Tissue region image recognition method and system based on block division

Abstract

The present invention relates to the field of data processing technology, and discloses a tissue region image recognition method and system based on block division, the method comprising: performing block division on a biological tissue image to be identified to obtain a plurality of image blocks; inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types; screening out at least one region combination from the plurality of tissue regions according to a preset type bordering rule; the region combination includes a plurality of mutually bordering tissue regions; inputting the region image corresponding to the region combination and the corresponding region type into a region boundary update model to obtain a corresponding region update result; the region update result is used to re-indicate the region boundaries of different tissue regions in the region image. It can be seen that the present invention can improve the accuracy and efficiency of tissue region recognition and provide an accurate data basis for disease treatment or research.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a tissue region image recognition method and system based on block division. Background Art

[0002] With the development of image processing technology and the popularization of medical auxiliary imaging technology, more and more biological tissue images are used for disease research or treatment purposes. How to effectively identify different types of tissue regions in biological tissue images has become a core technical problem. However, the existing tissue region image recognition technology generally only uses the algorithm model obtained by training based on annotated images to achieve region recognition, and does not further combine the regional connection features related to the region type to make more detailed regional edge determination. Therefore, its region recognition efficiency and accuracy are lacking. It can be seen that the existing technology has defects that need to be solved urgently. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a tissue region image recognition method and system based on block division, which can fully utilize the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research.

[0004] In order to solve the above technical problems, the first aspect of the present invention discloses a tissue region image recognition method based on block division, the method comprising:

[0005] Dividing the biological tissue image to be identified into blocks to obtain a plurality of image blocks;

[0006] Inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types;

[0007] According to a preset type bordering rule, at least one region combination is selected from the multiple tissue regions; the region combination includes multiple tissue regions bordering each other;

[0008] The regional image corresponding to the regional combination and the corresponding regional type are input into the regional boundary update model to obtain the corresponding regional update result; the regional update result is used to re-indicate the regional boundaries of different tissue regions in the regional image.

[0009] As an optional implementation, in the first aspect of the present invention, the region recognition model is trained by the following steps:

[0010] Determine the pre-divided type region division annotation corresponding to the training tissue image;

[0011] Segmenting the training tissue image to obtain a plurality of training image blocks;

[0012] For each of the training image blocks, calculating the intersection area between the training image block and any type of division area in the type area division annotation;

[0013] Determine the region type corresponding to the type division region with the largest intersection area as the type label corresponding to the training image block;

[0014] Based on all the training image blocks and the corresponding type annotations, iteratively train the model of the preset CNN network architecture, and calculate the average value of the classification accuracy corresponding to each type annotation in each round of training to obtain a training performance parameter; the classification accuracy is an F1 score or an F2 score;

[0015] The weight of the model obtained in the round of training with the highest training performance parameter is determined as the final model weight to obtain the region recognition model.

[0016] As an optional implementation, in the first aspect of the present invention, the inputting the multiple image blocks into the region recognition model to obtain multiple tissue regions and corresponding region types includes:

[0017] Inputting each of the image blocks into a region recognition model to obtain a predicted region type corresponding to each of the image blocks;

[0018] All the image blocks are spliced ​​together based on the segmentation rule to obtain a spliced ​​tissue image;

[0019] Performing morphological operations on the spliced ​​tissue image to obtain a contour smoothed image; the contour smoothed image includes a plurality of tissue regions and corresponding region types; the tissue region includes a plurality of image blocks having the same corresponding predicted region type; the region type is the predicted region type corresponding to the image block in the tissue region;

[0020] For each of the tissue regions in the contour smoothed image, determining the image block corresponding to the tissue region whose predicted region type is different from the region type of the tissue region, and obtaining at least one abnormal image block;

[0021] Calculating a ratio of the number of all the abnormal image blocks to the total number of all the image blocks in the tissue region;

[0022] Determining whether the quantity ratio is less than a preset ratio threshold;

[0023] If so, modifying the predicted region types of all the abnormal image blocks to the region type of the tissue region;

[0024] If not, input each of the image blocks in the tissue region into the prediction model corresponding to the predicted region type of any of the abnormal image blocks to obtain the prediction probability of other types corresponding to each of the image blocks;

[0025] For each predicted region type of the abnormal image block, calculating an average value of other type prediction probabilities of all the image blocks in the tissue region corresponding to the predicted region type;

[0026] The predicted region type with the highest average value is determined as the region type of the tissue region.

[0027] As an optional implementation, in the first aspect of the present invention, the step of selecting at least one region combination from the plurality of tissue regions according to a preset type bordering rule comprises:

[0028] For any two tissue regions that have borders, calculating the difference between the regional images of the two tissue regions;

[0029] Calculating the bordering ratio of the region boundaries of the two tissue regions; the bordering ratio is the ratio of the length of the common region boundary of the two tissue regions to the total region boundary length of the two tissue regions;

[0030] Calculating the product of the difference and the bordering ratio to obtain the bordering parameters of the two tissue regions;

[0031] Based on the bordering parameters, the multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0032] As an optional embodiment, in the first aspect of the present invention, the multiple tissue regions are grouped according to a dynamic programming algorithm based on the bordering parameters to obtain at least one region combination, including:

[0033] Setting the objective function to minimize the number of all the region combinations and minimize the number of the tissue regions in the region combinations;

[0034] Setting restrictions includes:

[0035] The area types of any two tissue areas in the area combination meet a preset type condition;

[0036] The border parameter of any two tissue regions in the region combination is greater than a preset first parameter threshold;

[0037] The border parameter of any two tissue regions belonging to different region combinations is less than a preset second parameter threshold; the second parameter threshold is less than the first parameter threshold;

[0038] Based on the objective function and the restriction condition, the multiple tissue regions are iteratively grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0039] As an optional embodiment, in the first aspect of the present invention, the region type is a normal type, a necrotic type, a tumor type, an infiltration type or a background type; the region type of any two bordering tissue regions in the region combination satisfies a preset type condition; the type condition is used to limit the region types of the two bordering tissue regions to be different, and the region type combination of the two tissue regions is normal-necrosis, normal-tumor, necrosis-tumor, infiltration-tumor or infiltration-necrosis.

[0040] As an optional embodiment, in the first aspect of the present invention, the region boundary update model is obtained by training a training data set including multiple training region combination images and corresponding region combination type annotations and region boundary adjustment annotations; the region boundary adjustment annotations include original region boundaries and modified region boundaries.

[0041] As an optional implementation, in the first aspect of the present invention, the region boundary update model is trained by the following steps:

[0042] Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially;

[0043] The training data set is input into a preset CNN architecture model for iterative training, and parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

[0044] A second aspect of an embodiment of the present invention discloses a tissue region image recognition system based on block division, the system comprising:

[0045] A division module, used for dividing the biological tissue image to be identified into multiple image blocks;

[0046] A first recognition module, used for inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types;

[0047] A screening module, used for screening at least one area combination from the plurality of tissue areas according to a preset type bordering rule; the area combination includes a plurality of tissue areas bordering each other;

[0048] The second recognition module is used to input the regional image corresponding to the regional combination and the corresponding regional type into the regional boundary update model to obtain the corresponding regional update result; the regional update result is used to re-indicate the regional boundaries of different tissue regions in the regional image.

[0049] As an optional implementation, in the second aspect of the present invention, the region recognition model is trained by the following steps:

[0050] Determine the pre-divided type region division annotation corresponding to the training tissue image;

[0051] Segmenting the training tissue image to obtain a plurality of training image blocks;

[0052] For each of the training image blocks, calculating the intersection area between the training image block and any type of division area in the type area division annotation;

[0053] Determine the region type corresponding to the type division region with the largest intersection area as the type label corresponding to the training image block;

[0054] Based on all the training image blocks and the corresponding type annotations, iteratively train the model of the preset CNN network architecture, and calculate the average value of the classification accuracy corresponding to each type annotation in each round of training to obtain a training performance parameter; the classification accuracy is an F1 score or an F2 score;

[0055] The weight of the model obtained in the round of training with the highest training performance parameter is determined as the final model weight to obtain the region recognition model.

[0056] As an optional implementation, in the second aspect of the present invention, the first recognition module inputs the multiple image blocks into the region recognition model to obtain a specific manner of multiple tissue regions and corresponding region types, including:

[0057] Inputting each of the image blocks into a region recognition model to obtain a predicted region type corresponding to each of the image blocks;

[0058] All the image blocks are spliced ​​together based on the segmentation rule to obtain a spliced ​​tissue image;

[0059] Performing morphological operations on the spliced ​​tissue image to obtain a contour smoothed image; the contour smoothed image includes a plurality of tissue regions and corresponding region types; the tissue region includes a plurality of image blocks having the same corresponding predicted region type; the region type is the predicted region type corresponding to the image block in the tissue region;

[0060] For each of the tissue regions in the contour smoothed image, determining the image block corresponding to the tissue region whose predicted region type is different from the region type of the tissue region, and obtaining at least one abnormal image block;

[0061] Calculating a ratio of the number of all the abnormal image blocks to the total number of all the image blocks in the tissue region;

[0062] Determining whether the quantity ratio is less than a preset ratio threshold;

[0063] If so, modifying the predicted region types of all the abnormal image blocks to the region type of the tissue region;

[0064] If not, input each of the image blocks in the tissue region into the prediction model corresponding to the predicted region type of any of the abnormal image blocks to obtain the prediction probability of other types corresponding to each of the image blocks;

[0065] For each predicted region type of the abnormal image block, calculating an average value of other type prediction probabilities of all the image blocks in the tissue region corresponding to the predicted region type;

[0066] The predicted region type with the highest average value is determined as the region type of the tissue region.

[0067] As an optional implementation, in the second aspect of the present invention, the specific manner in which the screening module screens out at least one region combination from the multiple tissue regions according to a preset type bordering rule includes:

[0068] For any two tissue regions that have borders, calculating the difference between the regional images of the two tissue regions;

[0069] Calculating the bordering ratio of the region boundaries of the two tissue regions; the bordering ratio is the ratio of the length of the common region boundary of the two tissue regions to the total region boundary length of the two tissue regions;

[0070] Calculating the product of the difference and the bordering ratio to obtain the bordering parameters of the two tissue regions;

[0071] Based on the bordering parameters, the multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0072] As an optional implementation, in the second aspect of the present invention, the screening module groups the multiple tissue regions based on the bordering parameters according to a dynamic programming algorithm to obtain a specific method of at least one region combination, including:

[0073] Setting the objective function to minimize the number of all the region combinations and minimize the number of the tissue regions in the region combinations;

[0074] Setting restrictions includes:

[0075] The area types of any two tissue areas in the area combination meet a preset type condition;

[0076] The border parameter of any two tissue regions in the region combination is greater than a preset first parameter threshold;

[0077] The border parameter of any two tissue regions belonging to different region combinations is less than a preset second parameter threshold; the second parameter threshold is less than the first parameter threshold;

[0078] Based on the objective function and the restriction condition, the multiple tissue regions are iteratively grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0079] As an optional embodiment, in the second aspect of the present invention, the region type is a normal type, a necrotic type, a tumor type, an infiltration type or a background type; the region type of any two bordering tissue regions in the region combination satisfies a preset type condition; the type condition is used to limit the region types of the two bordering tissue regions to be different, and the region type combination of the two tissue regions is normal-necrosis, normal-tumor, necrosis-tumor, infiltration-tumor or infiltration-necrosis.

[0080] As an optional embodiment, in the second aspect of the present invention, the region boundary update model is obtained by training a training data set including multiple training region combination images and corresponding region combination type annotations and region boundary adjustment annotations; the region boundary adjustment annotations include original region boundaries and modified region boundaries.

[0081] As an optional implementation, in the second aspect of the present invention, the region boundary update model is trained by the following steps:

[0082] Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially;

[0083] The training data set is input into a preset CNN architecture model for iterative training, and parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

[0084] The third aspect of the present invention discloses another tissue region image recognition system based on block division, the system comprising:

[0085] A memory storing executable program code;

[0086] a processor coupled to the memory;

[0087] The processor calls the executable program code stored in the memory to execute part or all of the steps in the tissue region image recognition method based on block division disclosed in the first aspect of the present invention.

[0088] The fourth aspect of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the tissue region image recognition method based on block division disclosed in the first aspect of the present invention.

[0089] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0090] The present invention can first identify preliminary tissue regions and region types for the tissue image after dividing the blocks based on the region recognition model, and then after screening out the region combinations with suitable types, identify and update the region boundaries based on the bordering regions and types through a further region boundary update model, thereby being able to fully utilize the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0092] Figure 1 It is a flow chart of a tissue region image recognition method based on block division disclosed in an embodiment of the present invention.

[0093] Figure 2 It is a structural schematic diagram of a tissue region image recognition system based on block division disclosed in an embodiment of the present invention.

[0094] Figure 3 It is a structural schematic diagram of another tissue region image recognition system based on block division disclosed in an embodiment of the present invention.

[0095] Figure 4 It is a schematic diagram of image recognition results of a tissue region image recognition solution disclosed in an embodiment of the present invention.

[0096] Figure 5 It is an image recognition result of another tissue region image recognition solution disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0097] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. 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.

[0098] The terms "first", "second", etc. in the specification of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or equipment.

[0099] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0100] The present invention discloses a tissue region image recognition method and system based on block division, which can first identify preliminary tissue regions and region types based on the region recognition model of the tissue image after the block division, and then screen out the region combination with type adaptation, and then identify and update the region boundary based on the bordering region and type through a further region boundary update model, so as to make full use of the characteristics of the bordering region type to accurately identify the region boundary, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research. The following are detailed descriptions.

[0101] Embodiment 1

[0102] See also Figure 1 , Figure 1 is a flow chart of a tissue region image recognition method based on block division disclosed in an embodiment of the present invention. Figure 1 The described tissue region image recognition method based on tile division can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 1 As shown, the tissue region image recognition method based on block division may include the following operations:

[0103] 101. Divide the biological tissue image to be identified into blocks to obtain multiple image blocks.

[0104] 102. Input multiple image blocks into a region recognition model to obtain multiple tissue regions and corresponding region types.

[0105] 103. According to a preset type bordering rule, at least one region combination is selected from multiple organizational regions.

[0106] Optionally, the region combination includes a plurality of mutually adjacent tissue regions.

[0107] 104. Input the region image corresponding to the region combination and the corresponding region type into the region boundary update model to obtain the corresponding region update result.

[0108] Optionally, the region update result is used to re-indicate region boundaries of different tissue regions in the regional image.

[0109] It can be seen that the above-mentioned embodiments of the invention can first identify preliminary tissue regions and region types for the tissue image after dividing the blocks based on the region recognition model, and then after screening out the type-adapted region combinations, identify and update the region boundaries based on the bordering regions and types through a further region boundary update model, thereby making full use of the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research.

[0110] As an optional embodiment, in the above steps, the region recognition model is trained by the following steps:

[0111] Determine the pre-divided type region division annotation corresponding to the training tissue image;

[0112] Segmenting the training tissue image to obtain multiple training image blocks;

[0113] For each training image block, calculate the intersection area of ​​the training image block and any type of division area in the type area division annotation;

[0114] The region type corresponding to the type division region with the highest intersection area is determined as the type label corresponding to the training image block;

[0115] Based on all training image blocks and corresponding type annotations, the model of the preset CNN network architecture is iteratively trained, and the average classification accuracy corresponding to each type annotation is calculated in each round of training to obtain the training performance parameter; optionally, the classification accuracy is the F1 score or the F2 score;

[0116] The weight of the model obtained in the round of training with the highest training performance parameter is determined as the final model weight to obtain the region recognition model.

[0117] In a specific implementation, a region recognition model is implemented, and its purpose is to find the tumor region in the tissue and the infiltration region around the tumor for subsequent further boundary update. Specifically, the objects that the region recognition model needs to recognize are the tumor region, the normal epithelial region and the necrotic region. Among them, the normal epithelial and necrotic regions are mainly used to distinguish from the tumor region and the infiltration region around the tumor.

[0118] Specifically, the digital pathology slides of the whole slide were randomly divided into a training set, a validation set, and a test set, of which the training set consisted of 414 slides, the validation set consisted of 60 slides, and the test set consisted of 250 slides. Each digital pathology slide had the region outlines manually annotated by the pathologist, with the annotated categories including “normal epithelium”, “tumor”, and “necrosis”, as well as the corresponding machine-recognized regions, such as Figure 4 As shown, the red outline is the tumor area recognized by the machine, and the yellow area is the normal epithelial area recognized by the machine.

[0119] Specifically, due to the huge size of digital pathology slices, it is impossible to put the entire image into the GPU for training. Therefore, the image segmentation training method is adopted, that is, the large image is traversed with a 512x512 window and the entire large image is segmented into small tiles of 512*512 size. If the non-tissue area pixels in a small tile ( Figure 4 If the proportion of black pixels in the image block to the total number of pixels in the image block is less than 5%, the image block is retained to ensure that the small image block finally obtained contains sufficient tissue area. A total of 1.46 million and 670,000 image blocks were collected from the training set and the validation set, respectively. Specifically, the category of the small image block is determined based on its intersection with the contour marked by the doctor, and the determination method has been disclosed in the above embodiment.

[0120] Specifically, the region recognition model uses the EfficientNetV2 structure, and the classification model is trained based on the small tiles generated above. The classification categories are four: tumor, normal epithelium, necrosis, and background. Optionally, various classification model architectures can be used here, but the classic EfficientNetV2 structure is selected based on the performance and prediction efficiency of the comprehensive model.

[0121] Specifically, the input image is first preprocessed, including: channel extraction, extracting only the DAPI channel and PanCK+ channel from the original four-channel image, and adding an additional channel with all zeros, so that the final input image becomes three-channel, so that the pre-trained weights of ImageNet can be loaded; data compression, compressing the original image data of each channel to 8 bits, thereby improving the efficiency of training and prediction.

[0122] Specifically, during the training process, each round of training was randomly iterated 7,000 times, and the validation was randomly iterated 3,000 times, for a total of 400 rounds of training. Since the number of tiles in the four categories was not balanced, class balanced sampling was used in each round of training to ensure that the tile data of the four categories would not differ too much. At the same time, random data enhancement was used in the training process, and there was a certain probability of rotating, flipping, scaling, color changing, etc. the tiles, thereby increasing the generalization ability of the model. After each round of training, the mean F1 score of the three categories of tumor, normal epithelium, and necrosis on the validation set was calculated, and the weight of the round with the largest mean was selected as the optimal weight of the final classification model.

[0123] It can be seen that through the above optional embodiments, the entire training tissue image can be divided and labeled based on the judgment of the doctor or the intervention of the technician, and then the type labeling can be automatically determined based on the intersection area of ​​the divided image block and the different divided regions, so as to simplify the labeling procedure on the one hand and ensure the labeling accuracy on the other hand. The CNN network is then trained using the segmented training tissue image and type labeling, and the model weights are selected based on the training performance related to the classification accuracy to train a more accurate regional recognition model, which helps to make full use of the characteristics of the bordering area types to accurately identify the regional boundaries, improve the accuracy and efficiency of tissue area recognition, and provide an accurate data basis for disease treatment or research.

[0124] As an optional embodiment, in the above step, multiple image blocks are input into the region recognition model to obtain multiple tissue regions and corresponding region types, including:

[0125] Input each image block into the region recognition model to obtain the predicted region type corresponding to each image block;

[0126] All image blocks are spliced ​​based on the segmentation rule to obtain a spliced ​​tissue image;

[0127] Performing morphological operations on the spliced ​​tissue image to obtain a contour smoothed image; optionally, the contour smoothed image includes a plurality of tissue regions and corresponding region types; the tissue region includes a plurality of image blocks with the same corresponding predicted region type; the region type is the predicted region type corresponding to the image block in the tissue region;

[0128] For each tissue region in the contour smoothed image, determining an image block in the tissue region whose corresponding predicted region type is different from the region type of the tissue region, and obtaining at least one abnormal image block;

[0129] Calculating the ratio of the number of all abnormal image blocks to the total number of all image blocks in the tissue region;

[0130] Determine whether the quantity ratio is less than a preset ratio threshold;

[0131] If so, modify the predicted region types of all abnormal image blocks to the region type of the tissue region;

[0132] If not, each image block in the tissue region is input into the prediction model corresponding to the prediction region type of any abnormal image block to obtain the prediction probability of other types corresponding to each image block;

[0133] For each predicted region type of an abnormal image block, calculating an average of other types of predicted probabilities of all image blocks in the tissue region corresponding to the predicted region type;

[0134] The predicted region type with the highest average value is determined as the region type of the tissue region.

[0135] Specifically, in the above implementation scheme, the scheme in this embodiment is implemented. Specifically, when the region recognition model is predicted, a 512×512 block is also used to traverse the entire image organization range, and the classification result of each block is obtained. The classified blocks are spliced ​​and restored to the large image, and morphological operations are performed, such as dilation first and then corrosion, so that the contours of different categories of regions are smoother.

[0136] Specifically, the solution based on the above embodiment can be achieved: if there are tumor areas in the normal tissue area, these tumor areas will be changed into normal tissue areas, thereby excluding the tumor areas misdetected in the normal area; if there are normal epithelial areas in the tumor area, these normal epithelial areas will be changed into tumor areas, thereby excluding the normal epithelial areas misdetected in the tumor area.

[0137] It can be seen that through the above-mentioned optional embodiments, it is possible to first predict the image blocks based on the region recognition model, and then re-splice and judge the regions of the image blocks to achieve more accurate preliminary region type recognition, assist in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research.

[0138] As an optional embodiment, in the above step, selecting at least one region combination from multiple tissue regions according to a preset type bordering rule includes:

[0139] For any two tissue regions that have borders, the difference between the regional images of the two tissue regions is calculated;

[0140] Calculating the bordering ratio of the region boundaries of the two tissue regions; optionally, the bordering ratio is the ratio of the length of the common region boundary of the two tissue regions to the total region boundary length of the two tissue regions;

[0141] Calculate the product of the difference and the bordering ratio to obtain the bordering parameters of the two tissue regions;

[0142] Based on the bordering parameters, multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0143] It can be seen that through the above-mentioned optional embodiments, the degree of bordering sufficiency can be obtained by calculating the image difference between the bordering tissue regions and the bordering boundary ratio, and then divided based on the dynamic programming algorithm to screen out more differentiated region combinations, which is convenient for subsequent region boundary identification, and assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0144] As an optional embodiment, in the above steps, based on the bordering parameters, the multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination, including:

[0145] The objective function is set to minimize the number of all regional combinations and minimize the number of organizational regions in the regional combinations;

[0146] Setting restrictions includes:

[0147] The regional types of any two organizational regions in the regional combination meet the preset type conditions;

[0148] The bordering parameter of any two tissue regions in the region combination is greater than a preset first parameter threshold;

[0149] The bordering parameter of any two tissue regions belonging to different region combinations is less than a preset second parameter threshold; optionally, the second parameter threshold is less than the first parameter threshold;

[0150] Based on the objective function and the restriction conditions, multiple tissue regions are iteratively grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0151] It can be seen that through the above-mentioned optional embodiments, it is possible to perform iterative grouping based on the preset objective function and constraints through a dynamic programming algorithm to screen for more differentiated and accurate region combinations, which is convenient for subsequent region boundary identification, and assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, thereby improving the accuracy and efficiency of tissue region identification and providing an accurate data basis for disease treatment or research.

[0152] As an optional embodiment, in the above steps, the region type is a normal type, a necrosis type, a tumor type, an infiltration type or a background type; the region type of any two bordering tissue regions in the region combination satisfies a preset type condition; the type condition is used to limit the region types of the two bordering tissue regions to be different, and the region type combination of the two tissue regions is normal-necrosis, normal-tumor, necrosis-tumor, infiltration-tumor or infiltration-necrosis.

[0153] It can be seen that through the above optional embodiments, the type conditions of the tissue regions in the region combination are defined, which facilitates the subsequent identification of region boundaries, assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improves the accuracy and efficiency of tissue region identification, and provides an accurate data basis for disease treatment or research.

[0154] As an optional embodiment, in the above steps, the region boundary update model is trained by a training data set including multiple training region combination images and corresponding region combination type annotations and region boundary adjustment annotations; the region boundary adjustment annotations include original region boundaries and modified region boundaries.

[0155] It can be seen that through the above-mentioned optional embodiments, the training data set details of the region boundary update model are limited, so that after training, it can be effectively used to refine the region boundaries of specific types, fully utilize the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0156] As an optional embodiment, in the above steps, the region boundary update model is trained by the following steps:

[0157] Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; optionally, the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially;

[0158] The training data set is input into the preset CNN architecture model for iterative training, and the parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

[0159] Specifically, an example of identifying a trained region boundary update model can be found in Figure 5 , where the left side is the region boundary before updating, and the right side is the region boundary after updating (the green outline is the tumor area, the purple outline is the normal epithelial area, and the pink outline is the necrotic area).

[0160] In an experiment, for each tissue region finally generated, the DICE indicator was used on the whole image to evaluate its consistency with the manual annotation. It was seen that the DICE values ​​of the main target tumor region in the training, validation and test sets were all greater than 0.8, proving the good generalization ability of the model.

[0161] It can be seen that through the above-mentioned optional embodiments, the priorities of different boundary annotations when updating the boundaries can be effectively adjusted based on the preset loss function, so as to train a regional boundary update model that can more accurately update the boundaries of tissue regions, make full use of the characteristics of the bordering region types to accurately identify regional boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0162] Embodiment 2

[0163] See also Figure 2 , Figure 2 : is a schematic diagram of a tissue region image recognition system based on block division disclosed in an embodiment of the present invention. Figure 2 The tissue region image recognition system based on tile division described above can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 2 As shown, the tissue region image recognition system based on block division may include:

[0164] The division module 201 is used to divide the biological tissue image to be identified into multiple image blocks.

[0165] The first recognition module 202 is used to input multiple image blocks into a region recognition model to obtain multiple tissue regions and corresponding region types.

[0166] The screening module 203 is used to screen out at least one region combination from multiple tissue regions according to a preset type bordering rule.

[0167] Optionally, the region combination includes a plurality of mutually adjacent tissue regions.

[0168] The second recognition module 204 is used to input the region image corresponding to the region combination and the corresponding region type into the region boundary update model to obtain the corresponding region update result.

[0169] Optionally, the region update result is used to re-indicate region boundaries of different tissue regions in the regional image.

[0170] It can be seen that the above-mentioned embodiments of the invention can first identify preliminary tissue regions and region types for the tissue image after dividing the blocks based on the region recognition model, and then after screening out the type-adapted region combinations, identify and update the region boundaries based on the bordering regions and types through a further region boundary update model, thereby making full use of the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research.

[0171] As an optional embodiment, the region recognition model is trained by the following steps:

[0172] Determine the pre-divided type region division annotation corresponding to the training tissue image;

[0173] Segmenting the training tissue image to obtain multiple training image blocks;

[0174] For each training image block, calculate the intersection area of ​​the training image block and any type of division area in the type area division annotation;

[0175] The region type corresponding to the type division region with the highest intersection area is determined as the type label corresponding to the training image block;

[0176] Based on all training image blocks and corresponding type annotations, the model of the preset CNN network architecture is iteratively trained, and the average classification accuracy corresponding to each type annotation is calculated in each round of training to obtain the training performance parameter; optionally, the classification accuracy is the F1 score or the F2 score;

[0177] The weight of the model obtained in the round of training with the highest training performance parameter is determined as the final model weight to obtain the region recognition model.

[0178] It can be seen that through the above optional embodiments, the entire training tissue image can be divided and labeled based on the judgment of the doctor or the intervention of the technician, and then the type labeling can be automatically determined based on the intersection area of ​​the divided image block and the different divided regions, so as to simplify the labeling procedure on the one hand and ensure the labeling accuracy on the other hand. The CNN network is then trained using the segmented training tissue image and type labeling, and the model weights are selected based on the training performance related to the classification accuracy to train a more accurate regional recognition model, which helps to make full use of the characteristics of the bordering area types to accurately identify the regional boundaries, improve the accuracy and efficiency of tissue area recognition, and provide an accurate data basis for disease treatment or research.

[0179] As an optional embodiment, the first recognition module inputs multiple image blocks into the region recognition model to obtain multiple tissue regions and corresponding region types in a specific manner, including:

[0180] Input each image block into the region recognition model to obtain the predicted region type corresponding to each image block;

[0181] All image blocks are spliced ​​based on the segmentation rule to obtain a spliced ​​tissue image;

[0182] Performing morphological operations on the spliced ​​tissue image to obtain a contour smoothed image; optionally, the contour smoothed image includes a plurality of tissue regions and corresponding region types; the tissue region includes a plurality of image blocks with the same corresponding predicted region type; the region type is the predicted region type corresponding to the image block in the tissue region;

[0183] For each tissue region in the contour smoothed image, determining an image block in the tissue region whose corresponding predicted region type is different from the region type of the tissue region, and obtaining at least one abnormal image block;

[0184] Calculating the ratio of the number of all abnormal image blocks to the total number of all image blocks in the tissue region;

[0185] Determine whether the quantity ratio is less than a preset ratio threshold;

[0186] If so, modify the predicted region types of all abnormal image blocks to the region type of the tissue region;

[0187] If not, each image block in the tissue region is input into the prediction model corresponding to the prediction region type of any abnormal image block to obtain the prediction probability of other types corresponding to each image block;

[0188] For each predicted region type of an abnormal image block, calculating an average of other types of predicted probabilities of all image blocks in the tissue region corresponding to the predicted region type;

[0189] The predicted region type with the highest average value is determined as the region type of the tissue region.

[0190] It can be seen that through the above-mentioned optional embodiments, it is possible to first predict the image blocks based on the region recognition model, and then re-splice and judge the regions of the image blocks to achieve more accurate preliminary region type recognition, assist in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improve the accuracy and efficiency of tissue region recognition, and provide an accurate data basis for disease treatment or research.

[0191] As an optional embodiment, the specific manner in which the screening module screens out at least one region combination from a plurality of tissue regions according to a preset type bordering rule includes:

[0192] For any two tissue regions that have borders, the difference between the regional images of the two tissue regions is calculated;

[0193] Calculating the bordering ratio of the region boundaries of the two tissue regions; optionally, the bordering ratio is the ratio of the length of the common region boundary of the two tissue regions to the total region boundary length of the two tissue regions;

[0194] Calculate the product of the difference and the bordering ratio to obtain the bordering parameters of the two tissue regions;

[0195] Based on the bordering parameters, multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0196] It can be seen that through the above-mentioned optional embodiments, the degree of bordering sufficiency can be obtained by calculating the image difference between the bordering tissue regions and the bordering boundary ratio, and then divided based on the dynamic programming algorithm to screen out more differentiated region combinations, which is convenient for subsequent region boundary identification, and assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0197] As an optional embodiment, the screening module groups the multiple tissue regions based on the bordering parameters according to the dynamic programming algorithm to obtain a specific method of at least one region combination, including:

[0198] The objective function is set to minimize the number of all regional combinations and minimize the number of organizational regions in the regional combinations;

[0199] Setting restrictions includes:

[0200] The regional types of any two organizational regions in the regional combination meet the preset type conditions;

[0201] The bordering parameter of any two tissue regions in the region combination is greater than a preset first parameter threshold;

[0202] The bordering parameter of any two tissue regions belonging to different region combinations is less than a preset second parameter threshold; optionally, the second parameter threshold is less than the first parameter threshold;

[0203] Based on the objective function and the restriction conditions, multiple tissue regions are iteratively grouped according to a dynamic programming algorithm to obtain at least one region combination.

[0204] It can be seen that through the above-mentioned optional embodiments, it is possible to perform iterative grouping based on the preset objective function and constraints through a dynamic programming algorithm to screen for more differentiated and accurate region combinations, which is convenient for subsequent region boundary identification, and assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, thereby improving the accuracy and efficiency of tissue region identification and providing an accurate data basis for disease treatment or research.

[0205] As an optional embodiment, the region type is a normal type, a necrosis type, a tumor type, an infiltration type or a background type; the region type of any two bordering tissue regions in the region combination satisfies a preset type condition; the type condition is used to limit the region types of the two bordering tissue regions to be different, and the region type combination of the two tissue regions is normal-necrosis, normal-tumor, necrosis-tumor, infiltration-tumor or infiltration-necrosis.

[0206] It can be seen that through the above optional embodiments, the type conditions of the tissue regions in the region combination are defined, which facilitates the subsequent identification of region boundaries, assists in making full use of the characteristics of the bordering region types to accurately identify region boundaries, improves the accuracy and efficiency of tissue region identification, and provides an accurate data basis for disease treatment or research.

[0207] As an optional embodiment, the region boundary update model is trained by a training data set including multiple training region combination images and corresponding region combination type annotations and region boundary adjustment annotations; the region boundary adjustment annotations include original region boundaries and modified region boundaries.

[0208] It can be seen that through the above-mentioned optional embodiments, the training data set details of the region boundary update model are limited, so that after training, it can be effectively used to refine the region boundaries of specific types, fully utilize the characteristics of the bordering region types to accurately identify the region boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0209] As an optional embodiment, the region boundary update model is trained by the following steps:

[0210] Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; optionally, the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially;

[0211] The training data set is input into the preset CNN architecture model for iterative training, and the parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

[0212] It can be seen that through the above-mentioned optional embodiments, the priorities of different boundary annotations when updating the boundaries can be effectively adjusted based on the preset loss function, so as to train a regional boundary update model that can more accurately update the boundaries of tissue regions, make full use of the characteristics of the bordering region types to accurately identify regional boundaries, improve the accuracy and efficiency of tissue region identification, and provide an accurate data basis for disease treatment or research.

[0213] Embodiment 3

[0214] See also Figure 3 , Figure 3 It is another tissue region image recognition system based on block division disclosed in an embodiment of the present invention. Figure 3 The described tissue region image recognition system based on tile division is applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 3 As shown, the tissue region image recognition system based on block division may include:

[0215] A memory 301 storing executable program codes;

[0216] a processor 302 coupled to the memory 301;

[0217] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the tissue region image recognition method based on block division described in the first embodiment.

[0218] Embodiment 4

[0219] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the tissue region image recognition method based on block division described in the first embodiment.

[0220] Embodiment 5

[0221] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps of the tissue region image recognition method based on block division described in the first embodiment.

[0222] The above describes specific embodiments of the present specification. In some cases, the actions or steps described can be performed in a different order than those in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily have to be performed in the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0223] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0224] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0225] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification may be in the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of this specification may be in the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0226] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0227] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0228] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0229] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0230] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0231] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0232] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0233] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0234] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0235] Finally, it should be noted that the tissue region image recognition method and system based on block division disclosed in the embodiment of the present invention discloses only the preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A tissue region image recognition method based on block division, characterized in that: The method comprises: Dividing the biological tissue image to be identified into blocks to obtain a plurality of image blocks; Inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types; According to a preset type bordering rule, at least one region combination is selected from the plurality of tissue regions; the region combination includes a plurality of tissue regions bordering each other; The regional image corresponding to the regional combination and the corresponding regional type are input into the regional boundary update model to obtain the corresponding regional update result; the regional update result is used to re-indicate the regional boundaries of different tissue regions in the regional image; the regional boundary update model is trained by a training data set including a plurality of training regional combination images and corresponding regional combination type annotations and regional boundary adjustment annotations; the regional boundary adjustment annotations include the original regional boundary and the modified regional boundary; the regional boundary update model is trained by the following steps: Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially; The training data set is input into a preset CNN architecture model for iterative training, and parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

2. The tissue region image recognition method based on block division according to claim 1, characterized in that: The region recognition model is trained by the following steps: Determine the pre-divided type region division annotation corresponding to the training tissue image; Segmenting the training tissue image to obtain a plurality of training image blocks; For each of the training image blocks, calculating the intersection area between the training image block and any type of division area in the type area division annotation; Determine the region type corresponding to the type division region with the largest intersection area as the type label corresponding to the training image block; Based on all the training image blocks and the corresponding type annotations, iteratively train the model of the preset CNN network architecture, and calculate the average value of the classification accuracy corresponding to each type annotation in each round of training to obtain a training performance parameter; the classification accuracy is an F1 score or an F2 score; The weight of the model obtained in the round of training with the highest training performance parameter is determined as the final model weight to obtain the region recognition model.

3. The tissue region image recognition method based on block division according to claim 2 is characterized in that: The step of inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types includes: Inputting each of the image blocks into a region recognition model to obtain a predicted region type corresponding to each of the image blocks; All the image blocks are spliced ​​together based on the segmentation rule to obtain a spliced ​​tissue image; Performing morphological operations on the spliced ​​tissue image to obtain a contour smoothed image; the contour smoothed image includes a plurality of tissue regions and corresponding region types; the tissue region includes a plurality of image blocks having the same corresponding predicted region type; the region type is the predicted region type corresponding to the image block in the tissue region; For each of the tissue regions in the contour smoothed image, determining the image block corresponding to the tissue region whose predicted region type is different from the region type of the tissue region, and obtaining at least one abnormal image block; Calculating a ratio of the number of all the abnormal image blocks to the total number of all the image blocks in the tissue region; Determining whether the quantity ratio is less than a preset ratio threshold; If so, modifying the predicted region types of all the abnormal image blocks to the region type of the tissue region; If not, input each of the image blocks in the tissue region into the prediction model corresponding to the predicted region type of any of the abnormal image blocks to obtain the prediction probability of other types corresponding to each of the image blocks; For each predicted region type of the abnormal image block, calculating an average value of other type prediction probabilities of all the image blocks in the tissue region corresponding to the predicted region type; The predicted region type with the highest average value is determined as the region type of the tissue region.

4. The tissue region image recognition method based on block division according to claim 1, characterized in that: The step of selecting at least one region combination from the plurality of tissue regions according to a preset type bordering rule includes: For any two tissue regions that have borders, calculating the difference between the regional images of the two tissue regions; Calculating the bordering ratio of the region boundaries of the two tissue regions; the bordering ratio is the ratio of the length of the common region boundary of the two tissue regions to the total region boundary length of the two tissue regions; Calculating the product of the difference and the bordering ratio to obtain the bordering parameters of the two tissue regions; Based on the bordering parameters, the multiple tissue regions are grouped according to a dynamic programming algorithm to obtain at least one region combination.

5. The tissue region image recognition method based on block division according to claim 4 is characterized in that: The step of grouping the plurality of tissue regions based on the bordering parameters according to a dynamic programming algorithm to obtain at least one region combination includes: Setting the objective function to minimize the number of all the region combinations and minimize the number of the tissue regions in the region combinations; Setting restrictions includes: The area types of any two tissue areas in the area combination meet a preset type condition; The border parameter of any two tissue regions in the region combination is greater than a preset first parameter threshold; The border parameter of any two tissue regions belonging to different region combinations is less than a preset second parameter threshold; the second parameter threshold is less than the first parameter threshold; Based on the objective function and the restriction condition, the multiple tissue regions are iteratively grouped according to a dynamic programming algorithm to obtain at least one region combination.

6. The tissue region image recognition method based on block division according to claim 1, characterized in that: The region type is a normal type, a necrosis type, a tumor type, an infiltration type or a background type; the region types of any two adjacent tissue regions in the region combination meet a preset type condition; The type condition is used to define that the region types of the two adjacent tissue regions are different, and the region type combination of the two tissue regions is normal-necrosis, normal-tumor, necrosis-tumor, infiltration-tumor or infiltration-necrosis.

7. A tissue region image recognition system based on block division, characterized in that: The system comprises: A division module, used for dividing the biological tissue image to be identified into multiple image blocks; A first recognition module, used for inputting the plurality of image blocks into a region recognition model to obtain a plurality of tissue regions and corresponding region types; A screening module, used for screening at least one area combination from the plurality of tissue areas according to a preset type bordering rule; the area combination includes a plurality of tissue areas bordering each other; The second recognition module is used to input the regional image corresponding to the regional combination and the corresponding regional type into the regional boundary update model to obtain the corresponding regional update result; the regional update result is used to re-indicate the regional boundaries of different tissue regions in the regional image; the regional boundary update model is trained by a training data set including a plurality of training regional combination images and corresponding regional combination type annotations and regional boundary adjustment annotations; the regional boundary adjustment annotations include original regional boundaries and modified regional boundaries; the regional boundary update model is trained by the following steps: Determine the loss function as the weighted sum of the first cross entropy and the second cross entropy; the first cross entropy is the cross entropy between the recognition boundary output by the model and the original region boundary; the second cross entropy is the cross entropy between the recognition boundary output by the model and the modified region boundary; wherein the weight of the first cross entropy is less than the second cross entropy; the weight of the first cross entropy is the product of the first weight and the second weight; the first weight is proportional to the boundary length of the original region boundary; the second weight is proportional to the total number of all different region types in the corresponding training region combination image; the weight of the second cross entropy is related to the corresponding region combination type label, and when the region combination type labels are normal-tumor, normal-necrosis, infiltration-necrosis, necrosis-tumor or infiltration-tumor, the weight of the second cross entropy increases sequentially; The training data set is input into a preset CNN architecture model for iterative training, and parameters are updated based on the loss function and gradient descent algorithm during training until the training converges to obtain the region boundary update model.

8. A tissue region image recognition system based on block division, characterized in that: The system comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the tissue region image recognition method based on block division as described in any one of claims 1-6.

Images