A tumor cell screening detection processing method, system and storage medium

Abstract

The present invention discloses a tumor cell screening detection processing method, system and storage medium, the method comprising: obtaining an ovarian image to be detected and obtaining a historical ovarian image set; respectively selecting feature images of the historical ovarian image set and the ovarian image to be detected, obtaining a historical target image and a target frame image to be detected, and comparing them to obtain a first matching value; extracting the contour of the ovarian image to be detected, obtaining an ovarian contour image to be detected; comparing and matching the historical target image with the detected ovarian contour image, obtaining a second matching value; comparing the first matching value with the second matching value, obtaining the ovarian state of the ovarian image to be detected, which is a normal ovary, a suspected cystic ovary, a confirmed cystic ovary, a mildly cystic ovary, or a severely cystic ovary, so as to perform medical treatment for different ovarian states.

Description

Technical Field

[0001] The present invention relates to the field of electronic image recognition of ovarian cells, and in particular to a tumor cell screening, detection and processing method, system and storage medium. Background Art

[0002] Research has found that when examining ovarian cancer with color Doppler ultrasound, irregular masses will be found on the ovaries, and the boundaries of the irregular masses are unclear and spread to the surrounding areas; while continuing to observe electronic images, irregular liquid dark areas can be seen in the masses, and there are also messy light clusters and light spots in the dark areas. Electronic images show that there are obvious and abundant blood flow signals in the masses; if auxiliary testing is continued and tumor markers are used to test the mass to confirm the mass, tumor markers such as CA125 will be found to be significantly elevated. Therefore, screening for malignant tumor cells in the ovaries is very important, but screening for benign tumors in ovarian cells should not be ignored either;

[0003] Among them, ovarian cysts are benign tumors in ovarian cells. Whether the ovarian cyst is serious needs to be judged in combination with the volume of the ovarian cyst. If the ovarian cyst is relatively small, the diameter does not exceed 5 cm, and there is no obvious discomfort, it is generally not serious; but if the ovarian cyst is large, the diameter exceeds 5 cm, or the ovarian cyst ruptures and the pedicle torsions, it may cause severe pain, bleeding and other symptoms. In this case, if the treatment is not timely, it may be life-threatening and is generally more serious.

[0004] Ovarian cysts are benign tumor cells, but screening is equally important. Research has found that due to inaccurate detection and identification of the ovaries, the state of the ovarian cells is incorrectly judged, such as: normal ovarian cell state, cells with lesions that produce mild cysts, or severe cyst cells. These changes in ovarian cells cannot be effectively detected, thus delaying subsequent treatment that cannot target the true state of the ovarian cyst.

[0005] Therefore, due to the complexity and diversity of ovarian cysts, there is a problem of inaccurate identification. This may lead to misdiagnosis, that is, misidentifying normal ovarian cells as cystic cells, or misclassifying cystic cells as other states. For example, it is impossible to accurately determine whether ovarian cells are normal, mildly cystic, or severely cystic, which may lead to delayed treatment or incorrect treatment decisions.

[0006] With the advancement of technology, more optimized image processing algorithms are also used in existing technologies; for example, deep learning and training models are combined to automatically identify and classify different cells in ovarian images. However, researchers further believe that it is not enough to simply classify and identify whether it is a normal ovarian cell or an ovarian image of a cyst. Such rough detection is still needed to further accurately analyze the specific survival status information of the ovarian image, which is more meaningful for subsequent diagnosis and treatment. Summary of the invention

[0007] The object of the present invention is to provide a tumor cell screening, detection and processing method, system and storage medium, which solve the above-mentioned technical problems pointed out in the prior art.

[0008] The present invention provides a tumor cell screening, detection and processing method, comprising the following steps:

[0009] Obtain an ovarian image to be detected and obtain a historical ovarian image set;

[0010] The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images;

[0011] Feature images of the historical ovarian image set and the ovarian image to be detected are respectively framed to obtain historical target images and target frame images to be detected, and compared to obtain a first matching value; contours of the ovarian image to be detected are extracted to obtain an ovarian contour image to be detected; the historical target image is compared and matched with the detected ovarian contour image to obtain a second matching value; the first matching value is compared with the second matching value to obtain the ovarian state of the ovarian image to be detected.

[0012] Preferably, feature images are selected for the historical ovarian image set and the ovarian image to be detected respectively to obtain a historical target image and a target frame image to be detected, and the comparison is performed to obtain a first matching value. The specific operation steps are as follows:

[0013] Obtain a preset YOLO model; select a candidate frame for the ovarian image to be detected according to the YOLO model to obtain multiple candidate frames to be detected, each of which contains different category labels; annotate the multiple candidate frames to be detected with confidence scores, and divide the confidence score of each candidate frame to be detected into different score intervals;

[0014] Dividing the interval levels according to the different category labels and the different score intervals;

[0015] Calculate the average value based on the confidence scores of all candidate boxes to be detected within each interval score;

[0016] The average of the confidence scores of all candidate boxes to be detected within each interval score corresponds to the interval level;

[0017] Select the candidate frames to be detected that are close to the average value of the confidence scores within each interval score, determine them as the target frames to be detected within each interval score, and crop the target frames to be detected to obtain the target images to be detected;

[0018] Performing candidate box selection on each historical ovarian image in the historical ovarian image set by using the YOLO model, obtaining multiple historical ovarian candidate boxes for each historical ovarian image, and marking confidence scores for the historical ovarian candidate boxes; selecting the historical ovarian candidate box with the highest confidence score as the historical target image;

[0019] The historical target images include: historical normal ovarian target images, historical suspected cystic ovarian images, historical confirmed cystic ovarian images, historical mild cystic ovarian images, and historical severe cystic ovarian images;

[0020] Each historical image in the historical target images is compared with the target frame image to be detected within each interval score to obtain a first matching value.

[0021] Preferably, the interval levels are divided into normal ovarian interval level, suspected cystic ovarian interval level, confirmed cystic ovarian interval level, mild cystic ovarian interval level, and severe cystic ovarian interval level.

[0022] Preferably, the contour of the ovarian image to be detected is extracted to obtain the ovarian contour image to be detected; the historical target image is compared and matched with the detected ovarian contour image to obtain a second matching value. The specific operation steps are as follows:

[0023] Extracting contour features of the ovarian image to be detected by edge detection algorithm to obtain the ovarian contour to be detected; expanding and filling the ovarian contour to be detected to obtain the contour image to be detected;

[0024] The to-be-detected contour image is compared with each of the historical target images to obtain a second matching value.

[0025] Preferably, the first matching value is compared with the second matching value to obtain the ovarian status of the ovarian image to be detected, specifically including: performing weighted evaluation calculation on the first matching value and the second matching value to obtain a comprehensive matching value; and determining the ovarian status of the ovarian image to be detected according to the comprehensive matching value.

[0026] Preferably, a weighted evaluation calculation is performed on the first matching value and the second matching value to obtain a comprehensive matching value. The specific operation steps are as follows:

[0027] Normalize the first matching value and the second matching value; weight the normalized first matching value and the second matching value, set the weight of the first matching value to w1, and the weight of the second matching value to w2, and calculate the comprehensive matching value C, the calculation formula is:

[0028] C = w1×first matching value+w2×second matching value;

[0029] The comprehensive matching values ​​include: normal ovary comprehensive matching value, suspected comprehensive matching value, confirmed comprehensive matching value, slight comprehensive matching value, and severe comprehensive matching value.

[0030] Preferably, the ovarian status of the ovarian image to be detected is determined according to the comprehensive matching value, and the specific operation steps are as follows:

[0031] The normal ovary threshold q, the suspected cystic ovary threshold w, the confirmed cystic ovary threshold e, and the mild cystic ovary threshold r are preset;

[0032] Determine whether the normal ovary comprehensive matching value is less than the normal ovary threshold q;

[0033] If yes, it is determined that the ovarian image to be detected is a normal ovary;

[0034] If not, determining whether the suspected comprehensive matching value is less than the suspected cystic ovarian threshold w;

[0035] If yes, it is determined that the ovarian image to be detected is a suspected cystic ovary;

[0036] If not, determining whether the determined comprehensive matching value is less than the determined cystic ovarian threshold e;

[0037] If yes, it is determined that the ovarian image to be detected is a confirmed cystic ovary;

[0038] If not, determining whether the mild comprehensive matching value is less than the mild cystic ovarian threshold r;

[0039] If yes, it is determined that the ovarian image to be detected is a mild cystic ovary;

[0040] If not, it is determined that the ovarian image to be detected is a severe cystic ovary.

[0041] Preferably, the expansion and filling of the to-be-detected ovarian contour to obtain the to-be-detected contour image specifically includes the following operations:

[0042] First, the initial expansion and filling parameters are preset; the expansion and filling parameters include the size and shape of the structure element;

[0043] Directly call the initial expansion and filling parameters; directly expand and fill the current ovarian contour to be detected using the current initial expansion and filling parameters to obtain the initial contour image to be detected; detect whether the initial contour image to be detected has excessive expansion and filling, and if so, adjust the expansion and filling parameters, continue to use multiple iterations to gradually expand and fill the current ovarian contour to be detected, and finally output the contour image to be detected.

[0044] Preferably, the expansion and filling parameters are adjusted, and the expansion and filling of the current ovarian contour to be detected is gradually performed using multiple iterations, and finally the contour image to be detected is output, which specifically includes the following steps:

[0045] The size of the structural element in the current initial expansion filling parameter is proportionally reduced T times to obtain the target value of the structural element required for each iteration;

[0046] First, the maximum value among the target values ​​of several structural elements is called and used as the parameter of the first iteration of expansion and filling; then the first expansion and filling is performed on the current ovarian contour to be detected to obtain the contour image to be detected of the first iteration; it is determined whether the contour image to be detected of the current first iteration has excessive expansion and filling, and if it still has excessive expansion and filling; the second largest value among the target values ​​of several structural elements is continued to be selected, and the second expansion and filling is continued to be obtained to obtain the contour image to be detected of the second iteration; at the same time, it is continuously iterated to detect whether the contour image to be detected obtained by the current iteration has excessive expansion and filling, until the current contour image to be detected is determined to have no excessive expansion and filling, and the contour image to be detected obtained by the last iteration filling is determined as the final contour image to be detected.

[0047] Accordingly, the present invention also proposes a tumor cell screening, detection and processing system, comprising: an image acquisition module, a matching degree module, and a final result module;

[0048] The image acquisition module is used to acquire the ovarian image to be detected and acquire the historical ovarian image set;

[0049] The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images;

[0050] The matching degree module is used to respectively select feature images of the historical ovarian image set and the ovarian image to be detected, obtain the historical target image and the target frame image to be detected, and compare them to obtain a first matching degree value; extract the contour of the ovarian image to be detected to obtain the ovarian contour image to be detected; compare and match the historical target image with the detected ovarian contour image to obtain a second matching degree value;

[0051] The final result module is used to compare the first matching value with the second matching value to obtain the ovarian status of the ovarian image to be detected.

[0052] Accordingly, the present application also relates to a storage medium, which includes:

[0053] Memory for storing computer programs;

[0054] A communication interface, used to connect the memory to the processor;

[0055] The processor is used to execute the computer program to implement the steps of the above-mentioned tumor cell screening, detection and processing method.

[0056] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0057] From the analysis of the above-mentioned tumor cell screening detection processing method provided by the present invention, it can be known that in specific applications, the candidate frame is selected for the ovarian image to be detected through the YOLO model, and each candidate frame will have a confidence score, and the interval score is divided by the confidence score, and the interval level is set by the interval score. Because, in the ovarian image to be detected, the candidate frame of the confidence score will be in each corner of the ovarian image to be detected, so different confidence scores will appear. The ovarian cell state at each position in the ovarian image to be detected can be judged by different confidence scores, so that the candidate frame of the confidence score is divided by the interval level, and the candidate frame with the average confidence score is selected as the representative candidate frame of the score interval, and at the same time as the cell state of each score interval; and the candidate frame is selected for the historical ovarian image set to obtain the historical target image, and finally matched to obtain the first matching value;

[0058] Further, the contour of the ovarian image to be detected is extracted by an edge detection algorithm, and the contour is expanded and filled with pixels to obtain the contour image to be detected. By filling the contour with pixels, the contour image to be detected can be matched with the historical target image, so that the incompatibility when matching the contour and the image can be avoided, and the matching can be smoothly performed to obtain the second matching value;

[0059] Furthermore, the processes for obtaining the first matching value and the second matching value are different, so the first matching value and the second matching value cannot be matched directly. Therefore, it is necessary to calculate the comprehensive matching value through a weighted method, and judge whether the ovary is a normal ovary, a suspected cystic ovary, a confirmed cystic ovary, a mild cystic ovary, or a severe cystic ovary through the comprehensive matching value and the preset ovarian cyst threshold, so as to carry out medical treatment for different ovarian conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0061] Figure 1 This is an overall flow chart of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0062] Figure 2 A matching degree flow chart of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0063] Figure 3 A schematic diagram of a confidence score of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0064] Figure 4 A schematic diagram of score intervals for a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0065] Figure 5 A schematic diagram of interval levels of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0066] Figure 6 A flow chart of determining the ovarian state of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0067] Figure 7 A contour filling flow chart of a tumor cell screening, detection and processing method provided in Example 1 of the present invention;

[0068] Figure 8 It is a contour filling process iteration diagram of a tumor cell screening detection processing method provided in the first embodiment of the present invention;

[0069] Fig. 9 A flow chart of a tumor cell screening, detection and processing system provided in Embodiment 2 of the present invention;

[0070] Fig.10 A storage medium provided in Embodiment 3 of the present invention.

[0071] Markings: image acquisition module 10, matching degree module 20, final result module 30; processor 1110; communication interface 1120; memory 1130; computer storage medium 1140. DETAILED DESCRIPTION

[0072] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0074] Embodiment 1

[0075] like Figure 1 As shown, the present invention proposes a tumor cell screening, detection and processing method, comprising the following steps:

[0076] S10: obtaining an ovarian image to be detected and obtaining a historical ovarian image set;

[0077] The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images;

[0078] It should be noted that the historical ovarian image set obtained includes the gradual change of the ovary from the normal ovary in history to the suspected cystic ovary (that is, there may be a risk of cystic ovary), then to the confirmed detection of cystic ovary, and then to the changes of ovarian cells in different stages from mild to severe deterioration of ovarian cells; through the five changes of ovarian cells in history, it is possible to determine which ovarian disease state the ovarian image to be detected belongs to, and different detection methods are implemented according to different ovarian disease states;

[0079] S20: Select feature images of the historical ovarian image set and the ovarian image to be detected respectively, obtain a historical target image and a target frame image to be detected, and compare them to obtain a first matching value; extract the contour of the ovarian image to be detected to obtain an ovarian contour image to be detected; compare and match the historical target image with the detected ovarian contour image to obtain a second matching value; compare the first matching value with the second matching value to obtain the ovarian state of the ovarian image to be detected;

[0080] It should be noted that the YOLO model is used to select features of candidate frames for the historical ovarian image set and the ovarian image to be detected. For feature frame selection of each ovarian image in the historical ovarian image set, since the historical ovarian image set is a complete and clear image, the ovarian position of a normal ovary or a cyst can be directly framed. At the same time, the ovarian image to be detected is also subjected to candidate frame selection, and the interval score is set by the confidence score. Multiple candidate frames in the ovarian image to be detected are selected by the interval score as the target frame to be detected. The historical ovarian image set and the target frame to be detected are cropped to form the historical target image and the target image to be detected, and the target image is compared to obtain the first matching value.

[0081] The contour of the ovarian image to be detected is extracted by edge detection algorithm to obtain the contour of the ovarian to be detected; the researchers believe that the contour of the ovarian to be detected cannot be matched and analyzed with the historical target image; therefore, the contour of the ovarian to be detected needs to be filled; the contour of the ovarian to be detected is filled by expansion operation, and the contour image to be detected can be obtained by filling by expansion operation; because, the contour image to be detected needs to be compared with the historical target image in the subsequent process to obtain a second matching value; and if it is the contour of the ovarian to be detected, there will be an inappropriate contrast difference (that is, one is a contour and the other is an image, so there will be an inappropriate situation when matching, and only after the contour or image is unified, there will be no inappropriate difference when matching), so the contour of the ovarian to be detected is filled by expansion operation to become the contour image to be detected, so that the images can be compared to obtain a second matching value;

[0082] Comprehensively judging the first matching value and the second matching value to judge the ovarian status of the ovarian image to be detected (i.e., whether it is a normal ovary, a mild cystic ovary, or a severe cystic ovary);

[0083] Specifically, Figure 2 As shown, in step S20, the historical ovarian image set and the ovarian image to be detected are respectively framed with feature images to obtain a historical target image and a target frame image to be detected, and the frame images are compared to obtain a first matching value; the contour of the ovarian image to be detected is extracted to obtain a detected ovarian contour image; the historical target image is compared and matched with the detected ovarian contour image to obtain a second matching value; the first matching value is compared with the second matching value to obtain the ovarian state of the ovarian image to be detected. The specific operation steps are as follows:

[0084] S21: Obtain a preset YOLO model; select candidate frames for the ovarian image to be detected according to the YOLO model to obtain multiple candidate frames to be detected, each of which contains different category labels; annotate the multiple candidate frames to be detected with confidence scores, and divide the confidence score of each candidate frame to be detected into different score intervals (such as Figure 3 shown);

[0085] S22: Divide the interval levels according to the different category labels and the different score intervals; the interval levels are divided into normal ovarian interval level, suspected cystic ovarian interval level, confirmed cystic ovarian interval level, mild cystic ovarian interval level, severe cystic ovarian interval level (such as Figure 5 shown);

[0086] It should be noted that, when the ovarian image to be detected is detected by the YOLO model (that is, the YOLO model is an existing model), many candidate frames will be selected, and each candidate frame will have a different category label, and the category label is normal ovary, suspected cystic ovary, confirmed cystic ovary, mild cystic ovary, and severe cystic ovary; because the ovarian image to be detected cannot determine the true state of the ovary, the candidate frame will select a lot of candidate frames to be detected, and each label of the category label may be the label state of the candidate frame to be detected;

[0087] And by marking the confidence score for each candidate frame to be detected and dividing different score intervals by the confidence score, the candidate frames to be detected can be classified more effectively, divided into 5 confidence score intervals. Because of the different confidence scores, the size of the scattered intervals will also be different. The score intervals are sorted from small to large, such as: the confidence score of the score interval is 0-20 for normal ovaries, the confidence score of the score interval is 20-40 for suspected cystic ovaries, the confidence score of the score interval is 40-60 for confirmed cystic ovaries, the confidence score of the score interval is 60-80 for mild cystic ovaries, and the confidence score of the score interval is 80-100 for severe cystic ovaries; the larger the confidence score of the score interval, the greater the possibility of ovarian cell lesions, such as Figure 4 As shown;

[0088] S23: Calculate the average value according to the confidence scores of all candidate boxes to be detected within each interval score;

[0089] The average value of the confidence scores of all candidate boxes to be detected within each interval score corresponds to the interval level (that is, because the interval level is obtained by dividing the interval score and the category, the average value of the confidence score of the interval score corresponds to the interval level);

[0090] Select the candidate frames to be detected that are close to the average value of the confidence scores within each interval score, determine them as the target frames to be detected within each interval score, and crop the target frames to be detected to obtain the target images to be detected;

[0091] It should be noted that different score intervals after classification can be used to determine the interval level with different category labels. Each score interval corresponds to a category label to form an interval level. The normal ovarian interval level is the smallest score interval, and so on. The severe cystic ovarian interval level is the largest score interval.

[0092] Therefore, it is necessary to calculate the average confidence scores within different score intervals, and determine the interval level by the average confidence scores, so the average confidence scores correspond to the interval level; by comparing the average confidence scores of different intervals, it is possible to more accurately determine which score interval corresponds to a more severe ovarian state;

[0093] Select candidate frames to be detected that are close to the average value of confidence scores within each interval score, select one candidate frame to be detected within each interval score as the target image to be detected, and select a total of 4 target images to be detected; each target image to be detected is within each corresponding interval level; and the candidate frames to be detected that are close to the average value of confidence scores are selected because they have relatively high representativeness in their corresponding category labels and interval levels, representing the average status of the candidate frames to be detected in the category labels and interval levels;

[0094] And because the average value of the confidence scores of all candidate boxes to be detected within each interval score corresponds to the interval level, the target image to be detected is also divided into normal ovary interval level, suspected cystic ovary interval level, confirmed cystic ovary interval level, mild cystic ovary interval level, and severe cystic ovary interval level (because they correspond to each other, each score interval has only one interval level, and they are arranged from small to large confidence scores; at the same time, the corresponding arrangement levels are also from normal ovary to suspected cystic ovary to the last severe cystic ovary);

[0095] S24: selecting a candidate frame for each historical ovarian image in the historical ovarian image set by using the YOLO model, obtaining multiple historical ovarian candidate frames for each historical ovarian image, and marking confidence scores for the historical ovarian candidate frames; selecting the historical ovarian candidate frame with the highest confidence score as the historical target image;

[0096] The historical target images include: historical normal ovarian target images, historical suspected cystic ovarian images, historical confirmed cystic ovarian images, historical mild cystic ovarian images, and historical severe cystic ovarian images;

[0097] It should be noted that the historical ovarian image set is the change from normal ovarian images to pathological and severe cystic ovarian images in history; therefore, after the YOLO model is used for frame selection, it is only necessary to select the historical ovarian candidate frame with the highest confidence score as the historical target image (that is, after the historical ovarian candidate frame with the highest confidence score is selected after frame selection, the historical ovarian candidate frame with the highest confidence score is cropped to form the historical target image);

[0098] S25: Compare each historical image in the historical target images with the target frame image to be detected within each interval score to obtain a first matching value;

[0099] It should be noted that the first matching value is obtained by comparing the historical normal ovarian target image, the historical suspected cystic ovarian image, the historical confirmed cystic ovarian image, the historical mild cystic ovarian image, the historical severe cystic ovarian image with the target frame image to be detected within each interval score by extracting features (such as color, texture, shape, etc.) (that is, it belongs to the prior art and will not be repeated here);

[0100] Because each target image to be detected corresponds to an interval level, the first matching value includes: the first matching value of the historical normal ovarian target image and the target frame image to be detected at the normal ovarian interval level, the first matching value of the historical suspected cystic ovarian image and the target frame image to be detected at the suspected cystic ovarian interval level, the first matching value of the historical determined cystic ovarian image and the target frame image to be detected at the determined cystic ovarian interval level, the first matching value of the historical mild cystic ovarian image and the target frame image to be detected at the mild cystic ovarian interval level, and the first matching value of the historical severe cystic ovarian image and the target frame image to be detected at the severe cystic ovarian interval level;

[0101] S26: extracting contour features from the image of the ovary to be detected by an edge detection algorithm to obtain a contour of the ovary to be detected; and expanding and filling the contour of the ovary to be detected to obtain a contour image to be detected;

[0102] It should be noted that the ovarian image to be detected is grayed, and the pixel points are clustered by the edge detection algorithm, and the pixel points are clustered by the change of the gray value, and the connected contour points are determined by setting a threshold, and the contour points are connected to obtain the ovarian contour to be detected (that is, it belongs to the prior art and will not be described in detail); because the ovarian contour to be detected is a contour, the expansion operation is filled to obtain the contour image to be detected, which can be compared with the subsequent historical target image to obtain the second matching value;

[0103] S27: Compare the contour image to be detected with each of the historical target images to obtain a second matching value;

[0104] It should be noted that, by comparing the historical normal ovarian target image, the historical suspected cystic ovarian image, the historical confirmed cystic ovarian image, the historical mild cystic ovarian image, the historical severe cystic ovarian image with the contour image to be detected (because the contour image to be detected is obtained by filling through the expansion operation, so features can be extracted) by extracting features (such as color, texture, shape, etc.) (that is, it belongs to the prior art and will not be repeated here), a second matching value is obtained;

[0105] At the same time, the contour image to be detected is the last one obtained after the contour is extracted, so there is only one image. The contour image to be detected is compared with multiple images of the historical target image one by one, and multiple second matching values ​​can be obtained (for example, the first matching value is obtained by obtaining each historical image (i.e., 5 images) in the historical target image and the target frame image to be detected (5 images) within each interval score, and 5 results of the first matching value are obtained (i.e., in the explanation of S25); the second matching value is the contour image to be detected (i.e., 1 image, because the ovarian image to be detected is directly extracted, so only the ovarian image to be detected is extracted). A contour image is compared with each historical image in the historical target image (i.e., 5 images) to obtain a plurality of second matching values ​​(i.e., the second matching values ​​include: a second matching value between the historical normal ovarian target image and the contour image to be detected, a second matching value between the historical suspected cystic ovarian image and the contour image to be detected, a second matching value between the historical confirmed cystic ovarian image and the contour image to be detected, a second matching value between the historical mild cystic ovarian image and the contour image to be detected, and a second matching value between the historical severe cystic ovarian image and the contour image to be detected);

[0106] S28: performing weighted evaluation calculation on the first matching value and the second matching value to obtain a comprehensive matching value; and determining the ovarian status of the ovarian image to be detected according to the comprehensive matching value;

[0107] It should be noted that the first matching value is obtained by the YOLO model by selecting candidate boxes for the ovarian image to be detected and the historical ovarian image set, and the first matching value obtained after the final cropping and comparison is obtained by image comparison; and the second matching value is obtained by comparing the contour image to be detected obtained by filling the contour image to be detected after the expansion operation after the contour is extracted from the ovarian image to be detected. Although the contour of the ovarian to be detected is filled, it may still affect the comparison between the final contour image to be detected and the historical target image (that is, when the contour of the ovarian to be detected is filled, the grayscale value of the internal pixel points may change or the texture and brightness may change, so that the final comprehensive matching value determines that the ovarian state of the ovarian image to be detected changes);

[0108] Therefore, the first matching value and the second matching value are different in the process of obtaining (i.e., one is an image obtained by cropping the candidate frame through the YOLO model, and the other is an image obtained by filling the contour expansion operation), so the first matching value and the second matching value need to be calculated for the comprehensive matching value, and the ovarian state of the ovarian image to be detected is judged by the final comprehensive matching value;

[0109] Specifically, Figure 6 As shown, in step S28, a comprehensive matching value is performed on the first matching value and the second matching value, and the ovarian state of the ovarian image to be detected is determined according to the comprehensive matching value. The specific operation steps are as follows:

[0110] S281: Normalize the first matching value and the second matching value; weight the normalized first matching value and the second matching value, set the weight of the first matching value to w1, and the weight of the second matching value to w2, and calculate the comprehensive matching value C, the calculation formula is:

[0111] C = w1×first matching value+w2×second matching value;

[0112] The comprehensive matching value includes: a normal ovary comprehensive matching value, a suspected comprehensive matching value, a confirmed comprehensive matching value, a slight comprehensive matching value, and a severe comprehensive matching value (i.e., the explanations in S25 and S27 have explained that the first matching value and the second matching value both include multiple matching values, and therefore, a corresponding calculation of the comprehensive matching value is performed (e.g., the first matching value obtained by the normal ovary and the second matching value obtained by the normal ovary are used to obtain the normal ovary comprehensive matching value));

[0113] The reason why the comprehensive matching value needs to be calculated has been explained in the explanation of step S28. The first matching value and the second matching value are different in the process of obtaining them, so they need to be normalized and unified, and weighted to calculate the comprehensive matching value;

[0114] S282: preset a normal ovary threshold q, a suspected cystic ovary threshold w, a confirmed cystic ovary threshold e, and a mild cystic ovary threshold r; it should be noted that the above thresholds refer to the thresholds of the comprehensive matching value;

[0115] Determine whether the normal ovary comprehensive matching value is less than the normal ovary threshold q;

[0116] If yes, it is determined that the ovarian image to be detected is a normal ovary;

[0117] If not, determining whether the suspected comprehensive matching value is less than the suspected cystic ovarian threshold w;

[0118] If yes, it is determined that the ovarian image to be detected is a suspected cystic ovary;

[0119] If not, determining whether the determined comprehensive matching value is less than the determined cystic ovarian threshold e;

[0120] If yes, it is determined that the ovarian image to be detected is a confirmed cystic ovary;

[0121] If not, determining whether the mild comprehensive matching value is less than the mild cystic ovarian threshold r;

[0122] If yes, it is determined that the ovarian image to be detected is a mild cystic ovary;

[0123] If not, it is determined that the ovarian image to be detected is a severe cystic ovary;

[0124] It should be noted that four thresholds are set (i.e., normal ovary threshold q, suspected cystic ovary threshold w, confirmed cystic ovary threshold e, and mild cystic ovary threshold r); the comprehensive matching values ​​are compared one by one;

[0125] Whether the comprehensive matching value of the normal ovary is less than the normal ovary threshold value q. If it is less than the normal ovary threshold value q, it means that the ovarian image to be detected is a normal ovary. If it is greater than the normal ovary threshold value q, it means that there may be lesions in the ovarian cells in the ovarian image to be detected, and the lesion cells have exceeded the range of normal ovarian cells.

[0126] At this time, it is necessary to further judge whether the suspected comprehensive matching value is less than the suspected cystic ovary threshold w (that is, if the normal ovary comprehensive matching value is less than the normal ovary threshold q, there is no need to judge whether the suspected comprehensive matching value is less than the suspected cystic ovary threshold w, because when it is less than the normal ovary threshold q, the ovarian image to be detected is already a normal ovarian image). If it is less than the suspected cystic ovary threshold w, it means that there may be lesions in the ovarian cells in the ovarian image to be detected; if it is greater than the suspected cystic ovary threshold w, it means that the probability of ovarian cells in the ovarian image to be detected is greater; at this time, it is necessary to judge whether the determined comprehensive matching value is less than the determined cystic ovary threshold e (that is, because the suspected comprehensive matching value has exceeded the range of the suspected cystic ovary threshold w, the suspected comprehensive matching value can no longer meet the judgment In order to determine the change state of ovarian cells in the ovarian image to be detected, it is necessary to further determine whether the comprehensive matching value is less than the cystic ovary threshold e), until it is determined whether the slight comprehensive matching value is less than the slight cystic ovary threshold r; if it is less than the slight cystic ovary threshold r, it means that there are slight pathological changes in the ovarian cells in the ovarian image to be detected, and when it is greater than the slight cystic ovary threshold r, the ovarian image to be detected will reach the most serious pathological state of the ovarian cells in the final stage, so there is no need to set the severe cystic ovary threshold; because even if the severe cystic ovary threshold is set, whether the severe comprehensive matching value is greater than or less than or equal to the severe cystic ovary threshold, the ovarian image to be detected will be in the severe pathological state of the final stage, and no more serious results than the severe cystic ovary state will occur.

[0127] Further research has found that the process of expanding and filling the ovarian contour to be detected may introduce some new technical defects, including the following common problems:

[0128] For example, over-expansion: The process of expansion and filling will expand the contour area outward, which may sometimes lead to over-expansion. This may enlarge the ovarian area, distort the boundary or overlap with surrounding structures, thus interfering with the subsequent ovarian status analysis.

[0129] For example, the border is blurred: during the expansion and filling process, the border of the contour may become blurred. This is because the expansion operation blurs the edge information, reducing the clarity of the border, thus affecting the precise positioning and morphological analysis of the ovarian contour.

[0130] Another example is morphological changes: expansion and filling may change the morphological characteristics of the ovarian contour, including area, perimeter, shape, etc. This morphological change may have a negative impact on subsequent feature extraction, matching, and classification tasks because there is a difference between the original contour and the contour after expansion and filling.

[0131] However, the study found that if a large area of ​​excessive expansion and filling occurs, it will have the greatest impact on subsequent analysis and judgment; therefore, it is necessary to further propose new solutions to this specific technical defect.

[0132] Specifically, Figure 7 As shown, in step S26, the ovarian contour to be detected is expanded and filled to obtain a contour image to be detected, which specifically includes the following operations:

[0133] S261, firstly preset initial expansion and filling parameters; the expansion and filling parameters include the size and shape of the structure element;

[0134] S262, directly calling the initial expansion and filling parameters; directly expanding and filling the current ovarian contour to be detected using the current initial expansion and filling parameters to obtain the initial contour image to be detected; detecting whether the initial contour image to be detected has excessive expansion and filling, and if so, adjusting the expansion and filling parameters, continuing to use multiple iterations to gradually expand and fill the current ovarian contour to be detected, and finally outputting the contour image to be detected.

[0135] Specifically, Figure 8 As shown, in step S262, the expansion and filling parameters are adjusted, and the expansion and filling of the current ovarian contour to be detected is gradually performed using multiple iterations, and finally the contour image to be detected is output, which specifically includes the following steps:

[0136] S2621, reducing the size of the structure element in the current initial expansion filling parameter by equal proportion T times to obtain the target value of the structure element required for each iteration;

[0137] S2622, first call the maximum value among the target values ​​of several structural elements, and use it as the parameter of the first iteration of expansion and filling; then perform the first expansion and filling on the current ovarian contour to be detected to obtain the contour image to be detected of the first iteration; determine whether the contour image to be detected of the current first iteration has excessive expansion and filling, if there is still excessive expansion and filling; continue to select the second largest value among the target values ​​of several structural elements, continue to perform the second expansion and filling to obtain the contour image to be detected of the second iteration; at the same time, continuously iteratively detect whether the contour image to be detected obtained in the current iteration has excessive expansion and filling, until the current contour image to be detected is determined to have no excessive expansion and filling, and determine the contour image to be detected obtained by the last iteration filling as the final contour image to be detected.

[0138] It should be noted that the parameters of the expansion filling include the size and shape of the structural element. When adjusting the parameters of the expansion filling, the size of the structural element is mainly adjusted;

[0139] Here is an example showing how to use the scipy library in Python to perform dilation and adjust the parameters:

[0140] from scipy.ndimage import binary_dilationimport numpy as np#define the structural element structure=np.ones((50,50),dtype=bool)#rectangular structural element#dilate and fill dilated_image=binary_dilation(image,structure=structure,iterations=1);

[0141] In this example, we use the binary_dilation function in the SciPy library to perform the dilation operation. The parameter image is the binary image to be processed, and structure is the structure element, which determines the shape and size of the dilation. Here, we use a 50x50 rectangular structure element (that is, the size of the structure element is 50x50, and the shape of the structure element is rectangular). By adjusting the size of the structure element, the degree of dilation can be controlled. It should be noted here that the shape of the structure element does not need to be changed when adjusting proportionally, but mainly to modify

[0142] That is, multiple step-by-step iterations are used: expansion and filling are performed through multiple iterations, and each iteration uses a smaller structural element, that is, the structural element is reduced a little to perform multiple downgraded expansion and filling. In this way, the filling can be performed step by step to reduce the possibility of over-expansion. After each iteration, the filling results can be checked and further downgraded iterations can be performed as needed. In summary, the expansion and filling process may introduce new technical defects, but through appropriate parameter settings, the impact of these defects on ovarian status analysis can be reduced, and the accuracy and reliability of the results can be improved. Dynamic control of the size of the expanded structural element can achieve control of the expansion degree and the filling effect, thereby reducing the impact of over-expansion and morphological changes.

[0143] Embodiment 2

[0144] like Fig. 9 As shown, accordingly, the present invention also proposes a tumor cell screening detection processing system, including: an image acquisition module 10, a matching module 20, and a final result module 30;

[0145] The image acquisition module 10 is used to acquire the ovarian image to be detected and acquire the historical ovarian image set;

[0146] The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images;

[0147] The matching degree module 20 is used to select feature images of the historical ovarian image set and the ovarian image to be detected respectively, obtain the historical target image and the target frame image to be detected, and compare them to obtain a first matching degree value; extract the contour of the ovarian image to be detected to obtain the ovarian contour image to be detected; compare and match the historical target image with the detected ovarian contour image to obtain a second matching degree value;

[0148] The final result module 30 is used to compare the first matching value with the second matching value to obtain the ovarian status of the ovarian image to be detected.

[0149] like Fig.10 As shown, on the other hand, this embodiment 3 is based on a tumor cell screening detection and processing method provided in the embodiment 1 of the invention, and further provides a computer storage medium 1140 (hereinafter referred to as storage medium). It is a schematic diagram of the computer storage medium structure framework provided in the embodiment 3 of the present invention, which includes:

[0150] A memory 1130, used for storing computer programs;

[0151] The communication interface 1120 is used to connect the memory 1130 to the processor 1110;

[0152] The processor 1110 is used to execute a computer program to implement a tumor cell screening, detection and processing method involved in Example 1 disclosed in combination with any of the above-mentioned implementation modes.

[0153] In summary, a tumor cell screening detection processing method, system and storage medium proposed in an example of the present invention selects candidate frames for the ovarian image to be detected through the YOLO model, and each candidate frame will have a confidence score, and the interval score is divided by the confidence score, and the interval level is set by the interval score. Because, in the ovarian image to be detected, the candidate frames with confidence scores will be in various corners of the ovarian image to be detected, so different confidence scores will appear. The ovarian cell states at various positions in the ovarian image to be detected can be judged by different confidence scores, so that the candidate frames with confidence scores are divided by interval levels, and the candidate frames with average confidence scores are selected as representative candidate frames of the score intervals, and at the same time as the cell states of each score interval; and the candidate frames are selected for the historical ovarian image set to obtain the historical target image, and finally matched to obtain the first matching value;

[0154] Further, the contour of the ovarian image to be detected is extracted by an edge detection algorithm, and the contour is expanded and filled with pixels to obtain the contour image to be detected. By filling the contour with pixels, the contour image to be detected can be matched with the historical target image, so that the incompatibility when matching the contour and the image can be avoided, and the matching can be smoothly performed to obtain the second matching value;

[0155] Furthermore, the processes for obtaining the first matching value and the second matching value are different, so the first matching value and the second matching value cannot be matched directly. Therefore, it is necessary to calculate the comprehensive matching value through a weighted method, and judge whether the ovary is a normal ovary, a suspected cystic ovary, a confirmed cystic ovary, a mild cystic ovary, or a severe cystic ovary through the comprehensive matching value and the preset ovarian cyst threshold, so as to carry out medical treatment for different ovarian conditions.

[0156] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace part or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for screening, detecting and processing tumor cells, characterized in that: The steps include: Obtain an ovarian image to be detected and obtain a historical ovarian image set; The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images; Select feature images of the historical ovarian image set and the ovarian image to be detected respectively to obtain a historical target image and a target frame image to be detected, and compare them to obtain a first matching value; extract the contour of the ovarian image to be detected to obtain an ovarian contour image to be detected, compare and match the historical target image with the detected ovarian contour image to obtain a second matching value; compare the first matching value with the second matching value to obtain the ovarian state of the ovarian image to be detected; The historical ovarian image set and the ovarian image to be detected are respectively framed to select feature images, to obtain the historical target image and the target frame image to be detected, and then compared to obtain a first matching value. The specific operation steps are as follows: Obtain a preset YOLO model; select a candidate frame for the ovarian image to be detected according to the YOLO model to obtain multiple candidate frames to be detected, each of which contains different category labels; annotate the multiple candidate frames to be detected with confidence scores, and divide the confidence score of each candidate frame to be detected into different score intervals; Dividing the interval levels according to the different category labels and the different score intervals; Calculate the average value based on the confidence scores of all candidate boxes to be detected within each interval score; The average of the confidence scores of all candidate boxes to be detected within each interval score corresponds to the interval level; Select the candidate frames to be detected that are close to the average value of the confidence scores within each interval score, determine them as the target frames to be detected within each interval score, and crop the target frames to be detected to obtain the target images to be detected; Performing candidate box selection on each historical ovarian image in the historical ovarian image set by using the YOLO model, obtaining multiple historical ovarian candidate boxes for each historical ovarian image, and marking confidence scores for the historical ovarian candidate boxes; selecting the historical ovarian candidate box with the highest confidence score as the historical target image; The historical target images include: historical normal ovarian target images, historical suspected cystic ovarian images, historical confirmed cystic ovarian images, historical mild cystic ovarian images, and historical severe cystic ovarian images; Comparing each historical image in the historical target images with the target frame image to be detected within each interval score to obtain a first matching value; The interval grades are divided into normal ovarian interval grade, suspected cystic ovarian interval grade, confirmed cystic ovarian interval grade, mild cystic ovarian interval grade, and severe cystic ovarian interval grade; The contour of the ovarian image to be detected is extracted to obtain the ovarian contour image to be detected, and the historical target image is compared and matched with the detected ovarian contour image to obtain a second matching value. The specific operation steps are as follows: Extracting contour features of the ovarian image to be detected by edge detection algorithm to obtain the ovarian contour to be detected; expanding and filling the ovarian contour to be detected to obtain the contour image to be detected; The contour image to be detected is compared with each historical image of the historical target image to obtain a second matching value; The first matching value is compared with the second matching value to obtain the ovarian state of the ovarian image to be detected, specifically including: Performing a weighted evaluation calculation on the first matching value and the second matching value to obtain a comprehensive matching value; The ovarian status of the ovarian image to be detected is determined according to the comprehensive matching value.

2. A tumor cell screening, detection and processing method according to claim 1, characterized in that: The first matching value and the second matching value are weightedly evaluated and calculated to obtain a comprehensive matching value. The specific operation steps are as follows: Normalize the first matching value and the second matching value; weight the normalized first matching value and the second matching value, set the weight of the first matching value to w1, and the weight of the second matching value to w2, and calculate the comprehensive matching value C, the calculation formula is: C=w1×first matching value+w2×second matching value; The comprehensive matching values ​​include: normal ovary comprehensive matching value, suspected comprehensive matching value, confirmed comprehensive matching value, slight comprehensive matching value, and severe comprehensive matching value.

3. A tumor cell screening, detection and processing method according to claim 2, characterized in that: The ovarian status of the ovarian image to be detected is determined according to the comprehensive matching value, and the specific operation steps are as follows: The normal ovary threshold q, the suspected cystic ovary threshold w, the confirmed cystic ovary threshold e, and the mild cystic ovary threshold r are preset; Determine whether the normal ovary comprehensive matching value is less than the normal ovary threshold q; If yes, it is determined that the ovarian image to be detected is a normal ovary; If not, determining whether the suspected comprehensive matching value is less than the suspected cystic ovarian threshold w; If yes, it is determined that the ovarian image to be detected is a suspected cystic ovary; If not, determining whether the determined comprehensive matching value is less than the determined cystic ovarian threshold e; If yes, it is determined that the ovarian image to be detected is a confirmed cystic ovary; If not, determining whether the mild comprehensive matching value is less than the mild cystic ovarian threshold r; If yes, it is determined that the ovarian image to be detected is a mild cystic ovary; If not, it is determined that the ovarian image to be detected is a severe cystic ovary.

4. A tumor cell screening, detection and processing method according to claim 3, characterized in that: The ovarian contour to be detected is expanded and filled to obtain a contour image to be detected, which specifically includes the following operations: Presetting initial expansion and filling parameters; the expansion and filling parameters include the size and shape of the structure element; Directly call the initial expansion and filling parameters; directly expand and fill the current ovarian contour to be detected using the current initial expansion and filling parameters to obtain the initial contour image to be detected; detect whether the initial contour image to be detected has excessive expansion and filling, and if so, adjust the expansion and filling parameters, continue to use multiple iterations to gradually expand and fill the current ovarian contour to be detected, and finally output the contour image to be detected.

5. A tumor cell screening, detection and processing method according to claim 4, characterized in that: Adjust the expansion and filling parameters, continue to use multiple iterations to gradually expand and fill the current ovarian contour to be detected, and finally output the contour image to be detected, which specifically includes the following steps: The size of the structural element in the current initial expansion filling parameter is proportionally reduced T times to obtain the target value of the structural element required for each iteration; First, the maximum value among the target values ​​of several structural elements is called and used as the parameter of the first iteration of expansion and filling; then the first expansion and filling is performed on the current ovarian contour to be detected to obtain the contour image to be detected of the first iteration; it is determined whether the contour image to be detected of the current first iteration has excessive expansion and filling, and if it still has excessive expansion and filling; the second largest value among the target values ​​of several structural elements is continued to be selected, and the second expansion and filling is continued to be obtained to obtain the contour image to be detected of the second iteration; at the same time, it is continuously iterated to detect whether the contour image to be detected obtained by the current iteration has excessive expansion and filling, until the current contour image to be detected is determined to have no excessive expansion and filling, and the contour image to be detected obtained by the last iteration filling is determined as the final contour image to be detected.

6. A tumor cell screening, detection and processing system, characterized in that: Used to implement the tumor cell screening, detection and processing method according to any one of claims 1 to 5, the system comprises: an image acquisition module, a matching module, and a final result module; The image acquisition module is used to acquire the ovarian image to be detected and acquire the historical ovarian image set; The historical ovarian image set includes: historical normal ovarian images, historical suspected ovarian cyst images, historical confirmed ovarian cyst images, historical mild ovarian cyst images, and historical severe ovarian cyst images; The matching degree module is used to respectively select feature images of the historical ovarian image set and the ovarian image to be detected, obtain the historical target image and the target frame image to be detected, and compare them to obtain a first matching degree value; extract the contour of the ovarian image to be detected to obtain the ovarian contour image to be detected; compare and match the historical target image with the detected ovarian contour image to obtain a second matching degree value; The final result module is used to compare the first matching value with the second matching value to obtain the ovarian status of the ovarian image to be detected.

7. A storage medium, characterized in that: It includes: Memory for storing computer programs; A communication interface, used to connect the memory to the processor; A processor is used to execute a computer program to implement the steps of the tumor cell screening, detection and processing method as described in any one of claims 1 to 5.

Images