 is specifically used to determine the roughness and CT value on both sides of the initial boundary line within the first window; and to determine the first feature similarity based on the roughness and the CT value.

[0079] Specifically, the similarity determination submodule 1021 can first perform grayscale processing on the CT image to obtain the grayscale value of each pixel in the CT image; then, based on the difference in grayscale values ​​of each pair of adjacent pixels in any side region within the first window, determine the roughness of any side region within the first window.

[0080] Based on the description of the above embodiments, when segmenting the CT image, the CT image has already been processed into grayscale. At this time, if the CT image is processed into grayscale again, the CT image can be compressed from 256 grayscale levels to 8 levels.

[0081] In one alternative implementation, in order to comprehensively calculate the coarseness of pixel pairs in all directions without repetition, the grayscale values ​​of pixel pairs between adjacent pixels in each of the four directions can be calculated separately.

[0082] Optionally, the four directions are 0° (right), 45° (bottom right), 90° (bottom), and 135° (bottom left).

[0083] It is understood that the contrast in the gray-level co-occurrence matrix can reflect the roughness of the image surface texture. In the embodiments of this application, the contrast can be used to measure the roughness of the image. The greater the roughness, the rougher the texture; the smaller the roughness, the smoother the texture.

[0084] In this embodiment of the application, a gray-level co-occurrence matrix can be constructed by scanning a grayscale image, recording the grayscale value of each pixel and its specified neighboring pixels, and obtaining gray-level co-occurrence matrices in four directions.

[0085] Optionally, the contrast in one direction satisfies the following formula:

[0086]

[0087] in, Indicates contrast in one direction. and These represent the grayscale values ​​of a single pixel. This represents the maximum grayscale value. Indicates grayscale value and The probability that pairs of pixels co-occur at a specific direction and distance.

[0088] For example, after performing grayscale processing on the CT image as described above, the grayscale is compressed to 8 levels. The value is 8.

[0089] Optionally, the The value can be obtained by counting and statistically analyzing the gray-level co-occurrence matrix.

[0090] Optionally, after obtaining the contrast in the four directions, the average of the four contrasts can be determined as the roughness.

[0091] In one implementation of this application, when the similarity determination submodule 1021 is used to determine the roughness difference and CT value difference on both sides of the initial boundary line within the first window, it is specifically used to determine the first feature similarity based on the roughness and the CT value.

[0092] Optionally, the average CT value on both sides of the initial boundary line within the first window can be determined first, and then the difference between the average CT values ​​can be determined.

[0093] Optionally, feature similarity is determined to satisfy the following formula:

[0094]

[0095] in, Indicates the area With the region Feature similarity, Indicates the region internal roughness, Indicates the region internal roughness, Indicates the area The average CT value within, Indicates the area The average CT value within.

[0096] It should be understood that the roughness and CT values ​​in the above calculation formulas are normalized dimensionless data, such as those normalized by the maximum and minimum values.

[0097] It should be understood that This indicates a region. With the region The difference in roughness, Indicates the area With the region The smaller the difference in roughness and the difference in CT values, the higher the feature similarity between the two regions.

[0098] Understandably, when When the value equals 0, it means that the features of the two regions are completely identical. In this case, the feature similarity can be calculated. The preset maximum similarity value is determined. When the value is not equal to 0, the feature similarity is calculated based on the above formula. .

[0099] Furthermore, the aforementioned preset maximum similarity value is a value greater than the feature similarity threshold. Because when A value of 0 indicates that the texture and density features on both sides of the boundary are completely identical (which can also be understood as belonging to the same tissue), and the similarity is extremely high. By assigning a maximum value (e.g., 10000) to the preset maximum similarity value, it can be ensured that the preset maximum similarity value is greater than the feature similarity threshold, thereby determining that this is not a boundary and prompting the algorithm to continue moving the window to find the true boundary.

[0100] The similarity determination submodule 1021 is further configured to determine the second feature similarity between the region within the first window and the initial stomach wall region, and the third feature similarity between the region and the initial lymph region, respectively, when the first feature similarity is greater than the feature similarity threshold.

[0101] Understandably, if the first feature similarity is greater than the feature similarity threshold, it means that the structures on both sides of the initial boundary line within the first window are quite similar and may be the same structure. The initial boundary line may be inaccurate. In this case, it can be determined whether the region within the first window is the stomach wall region or the lymphatic region.

[0102] It should be understood that the second feature similarity is the feature similarity between all regions within the first window and the initial stomach wall region, and the third feature similarity is the feature similarity between all regions within the first window and the initial lymphatic region.

[0103] It should be noted that the methods for determining the similarity of the second and third features are similar to those for determining the similarity of the first feature, and will not be repeated here.

[0104] Adjustment submodule 1022 is used to adjust the initial boundary line within the first window based on the second feature similarity and the third feature similarity.

[0105] It should be understood that, based on whether the area within the first window resembles a stomach wall region or a lymphatic region, it can be determined which side the initial boundary line is currently shifting to, and based on this shift, the initial boundary line within the first window is adjusted.

[0106] Specifically, if the similarity of the second feature is greater than that of the third feature, the initial boundary line within the first window is moved towards the initial lymphatic region based on a preset step size. If the similarity of the third feature is greater than that of the second feature, the initial boundary line within the first window is moved towards the initial gastric wall region based on a preset step size.

[0107] It is understandable that if the similarity of the second feature is greater than that of the third feature, it means that the region within the first window is more similar to the initial stomach wall region. The initial boundary line within the first window may shift towards the initial stomach wall region. In this case, in order to correct the initial boundary line, the initial boundary line can be moved towards the initial lymphatic region.

[0108] If the similarity of the third feature is greater than that of the second feature, it indicates that the region within the first window is more similar to the initial lymphatic region. The initial boundary line within the first window may shift towards the initial lymphatic region. In this case, to correct the initial boundary line, it can be moved towards the initial gastric wall region.

[0109] For example, the preset step size can be 5 pixels.

[0110] In an alternative implementation, feature similarity can be used as a preset step size weight, and the initial boundary line within the first window can be moved based on the weight of a larger feature similarity and the preset step size.

[0111] For example, assuming the preset step size can be 5 pixels, the second feature similarity is greater than the third feature similarity, and the second feature similarity is 80%, then the initial boundary line in the first window can be moved 4 pixels towards the initial lymphatic region.

[0112] In one implementation of this application, there may be a situation where the initial boundary line is frequently moved in the same direction, but the first feature similarity is always greater than the feature similarity threshold. In this case, in order to constrain the movement space of the stomach wall boundary, a preset adjustment width can be set, and adjustments can be made within the preset adjustment width.

[0113] Specifically, the aforementioned boundary line determination module 103 includes an adjustment width determination submodule and a target boundary line determination submodule.

[0114] The width adjustment determination submodule is used to determine the total adjustment width of the initial boundary line within the first window after each adjustment.

[0115] Understandably, after each adjustment of the initial boundary line within the first window, the current adjustment distance and the width between the current boundary line and the initial boundary line within the first window can be recorded.

[0116] The target boundary line determination submodule is used to determine the initial boundary line corresponding to the minimum value of the first feature similarity in the historical adjustment count as the target boundary line within the first window, when the first feature similarity is greater than the feature similarity threshold and the total adjustment width is greater than or equal to the preset adjustment width.

[0117] It should be understood that when the first feature similarity is greater than the feature similarity threshold, the boundary line adjustment module will iteratively adjust the initial boundary line. When the initial boundary line moves to the limit (i.e., the preset adjustment width), the iteration will stop, and the minimum first feature similarity in the historical adjustment count will be determined. The initial boundary line corresponding to the minimum first feature similarity will be determined as the target boundary line within the first window.

[0118] It should be noted that the lymphatic region includes the extramural lymphatic region and the intramural lymphatic region. Based on the intelligent analysis system of gastric CT images provided in this application embodiment, the boundary between the intramural lymphatic region and the gastric wall region, as well as the boundary between the extramural lymphatic region and the gastric wall region, can be determined respectively.

[0119] In summary, the intelligent analysis system for gastric CT images provided in this application first segments the gastric wall region and the lymphatic tissue region based on the different characteristics of the gastric wall region and the lymphatic tissue region in CT images, according to their positional relationship and average tissue density. It then establishes a similarity score between the boundary pixel region of the gastric wall and the features of the gastric wall region and the lymphatic tissue region. Utilizing the similarity between the boundary pixel region and the two regions, and through a dynamic sliding window, it dynamically adjusts the boundary contour between the two regions, thereby achieving precise segmentation of the gastric wall and the perigastric lymphatic tissue.

[0120] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0121] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An intelligent analysis system for gastric CT images, characterized in that, The system includes a region segmentation module, a boundary adjustment module, a boundary determination module, and an image analysis module. The region segmentation module is used to segment the CT image to obtain an initial boundary line, which is the boundary line between the initial gastric wall region and the initial lymphatic region. The boundary adjustment module is used to adjust the initial boundary line based on the feature similarity between the gastric wall region and the lymphatic region on both sides of the initial boundary line. The boundary line determination module is used to determine the initial boundary line as the target boundary line when the initial boundary line meets the preset conditions. The preset conditions are that the feature similarity on both sides of the initial boundary line is less than or equal to the feature similarity threshold, or the minimum feature similarity within a preset adjustment width. The image analysis module is used to perform image analysis on the CT image based on the target boundary line.

2. The intelligent analysis system for gastric CT images according to claim 1, characterized in that, The boundary line adjustment module is specifically used for: The initial boundary line is adjusted based on a sliding window so that the feature similarity on both sides of the initial boundary line within each window meets the preset condition.

3. The intelligent analysis system for gastric CT images according to claim 2, characterized in that, The boundary line adjustment module includes a similarity determination submodule and an adjustment submodule; The similarity determination submodule is used to determine the first feature similarity on both sides of the initial boundary line within the first window, wherein the first window is any one of the sliding windows on the initial boundary line. The similarity determination submodule is further configured to determine, when the first feature similarity is greater than the feature similarity threshold, the second feature similarity between the region within the first window and the initial stomach wall region, and the third feature similarity between the region and the initial lymph region. The adjustment submodule is used to adjust the initial boundary line within the first window based on the second feature similarity and the third feature similarity.

4. The intelligent analysis system for gastric CT images according to claim 3, characterized in that, The adjustment submodule is specifically used for: If the similarity of the second feature is greater than that of the third feature, the initial boundary line within the first window is moved toward the initial lymphatic region based on a preset step size. If the similarity of the third feature is greater than that of the second feature, the initial boundary line within the first window is moved toward the initial stomach wall region based on a preset step size.

5. The intelligent analysis system for gastric CT images according to claim 3, characterized in that, The boundary line determination module includes an adjustment width determination submodule and a target boundary line determination submodule; The adjustment width determination submodule is used to determine the total adjustment width of the initial boundary line in the first window after each adjustment of the initial boundary line in the first window; The target boundary determination submodule is used to determine the initial boundary corresponding to the minimum value of the first feature similarity in the historical adjustment count as the target boundary within the first window when the first feature similarity is greater than the feature similarity threshold and the total adjustment width is greater than or equal to the preset adjustment width.

6. The intelligent analysis system for gastric CT images according to claim 3, characterized in that, The similarity determination submodule is specifically used for: Determine the roughness and CT value on both sides of the initial boundary line within the first window, respectively; The first feature similarity is determined based on the roughness and the CT value.

7. The intelligent analysis system for gastric CT images according to claim 6, characterized in that, The similarity determination submodule is specifically used for: Determine the roughness difference and CT value difference on both sides of the initial boundary line within the first window; The first feature similarity is determined based on the roughness difference and the CT value difference.

8. The intelligent analysis system for gastric CT images according to claim 6, characterized in that, The similarity determination submodule is specifically used for: The CT image is converted to grayscale to obtain the grayscale value of each pixel in the CT image; The roughness of any one side of the first window is determined based on the difference in grayscale values ​​of each pair of adjacent pixels in any one side region.

9. The intelligent analysis system for gastric CT images according to claim 1, characterized in that, The region segmentation module is specifically used for: Threshold segmentation is performed on the CT image to obtain candidate regions of the stomach wall; The initial boundary line is obtained by performing a morphological closing operation on the candidate regions of the stomach wall.

10. The intelligent analysis system for gastric CT images according to claim 1, characterized in that, The image analysis module is specifically used for: Determine the distance of each lymphoid tissue from the target boundary line; Identify lymphoid tissues whose distance from the target boundary line is less than a distance threshold; The risk factor is determined based on the number and size of the lymphoid tissue. Based on the risk coefficient, the image analysis results of the CT image are determined.

Images