Postoperative pathological evaluation method, survival prediction method and system for esophageal carcinoma

Abstract

The present invention discloses a postoperative pathological assessment method and survival prediction method and system for esophageal cell carcinoma, comprising: acquiring and preprocessing WSI image data to obtain cropped images of the WSI image data at low, medium, and high magnification viewing angles, which are then sequentially input into a selective spatial state model through feature extraction; updating the diagnostic states through a long short-term memory network, and adjusting the contribution of each diagnostic state using selective memory weights; fusing all diagnostic states and outputting the probability of tissue type through a fully connected layer and a softmax function; calculating the percentage of residual viable tumor, and obtaining a pathological assessment result based on the percentage of residual viable tumor. The present invention can simulate the doctor's diagnostic process, reduce the workload of pathologists, improve the consistency and popularity of assessments, and construct a survival prediction model that can predict survival rates and assist in clinical treatment decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of pathological assessment and survival prediction, and in particular to a postoperative pathological assessment method, a survival prediction method and a system for esophageal cell carcinoma. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Data show that esophageal cancer is one of the most common malignant tumors with high morbidity and mortality. Esophageal squamous cell carcinoma (ESCC) is the predominant histological subtype of ESCC and carries a poor prognosis. Patients with locally advanced ESCC typically choose preoperative chemotherapy, but its efficacy has plateaued. The emergence of immunotherapy has brought new breakthroughs in esophageal cancer treatment, leading to the development of preoperative chemotherapy combined with immunotherapy.

[0004] However, there are currently many difficulties in evaluating the pathological response after preoperative chemotherapy combined with immunotherapy for ESCC. In order to obtain a more accurate description, pathologists need to conduct more diagnostic readings, which increases the workload of pathological evaluation. Due to differences in grading systems, diagnostic grading, and diagnostic descriptions, the evaluation consistency of existing technologies is poor. These inconsistencies may cause confusion and misjudgment in communication and learning between physicians, and even affect diagnosis and treatment decision-making. Furthermore, the evaluation of pathological responses requires experienced pathological diagnostic experts, who are in short supply even in regional pathology diagnostic centers. Furthermore, the training cycle is long, making it difficult to expand personnel in a short period of time. This makes the evaluation of pathological responses using existing technologies difficult to popularize. Furthermore, the pathological evaluation models in existing technologies cannot simulate the doctor's diagnostic process, resulting in low accuracy and poor evaluation results.

[0005] At the same time, current pathology image-based survival prediction for ESCC patients undergoing preoperative chemotherapy combined with immunotherapy also has certain limitations. Most existing technologies rely on extracting pathological features from untreated tissue samples to predict survival, which is not suitable for the preoperative neoadjuvant chemotherapy combined with immunotherapy model for esophageal cancer. Furthermore, there is limited evidence on the impact of these extracted histological features on survival prognosis, and there is a lack of strong theoretical support. Summary of the Invention

[0006] In order to solve the above problems, the present invention proposes a postoperative pathological assessment method, a survival prediction method and a system for esophageal cell carcinoma. The present invention can simulate the doctor's diagnostic process, reduce the workload of pathologists, improve the consistency and popularity of the assessment, and through the constructed survival prediction model, it can predict the survival rate and assist in clinical treatment decision-making.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a method for postoperative pathological assessment of esophageal cell carcinoma, comprising:

[0009] Acquire the patient's HE slice data, scan and generate WSI image data, and preprocess the WSI image data to obtain cropped thumbnails of the WSI image data at low, medium, and high magnification viewing angles.

[0010] Feature extraction is performed on the cropped small images at low, medium and high magnification viewing angles to obtain image features at different viewing angles, namely overall features, local features and detail features;

[0011] Image features from different perspectives are sequentially input into the selective spatial state model. The image features from each perspective are considered as the input at the current moment, and the corresponding diagnostic state is obtained. The diagnostic state is updated through a long short-term memory network, and the contribution of each diagnostic state is adjusted using selective memory weights. All diagnostic states are fused and the probability of tissue type is output through a fully connected layer and a softmax function.

[0012] According to the probability of the tissue type, the diagnosis result of the cropped image is obtained, and the residual viable tumor percentage is calculated according to the diagnosis result. According to the residual viable tumor percentage, the pathological evaluation result is obtained.

[0013] According to a further technical solution, after the WSI image data is generated, the tumor bed area and the residual tumor area are outlined, and then the WSI image data is preprocessed.

[0014] A further technical solution is to preprocess the WSI image data as follows: first, a thumbnail image of a preset magnification is obtained based on the read WSI image data, the thumbnail image is converted into a grayscale image, and then the grayscale image is divided into a tissue area and a background area; then, the WSI image data is cropped into different types of thumbnails according to the tumor bed area, residual tumor area, and tissue area, including: a tumor bed thumbnail, a residual tumor thumbnail, and other tissue thumbnails. After obtaining different types of thumbnails, the coordinates of each thumbnail are mapped to images of different magnifications, and cropped thumbnails of the WSI image data at low, medium, and high magnification viewing angles are obtained from the mapped areas; finally, color correction is performed on all thumbnails.

[0015] In a further technical solution, the updated diagnosis status includes the image features input at the current moment and the diagnosis status at the previous moment. The specific updating method is:

[0016] ;in, Indicates the updated diagnostic status; Indicates the diagnostic status at the previous moment; Represents the image features input at the current moment.

[0017] In a further technical solution, the selective memory weight is calculated by a Sigmoid activation function, and the formula is: ;in, represents the selective memory weight, ( ) represents the Sigmoid activation function, and Represents the learning parameters, which control the degree of memory enhancement for each perspective;

[0018] All diagnostic states are fused into global features, and the fused global features are expressed as:

[0019] ;in, Represents global features, Represents a learnable coefficient that controls the image features of each view Contribution to global features.

[0020] A further technical solution is to obtain the diagnostic results of the cropped image based on the probability of tissue type, determine the tumor bed area and the residual viable tumor area, and calculate the residual viable tumor percentage as follows: ;in, is the percentage of residual viable tumor, Sb is the area of the tumor bed in the HE section, St The percentage of residual viable tumor in the HE section was used for pathological evaluation, which was divided into no tumor cells, significant pathological response, partial pathological response, and no pathological response.

[0021] In a further technical solution, the loss function of the selective spatial state model is a cross entropy function, which is specifically calculated as follows: ;in, is the number of label types, y ( k ) is the data label in one-hot form, p ( k ) is the probability of tissue type output by the pathology assessment model.

[0022] In a second aspect, the present invention provides a postoperative pathological evaluation system for esophageal cell carcinoma, comprising:

[0023] The data acquisition and preprocessing module is configured to: acquire HE slice data of the patient, scan and generate WSI image data, and preprocess the WSI image data to obtain cropped thumbnails of the WSI image data at low, medium, and high magnification viewing angles;

[0024] The feature extraction module is configured to extract features of the cropped images at low, medium, and high magnification viewing angles to obtain image features at different viewing angles, namely, overall features, local features, and detail features;

[0025] The data processing module is configured to: sequentially input image features from different perspectives into the selective spatial state model, with the image features from each perspective being considered as input at the current moment, and obtain the corresponding diagnostic state, wherein the diagnostic state is updated via a long short-term memory network, and the contribution of each diagnostic state is adjusted using selective memory weights; all diagnostic states are fused and output as tissue type probabilities through a fully connected layer and a softmax function;

[0026] The pathology assessment module is configured to obtain a diagnosis result of the cropped image based on the probability of the tissue type, calculate the percentage of residual viable tumor based on the diagnosis result, and obtain a pathology assessment result based on the percentage of residual viable tumor.

[0027] In a third aspect, the present invention provides a method for predicting postoperative survival of esophageal cell carcinoma, based on the postoperative pathological assessment method for esophageal cell carcinoma described in any one of the first aspects, comprising:

[0028] Obtaining patient survival characteristics, and dividing the patients into high-risk group and low-risk group according to the median of the patient survival characteristics;

[0029] The Log-rank test was used to calculate the difference in survival curves between the high-risk group and the low-risk group, and the survival characteristics were preliminarily screened. The LASSO Cox regression model was used to further screen the survival characteristics. The LASSO Cox regression model was optimized using the minimum MSE of ten-fold cross-validation to obtain survival characteristics with non-zero coefficients, which were input into the multivariate Cox proportional hazard model to output the survival prediction results.

[0030] In a fourth aspect, the present invention provides a postoperative survival prediction system for esophageal cell carcinoma, comprising:

[0031] The survival feature acquisition module is configured to: acquire the patient's survival feature, and divide the patient into a high-risk group and a low-risk group according to the median of the patient's survival feature;

[0032] The survival prediction module is configured to: use the Log-rank test to calculate the difference in survival curves between the high-risk group and the low-risk group, and preliminarily screen the survival characteristics; use the LASSO Cox regression model to further screen the survival characteristics; use the minimum ten-fold cross-validation MSE to optimize the LASSO Cox regression model, obtain survival characteristics with non-zero coefficients, and input them into the multivariate Cox proportional hazard model to output the survival prediction results.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention constructs a Mamba model based on multi-perspective memory enhancement, whose core is the Selective Space State Model (SSM). This model can simulate the selective memory process of pathologists during the diagnosis process, thereby improving assessment accuracy: By simulating the doctor's diagnostic process, the present invention integrates image features from different perspectives for comprehensive diagnosis, and selectively enhances the memory of image features at different magnifications. Utilizing the Selective Space State Model for deep learning of pathological sections, the model can accurately identify pathological image features and the correlation between features, thereby obtaining the probabilities of different pathological tissue types and accurately assessing pathological reactions. This greatly improves the accuracy of pathological assessment and increases the diagnostic accuracy to over 90%.

[0035] 2. This invention uses machine learning to perform deep learning, eliminating the need for pathologists to perform extensive diagnostic film reading, significantly reducing the workload of pathology diagnosis. This invention avoids the problem of poor assessment consistency caused by different grading systems, diagnostic grading, and diagnostic descriptions, reduces confusion and misjudgment during communication and learning between physicians, and provides a more reliable basis for diagnosis and treatment decision-making. Furthermore, the pathology assessment model provided by this invention can be used by primary care hospitals to address the difficulties in pathology assessment caused by a shortage of professional talent, thereby increasing the popularity of pathology assessment after preoperative chemotherapy and immunotherapy for ESCC.

[0036] 3. The present invention constructs a postoperative prognosis survival prediction model based on the obtained pathological assessment results and combined with clinical characteristic factors, which can screen out the key factors affecting survival, thereby formulating personalized treatment plans for patients in a targeted manner, improving the accuracy and effectiveness of postoperative treatment, and also improving the accuracy of survival prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0038] Figure 1 This is a structural diagram of the pathology assessment model of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the following detailed description is illustrative and is intended to further illustrate the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as commonly understood by those of ordinary skill in the art to which the present invention belongs. The embodiments of the present invention and the features of the embodiments may be combined with each other unless otherwise specified.

[0041] Example 1

[0042] This embodiment provides a method for postoperative pathological evaluation of esophageal carcinoma, which specifically includes the following steps:

[0043] S1. Acquire HE slice data of the patient, scan and generate WSI image data, and preprocess the WSI image data to obtain cropped thumbnails of the WSI image data at low, medium, and high magnification viewing angles.

[0044] S2. Feature extraction is performed on the cropped images at low, medium, and high magnification viewing angles to obtain image features at different viewing angles, namely, overall features, local features, and detail features;

[0045] S3. Image features from different perspectives are sequentially input into the selective spatial state model. The image features from each perspective are considered as the input at the current moment, and the corresponding diagnostic state is obtained. The diagnostic state is updated using a long short-term memory network, and the contribution of each diagnostic state is adjusted using selective memory weights. All diagnostic states are fused and the probability of tissue type is output through a fully connected layer and a softmax function.

[0046] S4. Obtain a diagnosis result of the cropped image based on the probability of the tissue type, calculate the percentage of residual viable tumor based on the diagnosis result, and obtain a pathological assessment result based on the percentage of residual viable tumor.

[0047] In step S1, for each enrolled patient, multi-slice scanning of hematoxylin and eosinophilic (HE) slide data was performed using a Pannoramic MIDI slide scanner (3DHISTECH Ltd., Budapest, Hungary) at a resolution of 0.25 μm / pixel (40x magnification) to generate WSI image data for subsequent analysis. After the WSI data were generated, expert pathologists delineated the tumor bed and residual tumor areas as gold standards (i.e., classification labels) for the study. The WSI image data were then preprocessed.

[0048] Before building a deep learning model, you first need to preprocess the WSI image data to convert it into multiple small images that can be used for model analysis. The specific method for preprocessing the WSI image data is as follows:

[0049] First, the WSI image data is read from the OpenSlide library, and a thumbnail image with a preset magnification is obtained. In this embodiment, the preset magnification is 2.5 times (4 / pixel), then use OpenCV's cv2.COLOR_RGB2GRAY to convert the thumbnail image to a grayscale image, and then use cv2.THRESH_BINARY_INV at a threshold of 220 to segment the grayscale image into tissue areas (less than the threshold) and background areas (greater than the threshold); then subdivide the tissue area on the WSI image data into non-overlapping small images. Specifically: at 2.5 times magnification, based on the tumor bed area, residual tumor area, and tissue area on the WSI image data outlined by the pathologist, the WSI image data was cropped to a size of 224 224 Three different types of small images, including: tumor bed small image, residual tumor small image and other tissue small image. After obtaining different types of small images, if a small image contains more than one tissue type, the tissue type with an area greater than 50% is used as the tissue type of its small image. Subsequently, the coordinates of each small image are mapped to images of different magnifications (5x, 10x, 20x, 40x), and 4, 16, 64 and 256 images of size 224 are cropped from the mapped area. 224 cropped images at different magnifications. 2.5x is low magnification, 5x and 10x are medium magnification, and 20x and 40x are high magnification.

[0050] After obtaining small images of different types, the coordinates of each small image are mapped to images of different magnifications, and cropped small images of different magnification perspectives are obtained from the mapped areas. Finally, color correction is performed on all small images. Specifically, in this embodiment, the Vahadane method of the StainTools tool is used to perform color correction on all small images, and its staining density map is combined with the color distribution of the target image selected by the pathologist, so as to unify the color distribution of all small images, reduce the impact of staining differences on subsequent analysis, and retain the structural information of the image at the same time, so that the model can focus on the structural differences in the image and improve the tissue identification ability of the model.

[0051] After the above preprocessing, in step S2, feature extraction is performed on the cropped images at low, medium, and high magnification viewing angles to obtain image features at different viewing angles. The cropped images at each viewing angle are processed by a convolutional neural network (CNN) to extract image features from the overall to the local and then the detailed perspectives from low to medium to high magnification. The low magnification perspective (2.5x) focuses on the overall tissue structure, while the high magnification perspective (40x) focuses on the detailed features at the cellular level. To this end, different convolutional neural networks are used to extract effective features from images of different magnifications. Specifically:

[0052] Low magnification (2.5x): Use the pre-trained ResNet50 network to extract overall features, mainly learning the macroscopic structure of the tissue, such as shape, size, and distribution.

[0053] Medium magnification (5x, 10x): A deeper ResNet152 network is used to capture local features and focus on detailed differences between tissues, such as tumor boundaries and blood vessels.

[0054] High-magnification perspective (20x, 40x): Use a network with detail-capturing capabilities (U-Net) to extract detailed features at the cellular level, focusing on microscopic information such as cell morphology and cell spacing.

[0055] Finally, the features extracted from images at different perspectives are F i ,in i 2.5, 5, 10, 20 and 40 times magnification respectively.

[0056] In step S3, in this embodiment, a Mamba model based on multi-perspective memory enhancement is constructed, and its core structure is a selective state space model (SSM). In order to simulate the selective memory process of pathologists during the diagnosis process, this embodiment uses a parameterized selective SSM to learn the importance of image features of each perspective to the diagnosis. Each perspective (magnification) is regarded as a time point and a time step, and the image features under different perspectives are input into the selective state space model in sequence, that is, the low-magnification image features (such as 2.5 times) are used as the earliest input, and the high-magnification image features (such as 40 times) are used as the latest input. All image features are input into the selective state space model in chronological order. The core of this model is to strengthen the memory of the diagnosis process through the image features of each perspective. Image features of each perspective F i is considered as the current moment (time point t= i ) input, each time the image feature is input F iThe corresponding diagnostic status will be obtained, and the diagnostic status is updated through the Long Short-Term Memory (LSTM) network. i , its image features F i and the diagnostic status at the previous moment s i-1 It is input into the LSTM network for updating. The updated diagnostic state includes the image features input at the current moment and the diagnostic state at the previous moment. The updated diagnostic state si can be expressed by the following formula: ;in, Indicates the updated diagnostic status; Indicates the diagnostic status at the previous moment; Represents the image features input at the current moment.

[0057] In order to simulate the pathologist's memory of different magnification image features during the diagnosis process, this embodiment introduces a selective memory mechanism. To adjust the contribution of each diagnostic state. Memory weight Calculated by Sigmoid activation function, the formula is as follows: ;in, represents the selective memory weight, ( ) represents the Sigmoid activation function, and Represents the learning parameters that control the degree of memory reinforcement for each perspective; in this way, small crops at low magnification perspectives (such as 2.5x) may be given lower weights, while small crops at high magnification perspectives (such as 40x) may be given higher weights. The model dynamically adjusts the importance of the image features of each perspective to the final diagnosis.

[0058] The diagnostic status of each perspective is updated and its importance is adjusted through the selective memory mechanism. Then the diagnostic status of all perspectives needs to be fused into a global feature for the final classification decision. Feature fusion is performed by weighted summation. The fused global feature It can be expressed as:

[0059] ;in, Represents global features, Represents a learnable coefficient that controls the image features of each view Contribution to global features.

[0060] The Mamba model in this embodiment adopts a hardware-aware design. Considering the processing requirements of large-scale pathological images, the model maximizes parallel computing during inference to improve efficiency. During training, parallel computing is performed using efficient GPU clusters to accelerate the processing of large-scale image data. Furthermore, the model architecture is optimized to adapt to hardware resources, ensuring efficient operation even with limited computing resources. The model in this embodiment was trained and validated using the Pytorch 1.9.0 framework on an Ubuntu server equipped with an NVIDIA GeForce RTX 4090 GPU.

[0061] During the training process of this embodiment, the optimization goal is to minimize the classification loss of tissue types. The loss function of the model is the cross entropy (CE) function, which is specifically calculated as follows: ;in, is the number of label types, y ( k ) is the data label in one-hot form, p ( k ) is the probability of tissue type output by the pathology assessment model.

[0062] During the training process, the Adam optimizer is used, and the formula is:

[0063] ;

[0064] in, are the parameters of the current model, is the learning rate, is the gradient of the loss function.

[0065] In this example, all patients were assigned to training, testing, and validation sets in a 3:1:1 ratio. The training set was used to optimize the hyperparameters of the deep learning network; the testing set was used to select the best-performing model; and the validation set was used to verify the model's diagnostic capabilities. Standard evaluation metrics (accuracy, precision, recall, F1 score, etc.) were used to validate the model's performance.

[0066] In step S3, based on the probability of tissue type, the diagnostic result of the cropped image is obtained, and the tumor bed area and the residual viable tumor area are determined. The residual viable tumor percentage is calculated as follows: ;in, is the percentage of residual viable tumor, Sb is the area of the tumor bed in the HE section, StIt is the area of residual viable tumor in the HE section. According to the percentage of residual viable tumor, the pathological evaluation results are divided into no tumor cells seen, significant pathological remission (RVT ≤ 10%), partial pathological remission (10% < RVT < 90%), and no pathological remission (RVT ≥ 90%).

[0067] Both the training and testing of the model are completed at the sub-image level. After the model is trained, each sub-image at 40x magnification is combined with its corresponding four lower-magnification sub-images (20x, 10x, 5x, 2.5x). After feature extraction, an image feature matrix is formed and input into the trained Mamba model for inference to predict its tissue type, thereby obtaining the tissue type of each sub-image at 40x magnification on the entire pathological slide. If the number of sub-images predicted as tumor regions is and the number of sub-images predicted as tumor beds is then, based on the deep learning model is calculated as follows:

[0068] ; finally, referring to the Cottrell pathological response evaluation system, the tumor regression grade is determined.

[0069] Example 2

[0070] In this example, a postoperative pathological evaluation system for esophageal cell carcinoma is disclosed, including:

[0071] A data acquisition and preprocessing module, configured to: acquire patient HE section data, scan to generate WSI image data, and preprocess the WSI image data to obtain cropped sub-images of the WSI image data at low magnification, medium magnification, and high magnification perspectives;

[0072] A feature extraction module, configured to: perform feature extraction on the cropped sub-images at low magnification, medium magnification, and high magnification perspectives respectively to obtain image features at different perspectives, namely global features, local features, and detail features;

[0073] A data processing module, configured to: sequentially input the image features at different perspectives into a selective spatial state model. The image features of each perspective are regarded as the input at the current moment and the corresponding diagnostic state is obtained. Among them, the diagnostic state is updated through a long short-term memory network, and the contribution degree of each diagnostic state is adjusted through selective memory weights; all diagnostic states are fused, and the probability of the tissue type is output through a fully connected layer and a Softmax function;

[0074] The pathology assessment module is configured to obtain a diagnosis result of the cropped image based on the probability of the tissue type, calculate the percentage of residual viable tumor based on the diagnosis result, and obtain a pathology assessment result based on the percentage of residual viable tumor.

[0075] Example 3

[0076] This embodiment discloses a method for predicting postoperative survival of esophageal cell carcinoma, based on the pathological evaluation method after preoperative chemotherapy combined with immunotherapy for esophageal cell carcinoma described in any one of the first embodiments, comprising:

[0077] First, the patient survival characteristics were obtained, including survival information, pathological assessment results, gender, age, ECOG score, BMI, lesion location, lesion length, TNM clinical stage before treatment, PD-L1 expression status, TMB status, MSI status, and immune microenvironment status in postoperative specimens.

[0078] Patients were then divided into high-risk and low-risk groups based on the median of each of the above survival characteristics. The log-rank test was used to calculate the difference in survival curves between the high-risk and low-risk groups, initially screening for survival characteristics, specifically those with a p < 0.05. The LASSO Cox regression model was then used to further screen for survival characteristics. The LASSO Cox regression model was optimized using the minimum mean squared error (MSE) of ten-fold cross-validation to obtain survival characteristics with nonzero coefficients. These were then input into a multivariate Cox proportional hazards model to output survival prediction results. The model's predictive ability was evaluated using the C-index, Kaplan-Meier survival curves, and receiver operating characteristic (ROC) curves.

[0079] Example 4

[0080] This embodiment discloses a postoperative survival prediction system for esophageal carcinoma, comprising:

[0081] The survival feature acquisition module is configured to: acquire the patient's survival feature, and divide the patient into a high-risk group and a low-risk group according to the median of the patient's survival feature;

[0082] The survival prediction module is configured to: use the Log-rank test to calculate the difference in survival curves between the high-risk group and the low-risk group, and preliminarily screen the survival characteristics; use the LASSO Cox regression model to further screen the survival characteristics; use the minimum ten-fold cross-validation MSE to optimize the LASSO Cox regression model, obtain survival characteristics with non-zero coefficients, and input them into the multivariate Cox proportional hazard model to output the survival prediction results.

[0083] It should be noted here that each module in this embodiment corresponds one-to-one to the method in Example 1, and the specific implementation process is the same, which will not be repeated here.

[0084] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0085] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A method for postoperative pathological evaluation of esophageal cell carcinoma, characterized in that: include: Acquiring HE slice data of the patient, scanning to generate WSI image data, and preprocessing the WSI image data. After generating the WSI image data, outlining the tumor bed area and the residual tumor area, and then preprocessing the WSI image data; After preprocessing, cropped images of the WSI image data at low, medium, and high magnification viewing angles are obtained; Feature extraction is performed on the cropped small images at low, medium and high magnification viewing angles to obtain image features at different viewing angles, namely overall features, local features and detail features; The image features from different perspectives are sequentially input into the selective spatial state model. The image features from each perspective are regarded as the input at the current moment, and the corresponding diagnostic state is obtained. The diagnostic state is updated through the long short-term memory network, and the contribution of each diagnostic state is adjusted by the selective memory weight. All diagnostic states are integrated and the probability of tissue type is output through the fully connected layer and the Softmax function. The selective memory weight is calculated using the Sigmoid activation function, and the formula is: ;in, represents the selective memory weight, ( ) represents the Sigmoid activation function, and Represents the learning parameters, which control the degree of memory enhancement for each perspective; All diagnostic states are fused into global features, and the fused global features are expressed as: ;in, Represents global features, Represents a learnable coefficient that controls the image features of each view Contribution to global features; According to the probability of the tissue type, the diagnosis result of the cropped image is obtained, and the residual viable tumor percentage is calculated according to the diagnosis result. According to the residual viable tumor percentage, the pathological evaluation result is obtained.

2. The postoperative pathological evaluation method for esophageal cell carcinoma according to claim 1, wherein: The method for preprocessing WSI image data is as follows: first, a thumbnail image with a preset magnification is obtained based on the read WSI image data, the thumbnail image is converted into a grayscale image, and then the grayscale image is segmented into tissue area and background area; The WSI image data is then cropped into different types of thumbnails based on the tumor bed, residual tumor, and tissue regions, including: tumor bed thumbnail, residual tumor thumbnail, and remaining tissue thumbnail. After obtaining different types of thumbnails, the coordinates of each thumbnail are mapped onto images of different magnifications. Cropped thumbnails of the WSI image data at low, medium, and high magnification viewing angles are obtained from the mapped areas. Finally, color correction is performed on all thumbnails.

3. The postoperative pathological evaluation method for esophageal cell carcinoma according to claim 1, wherein: The updated diagnosis status includes the image features input at the current moment and the diagnosis status at the previous moment. The specific update method is: ;in, Indicates the updated diagnostic status; Indicates the diagnostic status at the previous moment; Represents the image features input at the current moment.

4. The postoperative pathological evaluation method for esophageal cell carcinoma according to claim 1, wherein: Based on the probability of tissue type, the diagnostic results of the cropped image are obtained, and the tumor bed area and residual viable tumor area are determined. The percentage of residual viable tumor is calculated as follows: ;in, is the percentage of residual viable tumor, Sb is the area of the tumor bed in the HE section, St The percentage of residual viable tumor in the HE section was used for pathological evaluation, which was divided into no tumor cells, significant pathological response, partial pathological response, and no pathological response.

5. The postoperative pathological evaluation method for esophageal cell carcinoma according to claim 1, wherein: The loss function of the selective spatial state model is the cross entropy function, which is specifically calculated as follows: ;in, is the number of label types, y ( k ) is the data label in one-hot form, p ( k ) is the probability of tissue type output by the pathology assessment model.

6. A postoperative pathological evaluation system for esophageal cell carcinoma, characterized in that: include: The data acquisition and preprocessing module is configured to: acquire HE slice data of the patient, scan and generate WSI image data, and preprocess the WSI image data; after generating the WSI image data, outline the tumor bed area and the residual tumor area, and then preprocess the WSI image data; After preprocessing, cropped images of the WSI image data at low, medium, and high magnification viewing angles are obtained; The feature extraction module is configured to extract features of the cropped images at low, medium, and high magnification viewing angles to obtain image features at different viewing angles, namely, overall features, local features, and detail features; a data processing module configured to sequentially input image features from different perspectives into a selective spatial state model, wherein the image features from each perspective are regarded as input at the current moment, and a corresponding diagnostic state is obtained, wherein the diagnostic state is updated via a long short-term memory network, and the contribution of each diagnostic state is adjusted via a selective memory weight; All diagnostic states are integrated and the probability of tissue type is output through the fully connected layer and the Softmax function. The selective memory weight is calculated using the Sigmoid activation function, and the formula is: ;in, represents the selective memory weight, ( ) represents the Sigmoid activation function, and Represents the learning parameters, which control the degree of memory enhancement for each perspective; All diagnostic states are fused into global features, and the fused global features are expressed as: ;in, Represents global features, Represents a learnable coefficient that controls the image features of each view Contribution to global features; The pathology assessment module is configured to obtain a diagnosis result of the cropped image based on the probability of the tissue type, calculate the percentage of residual viable tumor based on the diagnosis result, and obtain a pathology assessment result based on the percentage of residual viable tumor.

7. A method for predicting postoperative survival of esophageal cell carcinoma, based on the postoperative pathological evaluation method for esophageal cell carcinoma according to any one of claims 1 to 5, characterized in that: include: Obtaining patient survival characteristics, and dividing the patients into high-risk group and low-risk group according to the median of the patient survival characteristics; The Log-rank test was used to calculate the difference in survival curves between the high-risk group and the low-risk group, and to preliminarily screen the survival characteristics; The LASSO Cox regression model was used to further screen survival characteristics. The LASSO Cox regression model was optimized using the minimum MSE of ten-fold cross-validation to obtain survival characteristics with non-zero coefficients, which were input into the multivariate Cox proportional hazard model to output survival prediction results.

8. A postoperative survival prediction system for esophageal carcinoma, based on the postoperative survival prediction method for esophageal carcinoma according to claim 7, characterized in that: include: The survival feature acquisition module is configured to: acquire the patient's survival feature, and divide the patient into a high-risk group and a low-risk group according to the median of the patient's survival feature; The survival prediction module is configured to: calculate the difference in survival curves between the high-risk group and the low-risk group using a Log-rank test, and preliminarily screen survival characteristics; The LASSO Cox regression model was used to further screen survival characteristics. The LASSO Cox regression model was optimized using the minimum MSE of ten-fold cross-validation to obtain survival characteristics with non-zero coefficients, which were input into the multivariate Cox proportional hazards model to output survival prediction results.

Images