Gynecological CT image detection method and system for auxiliary diagnosis

Abstract

The invention provides a gynecological CT image detection method and system for auxiliary diagnosis, and relates to the technical field of medical image processing, and the method comprises the steps: S1, obtaining a plain-scan CT image in the process of CT scanning of the pelvic cavity of a female patient; s2, obtaining an initial boundary line of the contour of the uterus region on the plain-scan CT image; s3, performing region fusion on pixel points on the initial boundary line according to the gray difference absolute value to obtain a pixel region block; s4, detecting whether the average gray value of the pixel region blocks is greater than a threshold value or not, if so, taking the pixel region block with the maximum average gray value as a suspected region, and turning to step S5; if not, turning to step S6; s5, pixel points in the suspected area and pixel points on the next boundary line are subjected to extended gray similarity matching in sequence; and S6, sequentially carrying out gray similarity matching on pixel points in the pixel region blocks and adjacent pixel points on the next boundary line.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a method and system for detecting gynecological CT images for auxiliary diagnosis. Background Art

[0002] Gynecological CT images include plain CT images and enhanced CT images. When taking plain CT images, it is not necessary to inject iodine contrast agent, while when taking enhanced CT images, it is necessary to inject iodine contrast agent, which can make the gray value difference between the parts flowing through the blood and the surrounding pixel points more obvious. Therefore, currently, it is usually to perform a preliminary localization of the lesion on the plain CT image, and then perform feature analysis of the lesion on the enhanced CT image based on the preliminary localization. However, since some lesions are not clear enough on the plain CT image, if target detection of the lesion is used, the range of the target detection is too narrow, resulting in incomplete position information obtained by localization, affecting subsequent feature analysis on the enhanced CT image, imperfect data, and inaccurate detection results. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for detecting gynecological CT images for auxiliary diagnosis. The gynecological CT image detection method mainly performs preliminary localization of the lesion on the plain CT image, expands the position information of the lesion while performing preliminary localization, ensures obtaining a large amount of data related to the target, and performs feature classification of the target on the enhanced CT image through data assistance to improve the accuracy of the detection result.

[0004] To solve the above technical problems, the present invention adopts the following solutions:

[0005] A method for detecting gynecological CT images for auxiliary diagnosis, the gynecological CT image detection method includes the following steps:

[0006] S1. Obtain a plain CT image during the process of CT scanning the pelvic cavity of a female patient;

[0007] S2. Obtain the contour of the uterine region through image recognition processing on the plain CT image, extend one pixel point inward from the contour of the uterine region to obtain an initial boundary line, and obtain the absolute value of the gray difference between all two adjacent pixel points on the initial boundary line;

[0008] S3. Perform regional fusion on the pixel points on the initial boundary line according to the absolute value of the gray difference to obtain several pixel region blocks with partial overlap;

[0009] S4. Detect whether the average gray value of the pixel region block is greater than a threshold. If so, take the pixel region block with the largest average gray value as the suspected region and go to step S5; if not, go to step S6;

[0010] S5. Extend one pixel point inward from the initial boundary line to obtain the next boundary line. Perform extended gray-scale similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region.

[0011] S6. Extend one pixel point inward from the initial boundary line to obtain the next boundary line. Perform gray-scale similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block, and then go to step S4.

[0012] Further, in S2, the process of the image recognition processing is as follows:

[0013] Preprocess the plain CT image, extract the image edge of the plain CT image through edge recognition, then extract the uterine region on the plain CT image through threshold segmentation and morphological operations, and then extract, screen, smooth and fill the contour of the uterine region to obtain the contour of the uterine region.

[0014] Further, in S2, the process of extending one pixel point inward from the contour of the uterine region to obtain the initial boundary line is as follows:

[0015] Extract all the contour pixel points from the contour of the uterine region, calculate the normal direction perpendicular to the contour for each contour pixel point, and thus move one pixel point inward along this normal direction to obtain the corresponding pixel point, and connect all the pixel points by a line to obtain the initial boundary line.

[0016] Further, the step S3 includes the following steps:

[0017] S31. Find the maximum absolute value of the gray-scale difference among all the absolute values of the gray-scale differences on the first boundary, and use it as the first absolute value of the gray-scale difference.

[0018] S32. Starting from both sides of the maximum absolute value of the gray-scale difference respectively, sequentially perform threshold judgment on the absolute values of the gray-scale differences to determine whether it is greater than the threshold. If so, record this absolute value of the gray-scale difference as the second absolute value of the gray-scale difference.

[0019] S33. Perform region fusion on the pixel points according to the first absolute value and the second absolute value of the gray-scale difference on the initial boundary line to obtain several pixel region blocks with partial overlaps.

[0020] Further, in S33, the process of the region fusion is as follows:

[0021] Obtain two corresponding adjacent pixel points according to the absolute value of the first gray - level difference, and draw a splitting line in the middle of the two adjacent pixel points; obtain two corresponding adjacent pixel points according to the absolute value of the second gray - level difference and draw a splitting line in the middle of the two adjacent pixel points;

[0022] Cut all the pixel points on the initial boundary line into several connected pixel - region blocks through the splitting line;

[0023] Fuse the adjacent pixel points in the connected pixel - region blocks corresponding to a pixel - region block into the pixel - region block through the splitting line to obtain several pixel - region blocks with partial overlap.

[0024] Further, in S4, the gray - level average value is the average value of the gray - level values of all pixel points in the pixel - region block.

[0025] Further, in S5, the process of the extended gray - level similarity matching is as follows:

[0026] Find the adjacent pixel points on the next boundary line by extending inwards from the pixel points in the suspected region;

[0027] Exclude the pixel points on both sides in the suspected region, and match the other pixel points in the suspected region with the first adjacent pixel point on the next boundary line;

[0028] Calculate the absolute value of the gray - level difference between the other pixel points in the suspected region and the first adjacent pixel point on the next boundary line, calculate the absolute value of the gray - level difference between the pixel points on both sides in the suspected region and the second adjacent pixel point on the next boundary line, and match the pixel points on both sides in the suspected region with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray - level differences;

[0029] Calculate the absolute value of the gray - level difference between the first pixel point and the second pixel point on the next boundary line, calculate the absolute value of the gray - level difference between the second pixel point and the third adjacent pixel point on the next boundary line, and match the third pixel point on the next boundary line with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray - level differences.

[0030] Further, in S6, the process of the gray - level similarity matching is as follows:

[0031] Find the adjacent pixel points on the next boundary line by extending inwards from the pixel points in the pixel - region block;

[0032] Exclude the pixel points on both sides in the pixel - region block, and match the other pixel points in the pixel - region block with the first adjacent pixel point on the next boundary line;

[0033] Calculate the absolute value of the gray - level difference between other pixel points in the pixel region block and the first adjacent pixel point on the next boundary line, calculate the absolute value of the gray - level difference between the pixel points on both sides in the pixel region block and the second adjacent pixel point on the next boundary line, and match the pixel points on both sides in the pixel region block with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray - level differences.

[0034] Further, it further includes step S7, which is after S6, specifically:

[0035] S7. Inject an iodine contrast agent during the CT scan of the pelvic cavity of a female patient to obtain an enhanced CT image, and extract the region in the enhanced CT image according to the position information of the suspected region in the plain - scan CT, which is convenient for detecting the current region.

[0036] A gynecological CT image detection system for auxiliary diagnosis, applying the described gynecological CT image detection method for auxiliary diagnosis, includes:

[0037] CT image acquisition module: Obtain a plain - scan CT image during the CT scan of the pelvic cavity of a female patient, and obtain an enhanced CT image by injecting an iodine contrast agent during the CT scan of the pelvic cavity of a female patient

[0038] Image recognition and processing module: Obtain the contour of the uterine region through image recognition and processing on the plain - scan CT image, extend one pixel point inward from the contour of the uterine region to obtain an initial boundary line, and obtain the absolute value of the gray - level difference between all two adjacent pixel points on the initial boundary line;

[0039] Region fusion module: Perform region fusion on the pixel points on the initial boundary line according to the absolute value of the gray - level difference to obtain several pixel region blocks with partial overlap;

[0040] Gray - level average value detection module: Detect whether the gray - level average value of the pixel region block is greater than the threshold;

[0041] Suspected region update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform extended gray - level similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region.

[0042] Pixel region block update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform gray - level similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block.

[0043] Advantages of the present invention:

[0044] The present invention provides a method and system for detecting gynecological CT images for auxiliary diagnosis, which improves the detection method in the prior art that uses plain scan CT images for preliminary lesion localization and then performs feature analysis on the lesions on enhanced CT images based on the preliminary localization. It mainly changes the target detection for plain scan CT images in the prior art to target extension detection. By means of target extension detection, while realizing the localization of the lesions, the position information is selectively extended to the surrounding according to the gray similarity, avoiding the loss of position information due to the unclear lesion targets on the plain scan CT images, ensuring a large amount of data related to the target, and then classifying the features of the target on the data-assisted enhanced CT images to improve the accuracy of the detection results. Description of the Drawings

[0045] Figure 1 It is a schematic flow chart of the method for detecting gynecological CT images in Embodiment 1 of the present invention. Detailed Embodiments

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The description of at least one exemplary embodiment below is actually only illustrative and in no way limits the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0047] Unless otherwise specifically stated, the relative arrangements, numerical expressions and values of the components and steps described in these embodiments do not limit the scope of the present invention.

[0048] At the same time, it should be understood that for the sake of description, the sizes of the various parts shown in the drawings are not drawn in actual proportional relationships.

[0049] In addition, for the sake of clarity and conciseness, the descriptions of well-known structures, functions and configurations may be omitted. Those of ordinary skill in the art will recognize that various changes and modifications can be made to the examples described herein without departing from the spirit and scope of the present disclosure.

[0050] The techniques, methods and devices known to those of ordinary skill in the relevant fields may not be discussed in detail, but where appropriate, the said techniques, methods and devices should be regarded as part of the authorization specification.

[0051] In all the examples shown and discussed here, any specific value should be construed as merely exemplary and not as a limitation. Thus, other examples of the exemplary embodiments may have different values.

[0052] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments:

[0053] Embodiment 1

[0054] Since gynecological diseases are important problems threatening women's health, early diagnosis and treatment are crucial. Therefore, in order to achieve automatic, accurate, and efficient diagnosis of gynecological diseases, a CT detection device is currently used to photograph female patients to obtain gynecological CT images. Among them, the gynecological CT images include plain scan CT images and enhanced CT images. Specifically, when taking plain scan CT images, no iodine contrast agent needs to be injected, while when taking enhanced CT images, an iodine contrast agent needs to be injected, which can make the gray value difference between the parts flowing through the blood and the surrounding pixel points more obvious.

[0055] Based on the above principle, currently, the target lesion is usually initially located on the plain scan CT image, and then the characteristics of the lesion are analyzed on the enhanced CT image based on the initial location. However, since some lesions are not clear enough on the plain scan CT image, if target detection of the lesion is used, the range of the target detection is too narrow, resulting in incomplete position information obtained by the positioning, affecting subsequent feature analysis on the enhanced CT image, imperfect data, and inaccurate detection results.

[0056] For example, when uterine fibroids appear on the plain scan CT image, if target recognition of the uterine fibroids is used, at this time, the density of the uterine fibroids on the plain scan CT image may be similar to that of the surrounding normal uterine tissue or other pelvic tissues (such as intestines, muscles), resulting in blurred boundaries and difficulty in distinction. If target detection of the uterine fibroids is used, the boundaries of the uterine fibroids are blurred, and the range of the target detection is too narrow, making it difficult for the detected area to contain all the characteristics of the uterine fibroids. The result of the target detection is incomplete, so that when the position information of the area is provided to the enhanced CT image subsequently, a region containing the complete uterine fibroids cannot be obtained on the enhanced CT image either. Thus, when automatically performing feature recognition and classification on the enhanced CT image, the basic data is insufficient, thereby affecting the accuracy of the detection results.

[0057] Therefore, in order to solve the problem that the detection of gynecological CT images is inaccurate due to unclear lesion targets on the plain scan CT image, the present invention proposes a method for detecting gynecological CT images for auxiliary diagnosis, as Figure 1 shown, the method for detecting gynecological CT images includes the following steps:

[0058] S1. Obtain a plain CT image during the CT scan of the pelvic cavity of a female patient;

[0059] S2. Obtain the contour of the uterine region through image recognition processing on the plain CT image, extend one pixel point inward from the contour of the uterine region to obtain an initial boundary line, and obtain the absolute value of the gray-level difference between all two adjacent pixel points on the initial boundary line;

[0060] S3. Perform regional fusion on the pixel points on the initial boundary line according to the absolute value of the gray-level difference to obtain several pixel region blocks with partial overlap;

[0061] S4. Detect whether the average gray value of the pixel region block is greater than the threshold. If so, use the pixel region block with the largest average gray value as the suspected region and go to step S5; if not, go to step S6;

[0062] S5. Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform extended gray-scale similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region.

[0063] S6. Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform gray-scale similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block, and go to step S4.

[0064] In one embodiment, it further includes step S7, which is after S6, specifically:

[0065] S7. Inject an iodine contrast agent during the CT scan of the pelvic cavity of a female patient to obtain an enhanced CT image, and perform regional extraction in the enhanced CT image according to the position information of the suspected region in the plain CT, so as to facilitate the detection of the current region. Specifically, the purpose of the suspected region is to provide position information for the enhanced CT image, so that the enhanced CT image can quickly locate the suspected region containing the lesion. Through the extended suspected region, it can assist the enhanced CT image to perform a comprehensive feature classification on the lesion and improve the detection accuracy.

[0066] Based on the above principle, the present invention selects to perform hierarchical detection on the plain CT image starting from the contour of the uterine region. The hierarchical detection refers to detecting by sequentially extending one pixel point inward from the contour of the uterine region. At this time, it extends inward from the contour of the uterine region. Since uterine fibroids are located within the uterus, uterine fibroids can be detected. After detecting each layer of pixel points, fusion or extended fusion is selected according to the gray-scale similarity between the pixel points. When the gray-scale similarity is met, the range of target detection is extended, which can ensure obtaining more target features. At the same time, target detection can also be performed on the extended region to ensure that the extended region contains the complete target. When the position information of the target is provided to the enhanced CT image, it can assist in locating the region containing the complete target on the enhanced CT image and assist in detecting the target features contained in this region, thereby improving the detection accuracy.

[0067] In one embodiment, in S2, the process of the image recognition and processing is as follows:

[0068] Preprocess the plain CT image, extract the image edge of the plain CT image through edge recognition, then extract the uterine region on the plain CT image through threshold segmentation and morphological operations, and then extract, screen, smooth, and fill the contour of the uterine region to obtain the contour of the uterine region.

[0069] Specifically, during the process of CT scanning the pelvic cavity of a female patient, the plain CT image is obtained. The pixel values of the read plain CT image are standardized within a fixed range, and then filtering algorithms such as Gaussian filtering and median filtering are used to remove image noise. Then, the image edge is extracted through an edge detection algorithm, and the edge detection algorithm can be Canny or Sobel. After that, through standard or multiple training results, the uterine region can be extracted according to the range of pixel values in the plain CT image. Next, morphological operations are used to remove noise or fill holes, and the morphological operations include closing operation, opening operation, etc. Finally, the contour of the uterine region is extracted and screened, smoothed, and filled to obtain the contour of the uterine region. Among them, the screening can be performed according to features such as the area and shape of the contour; the smoothing can use a contour approximation algorithm to smooth the contour; the filling refers to filling the interior of the contour to generate a mask of the uterine region.

[0070] In one embodiment, in S2, the process of extending one pixel point inward from the contour of the uterine region to obtain the initial boundary line is as follows:

[0071] Extract all the contour pixel points from the contour of the uterine region, calculate the normal direction perpendicular to the contour for each contour pixel point, and then move one pixel point inward along this normal direction to obtain the corresponding pixel point. Connect all the pixel points by lines to obtain the initial boundary line.

[0072] Specifically, in the present invention, first extend one pixel point inward from the contour of the uterine region to obtain the initial boundary line, and the initial boundary line is the closest to the contour. At this time, if the effect of extracting the contour of the uterine region is poor, and when the extracted contour of the uterine region is small, the initial boundary line may still be located on the inner wall of the uterus, resulting in a small absolute value of the gray-level difference between two adjacent pixel points on the initial boundary line; when the extracted contour of the uterine region is large, the initial boundary line exceeds the inner wall of the uterus and is located inside the uterus, and there may be mutation points on the initial boundary line, and the gray-level value of the pixel points is large, so there is a large absolute value of the gray-level difference.

[0073] Therefore, in the present invention, hierarchical detection is performed by extending inward from the contour of the uterine region. After detecting the initial boundary line, extend one pixel point inward from the initial boundary line to obtain the next boundary line, detect the next boundary line, and then, after detecting the next boundary line, extend one pixel point inward from the next boundary line, and so on, until all the pixel points inside the uterine region are detected and no further extension is possible.

[0074] In S5 and S6, when a suspected region or a pixel region block is updated at the next boundary line, one pixel point can be extended inward from the next boundary line to obtain the next-next boundary line, and then, hierarchical detection is performed on the next-next boundary line. Among them, in step S6, after the pixel region block is updated, return to step S4. The purpose is to detect whether the average gray level of each updated pixel region block is greater than the threshold. The pixel region block is the extended target region. Therefore, there is an overlapping relationship between the pixel region blocks. By detecting all the pixel region blocks at each layer through the average gray level, the average gray level of the uniform region can be preset. If the detected average gray level is greater than the threshold, it means that the pixel region block is non-uniform and the gray-level value of the pixel points is relatively high, and it may be a suspected region containing lesions. The larger the average gray level, the more non-uniform the region and the higher the gray-level value of the pixel points. Then, the pixel region block with the largest average gray level is used as the suspected region. After detecting the suspected region, return to step S5, and only expand the suspected region again under the condition of gray-level similarity in each layer of the boundary line in turn, so that the suspected region containing the complete lesion can be found quickly and accurately, and detection is performed based on this suspected region in the enhanced CT image.

[0075] In one embodiment, step S3 includes the following steps:

[0076] S31. Find the maximum absolute value of the gray - level differences among all the absolute values of gray - level differences on the first boundary, and use it as the first absolute value of gray - level difference.

[0077] S32. Starting from both sides of the maximum absolute value of gray - level difference respectively, sequentially perform a threshold judgment on the absolute values of gray - level differences to determine whether it is greater than the threshold. If so, record this absolute value of gray - level difference as the second absolute value of gray - level difference.

[0078] S33. Perform regional fusion on the pixel points according to the first absolute value of gray - level difference and the second absolute value of gray - level difference on the initial boundary line to obtain several pixel region blocks with partial overlap.

[0079] In one embodiment, in S33, the process of the regional fusion is as follows:

[0080] Obtain two corresponding adjacent pixel points according to the first absolute value of gray - level difference, and draw a cut - off line in the middle of the two adjacent pixel points; obtain two corresponding adjacent pixel points according to the second absolute value of gray - level difference and draw a cut - off line in the middle of the two adjacent pixel points.

[0081] Cut all the pixel points on the initial boundary line into several connected pixel region blocks through the cut - off lines.

[0082] Fuse the adjacent pixel points in the connected pixel region blocks corresponding to a pixel region block into this pixel region block through the cut - off lines to obtain several pixel region blocks with partial overlap.

[0083] Specifically, for example: the pixels on the current initial boundary line are marked as 1, 2, 3, 4, 5, 6, 7, and the grayscale values corresponding to the pixels are: 1, 1.4, 1.2, 1.9, 2.1, 1.5, 1.2, and the grayscale difference absolute values corresponding to two adjacent pixels are: 0.4, 0.2, 0.7, 0.2, 0.6, 0.3. Find the maximum grayscale difference absolute value, where the number of maximum grayscale difference absolute values may be 1 or more. At this time, the first grayscale difference absolute value is 0.7. Then, starting from both sides of 0.7, the grayscale difference absolute values are threshold-judged. If the second grayscale difference absolute value is 0.6, a dividing line is drawn between pixels 3 and 4 and between pixels 5 and 6 according to 0.7 and 0.6, respectively. Then, three partially overlapping pixel area blocks are obtained according to the dividing lines, respectively with pixel points 1 and 2 as the dividing lines. 1-4 is a pixel area block, pixel points 3-6 are a pixel area block, and pixel points 5-7 are a pixel area block. The purpose is to place mutation point 3 and mutation point 4 in different pixel area blocks. At this time, it is not known whether the pixel area block will evolve into a suspected area according to the pixel points on the next layer of boundary line. The pixel area block is expanded for the first time to increase the features of the pixel area block respectively; mutation point 5 and mutation point 6 are placed in different pixel area blocks. As above, the features of the pixel area block are increased respectively to ensure that each pixel area block contains mutation points. While containing mutation information, it also contains more features. Unstable information cannot be lost simply because of uniform pixels. In hierarchical analysis, more feature information is easily lost, resulting in poor auxiliary performance for enhanced CT images and failure to improve their detection accuracy.

[0084] In one embodiment, in S5, the process of extending the grayscale similarity matching is:

[0085] Extend the pixels in the suspected area inward to find the adjacent pixels on the next boundary line;

[0086] Excluding the pixels on both sides of the suspected area, and matching the other pixels in the suspected area with the first adjacent pixel on the next boundary line;

[0087] Calculate the absolute value of the grayscale difference between other pixels in the suspected area and the first pixel point adjacent to the next boundary line, calculate the absolute value of the grayscale difference between the pixels located on both sides of the suspected area and the second pixel point adjacent to the next boundary line, and match the pixels located on both sides of the suspected area with the second pixel point adjacent to the next boundary line according to the similarity between the two grayscale difference absolute values;

[0088] Calculate the absolute value of the gray - scale difference between the first pixel point and the second pixel point on the next boundary line, calculate the absolute value of the gray - scale difference between the second pixel point on the next boundary line and the third pixel point adjacent to it, and match the third pixel point on the next boundary line with the adjacent second pixel point on the next boundary line according to the similarity between the two absolute values of the gray - scale differences.

[0089] Specifically, the process of finding adjacent pixel points on the next boundary line by extending pixel points in the suspected area inward is as follows: Similar to the above, calculate the normal direction perpendicular to the contour for each pixel point. Thus, move one pixel point inward along this normal direction to obtain the corresponding pixel point. Connect all pixel points by lines to obtain the next boundary line. Then, by moving sequentially along the corresponding normal directions, the boundary lines at corresponding levels can be obtained. It can be seen that a pixel point found downward from a pixel point in the suspected area along the same normal line is called an adjacent pixel point on the next boundary line.

[0090] Then, exclude the pixel points located on both sides in the suspected area. Specifically, in a pixel area block with pixel points 3 - 6 as a block, exclude pixel point 3 and pixel point 6. These pixel point 3 and pixel point 6 are the points where mutations occur with respect to pixel point 4 and pixel point 5 respectively, and are used as suspected feature points in this suspected area. Since there is a certain similarity between pixel point 4 and pixel point 5, pixel point 4 and pixel point 5 are not excluded, and pixel point 4 and pixel point 5 are matched with the adjacent pixel points on the next boundary line.

[0091] Since there is similarity between other pixel points, calculate the absolute value of the gray - scale difference between other pixel points and the adjacent pixel points on the next boundary line, and judge whether the pixel points on both sides match the adjacent second pixel point on the next boundary line according to this absolute value of the gray - scale difference. Specifically, if the absolute value of the gray - scale difference between other pixel points and the adjacent pixel points on the next boundary line is relatively similar to the absolute value of the gray - scale difference between the pixel points on both sides and the adjacent second pixel point on the next boundary line, it means that the second pixel point has similarity with the suspected feature point, and then they are matched; otherwise, they are not matched. The purpose is to represent the basic difference between two - layer boundary lines by the absolute value of the gray - scale difference between pixel points with a certain similarity, and based on this basic difference, analyze whether the mutation points on both sides of the next - layer boundary line are suspected feature points. If so, they are matched; if not, they can be discarded.

[0092] At this time, if the matching is successful and there are also suspected feature points on the next boundary line, then the suspected feature points can be expanded again. This expansion refers to horizontal expansion. The points adjacent to the suspected feature points are judged according to the absolute value of the gray-scale difference between the pixel points on each layer. That is, the absolute value of the gray-scale difference between the first pixel point and the second pixel point is judged against the third pixel point adjacent to it. When the absolute value of the gray-scale difference between the first pixel point and the second pixel point is relatively similar to the absolute value of the gray-scale difference between the second pixel point and the third pixel point adjacent to it, it means that the third pixel point has a similarity with the suspected feature point, and then they are matched; otherwise, they are not matched. The purpose is to represent the absolute value of the gray-scale difference between the boundary lines of this layer through the absolute value of the gray-scale difference between pixel points with a certain similarity. Thus, horizontal expansion is carried out again to both sides according to the absolute value of the gray-scale difference to obtain more suspected feature points, so as to ensure obtaining a large amount of data related to the target. Then, the features of the target are classified on the data-assisted enhanced CT image to improve the accuracy of the detection result.

[0093] In one embodiment, in S6, the process of gray-scale similarity matching is as follows:

[0094] Find the adjacent pixel points on the next boundary line by extending the pixel points in the pixel region block inward;

[0095] Exclude the pixel points located on both sides in the pixel region block, and match the other pixel points in the pixel region block with the first pixel point adjacent to the next boundary line;

[0096] Calculate the absolute value of the gray-scale difference between the other pixel points in the pixel region block and the first pixel point adjacent to the next boundary line, calculate the absolute value of the gray-scale difference between the pixel points located on both sides in the pixel region block and the second pixel point adjacent to the next boundary line, and match the pixel points located on both sides in the pixel region block with the second pixel point adjacent to the next boundary line according to the similarity between the two absolute values of the gray-scale differences.

[0097] Specifically, in S6, since the detected average gray-scale values of the pixel region blocks are not greater than the threshold, it means that there is no target in the current pixel region block and it is relatively uniform. By setting the threshold, the feature points of the suspected target in the uterine region can be obtained through the gray-scale values of the pixel points. If not detected, the current pixel region block can not be expanded. Since the current pixel region block is an expanded region block and there is repetition between them, the current pixel region block can be directly traversed downward in turn, and every feature point will not be missed, ensuring the detection of every feature point.

[0098] Embodiment 2

[0099] A gynecological CT image detection system for auxiliary diagnosis, applying the described method for detecting gynecological CT images for auxiliary diagnosis, includes:

[0100] CT image acquisition module: Obtain plain scan CT images during the CT scan of the pelvic cavity of female patients, and obtain enhanced CT images by injecting iodine contrast agent during the CT scan of the pelvic cavity of female patients

[0101] Image recognition and processing module: Obtain the contour of the uterine region through image recognition and processing on the plain scan CT image, extend one pixel point inward from the contour of the uterine region to obtain the initial boundary line, and obtain the absolute value of the gray level difference between all two adjacent pixel points on the initial boundary line;

[0102] Region fusion module: Perform region fusion on the pixel points on the initial boundary line according to the absolute value of the gray level difference to obtain several pixel region blocks with partial overlap;

[0103] Gray average value detection module: Detect whether the gray average value of the pixel region block is greater than the threshold;

[0104] Suspected region update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform extended gray similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region.

[0105] Pixel region block update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform gray similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block.

[0106] As described above, it is only a preferred embodiment of the present invention, and does not impose any form of limitation on the present invention. Based on the technical essence of the present invention, within the spirit and principle of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiments still fall within the protection scope of the technical solution of the present invention.

Claims

1. A gynecological CT image detection method for auxiliary diagnosis, characterized in that, The described gynecological CT image detection method includes the following steps: S1. Obtain a plain scan CT image during the CT scan of a female patient's pelvic cavity; S2. Obtain the contour of the uterine region through image recognition processing on the plain scan CT image, extend one pixel point inward from the contour of the uterine region to obtain an initial boundary line, and obtain the absolute value of the gray level difference between all two adjacent pixel points on the initial boundary line; S3. Perform regional fusion on the pixel points on the initial boundary line according to the absolute value of the gray level difference to obtain several pixel region blocks with partial overlap; S4. Detect whether the average gray level of the pixel region block is greater than the threshold. If so, take the pixel region block with the largest average gray level as the suspected region and go to step S5; if not, go to step S6; S5. Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform extended gray level similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region. S6. Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform gray level similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block, and go to step S4.

2. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, wherein In S2, the process of the image recognition processing is as follows: Preprocess the plain scan CT image, extract the image edge of the plain scan CT image through edge recognition, then extract the uterine region on the plain scan CT image through threshold segmentation and morphological operations, and then extract, screen, smooth and fill the contour of the uterine region to obtain the contour of the uterine region.

3. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, characterized in that, In S2, the process of extending one pixel point inward from the contour of the uterine region to obtain the initial boundary line is as follows: Extract all contour pixel points from the contour of the uterine region, calculate the normal direction perpendicular to the contour for each contour pixel point, and thus move one pixel point inward along this normal direction to obtain the corresponding pixel point, and connect all pixel points by a line to obtain the initial boundary line.

4. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, characterized in that, The step S3 includes the following steps: S31. Find the largest absolute value of the gray level difference among all the absolute values of the gray level differences on the first boundary, and take it as the first absolute value of the gray level difference; S32. Starting from both sides of the largest absolute value of the gray level difference respectively, sequentially perform threshold judgment on the absolute values of the gray level differences to determine whether it is greater than the threshold. If so, record this absolute value of the gray level difference as the second absolute value of the gray level difference; S33. Perform regional fusion on the pixel points according to the first absolute value of the gray level difference and the second absolute value of the gray level difference on the initial boundary line to obtain several pixel region blocks with partial overlap.

5. The gynecological CT image detection method for auxiliary diagnosis according to claim 4, wherein In S33, the process of the regional fusion is as follows: Obtain the corresponding two adjacent pixel points according to the first absolute value of the gray level difference, and make a cut line in the middle of the two adjacent pixel points; obtain the corresponding two adjacent pixel points according to the second absolute value of the gray level difference and make a cut line in the middle of the two adjacent pixel points; All pixel points on the initial boundary line are divided into several connected pixel region blocks by the cutting line; Adjacent pixel points in the connected pixel region blocks corresponding to a pixel region block are fused into the pixel region block by the cutting line to obtain several pixel region blocks with partial overlap.

6. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, wherein, In S4, the average gray value is the average of the gray values of all pixel points in the pixel region block.

7. A gynecological CT image detection method for auxiliary diagnosis according to claim 1, characterized in that, In S5, the process of extended gray similarity matching is as follows: Find the adjacent pixel points on the next boundary line by extending inward from the pixel points in the suspected region; Exclude the pixel points located on both sides in the suspected region, and match the other pixel points in the suspected region with the first adjacent pixel point on the next boundary line; Calculate the absolute value of the gray difference between the other pixel points in the suspected region and the first adjacent pixel point on the next boundary line, calculate the absolute value of the gray difference between the pixel points located on both sides in the suspected region and the second adjacent pixel point on the next boundary line, and match the pixel points located on both sides in the suspected region with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray differences; Calculate the absolute value of the gray difference between the first pixel point and the second pixel point on the next boundary line, calculate the absolute value of the gray difference between the second pixel point and the third adjacent pixel point on the next boundary line, and match the third pixel point on the next boundary line with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray differences.

8. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, wherein In S6, the process of gray similarity matching is as follows: Find the adjacent pixel points on the next boundary line by extending inward from the pixel points in the pixel region block; Exclude the pixel points located on both sides in the pixel region block, and match the other pixel points in the pixel region block with the first adjacent pixel point on the next boundary line; Calculate the absolute value of the gray difference between the other pixel points in the pixel region block and the first adjacent pixel point on the next boundary line, calculate the absolute value of the gray difference between the pixel points located on both sides in the pixel region block and the second adjacent pixel point on the next boundary line, and match the pixel points located on both sides in the pixel region block with the second adjacent pixel point on the next boundary line according to the similarity between the two absolute values of the gray differences.

9. The gynecological CT image detection method for auxiliary diagnosis according to claim 1, wherein It further includes step S7, which is after S6, specifically: S7. Inject an iodine contrast agent during the CT scan of the pelvic cavity of a female patient to obtain an enhanced CT image, and perform region extraction in the enhanced CT image according to the position information of the suspected region in the plain CT scan to facilitate the detection of the current region.

10. A gynecological CT image detection system for auxiliary diagnosis, characterized in that, Apply a method for detecting gynecological CT images for auxiliary diagnosis as described in any one of claims 1-9, including: CT image acquisition module: Obtain a plain CT image during the CT scan of the pelvic cavity of a female patient, and obtain an enhanced CT image by injecting an iodine contrast agent during the CT scan of the pelvic cavity of a female patient Image recognition processing module: Obtain the contour of the uterine region through image recognition processing on the plain CT image, extend one pixel point inward from the contour of the uterine region to obtain the initial boundary line, and obtain the absolute value of the gray difference between all two adjacent pixel points on the initial boundary line; Region fusion module: Perform region fusion on the pixel points on the initial boundary line according to the absolute value of the gray difference to obtain several pixel region blocks with partial overlap; Gray average value detection module: Detect whether the gray average value of the pixel region block is greater than the threshold; Suspected region update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform extended gray similarity matching between the pixel points in the suspected region and the pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the suspected region to update the suspected region. Pixel region block update module: Extend one pixel point inward from the initial boundary line to obtain the next boundary line, perform gray similarity matching between the pixel points in the pixel region block and the adjacent pixel points on the next boundary line, and combine the pixel points on the next boundary line obtained by the matching with the pixel region block to update the pixel region block.

Images