An AI-assisted pathological sample diagnosis method and diagnosis system

Abstract

The application discloses an AI-assisted pathological sample diagnosis method and a diagnosis system, relates to the technical field of medical image assistance, and comprises the following steps: acquiring a pathological sample from a digital pathology image library through a data access layer, and acquiring user information corresponding to the pathological sample from a clinical information library; performing multi-scale attention feature extraction and representation learning on the pathological sample through a multi-scale feature extraction network; quantifying the dynamic uncertainty of the pathological sample through a Bayesian neural network; performing multi-modal knowledge fusion; performing diagnosis decision process visualization output through a visualization module; evaluating and verifying the diagnosis system based on the visualization output result; the recognition ability of microstructure in the pathological sample and the confidence evaluation accuracy of the diagnosis result are significantly improved, the explainability and clinical practicability of the system are enhanced, incremental learning and long-term verification are supported, and the robustness, generalization ability and cost-effectiveness of the diagnosis system are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of medical image-assisted technology, and in particular to an AI-assisted method and system for diagnosing pathological samples. Background Technology

[0002] Pathological sample diagnosis is the gold standard for disease diagnosis and has crucial clinical value for the early detection, accurate classification, and treatment planning of major diseases.

[0003] In existing technologies, pathological sample diagnosis methods mainly rely on pathologists to observe tissue sections under a microscope and make comprehensive judgments based on clinical information and professional experience. First, manual slide reading is time-consuming and labor-intensive, especially inefficient during large-scale screening or diagnosis of difficult cases. Moreover, the diagnostic results are easily affected by the physician's experience level and subjective factors, resulting in poor diagnostic consistency. Second, pathological images contain a massive amount of cellular and microscopic structural information, which is difficult for the human eye to fully capture and quantify, leaving a large amount of potential diagnostic information unutilized. Finally, pathological diagnosis requires the integration of multi-source data such as imaging and clinical information, but existing methods lack effective data fusion mechanisms and struggle to establish cross-modal semantic associations, limiting the comprehensiveness and accuracy of diagnosis.

[0004] Therefore, an AI-assisted method and system for diagnosing pathological samples are provided to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an AI-assisted pathological sample diagnosis method and system, which solves the problems of low efficiency, poor diagnostic consistency, insufficient fusion of multi-source data, and opaque decision-making process in the prior art, thereby achieving efficient, accurate, and interpretable pathological diagnosis assistance.

[0006] To achieve the above objectives, the present invention provides an AI-assisted method for diagnosing pathological samples, comprising the following steps: S1: Obtain pathological samples from the digital pathology image library and obtain user information corresponding to the pathological samples from the clinical information database through the data access layer; S2: Multi-scale attention feature extraction and representation learning of pathological samples are performed using a multi-scale feature extraction network; S3: Quantify the dynamic uncertainty of pathological samples using a Bayesian neural network; S4: Perform multimodal knowledge fusion; S5: Visualize the diagnostic decision-making process through the visualization module; S6: Evaluate and validate the diagnostic system based on the visualized output results.

[0007] Preferably, step S2 specifically includes the following steps: S21: The microstructure of pathological samples is identified through an attention mechanism CNN, and multi-resolution pyramid feature extraction is performed to obtain multi-scale feature representations. Multi-scale feature representation Specifically set as follows: ; in, This indicates the input pathological sample. Indicates the number of pyramid levels. Indicates the first The feature transformation function of the layer, Indicates the first The graph neural network modeling function of the layer, Indicates the first Feature extraction function of layer Indicates the first Parameters of the layer feature extraction network, Indicates the first Parameters of the layer feature transformation network; S22: Perform feature weighting for the attention mechanism, and calculate the... Layer Attention weight of each region , No. Layer Attention weight of each region Specifically set as follows: ; in, Represents an exponential function. The weight vector representing the attention mechanism. This represents the hyperbolic tangent activation function. The transformation matrix representing the attention mechanism. Indicates the first Layer Feature vectors of each region This indicates the bias term.

[0008] Preferably, step S3 specifically includes the following steps: S31: Predict pathological samples using a Bayesian neural network and quantify model uncertainty. Specifically set as follows: ; in, This represents the output value of the Bayesian neural network. Represents the parameters of a Bayesian neural network. This indicates the probability of the output value. This represents the model's uncertainty and prediction variance. Indicates the number of samples. Indicates the first The parameters corresponding to the next sampling , express The mean of the predictions; S32: Predict pathological samples using a Bayesian neural network and quantify data uncertainty. Specifically set as follows: ; in, This indicates the variance in the prediction of data uncertainty. S33: Calculate the dynamic uncertainty by comprehensively evaluating the uncertainty score. Dynamic uncertainty Specifically set as follows: ; ; in, Indicating model uncertainty The weighting coefficients, Indicating data uncertainty The weighting coefficients.

[0009] Preferably, in step S31, the prediction result Specifically set as follows: ; in, Represents the training data set. This represents the posterior probability distribution of the Bayesian neural network.

[0010] Preferably, step S4 specifically includes the following steps: S41: By using the cross-modal attention fusion module, pathological samples and corresponding user information are projected into a shared feature space to establish semantic associations between pathological samples and user information, enabling collaborative expression of heterogeneous data. S42: Based on the pathology knowledge base, clinical guidelines and historical research results, a hierarchical knowledge graph is constructed through the knowledge graph reasoning module. The hierarchical knowledge graph includes entity relationships, diagnostic rules and clinical pathways. S43: Store representative samples through an incremental learning module and dynamically evaluate the Bayesian neural network parameters corresponding to the representative samples. The importance of Bayesian neural network parameters based on importance ranking Strengthen constraints and monitor dynamic uncertainties. .

[0011] Preferably, step S5 specifically includes the following steps: S51: Display the lesion areas in the pathological sample using a color heatmap, with lesion areas of high abnormality set to dark color and lesion areas of low abnormality set to light color; S52: Retrieve cases similar to the current case, display the pathological images and diagnostic results of the current case and similar cases side by side, and determine the accuracy of the diagnostic results of the current case; S53: The diagnostic reasoning process is unfolded in the form of a tree diagram, and a complete thinking chain is constructed based on the decision logic of each branch of the tree diagram.

[0012] Preferably, step S6 specifically includes the following steps: S61: Evaluate the diagnostic accuracy, clinical usability, robustness, and generalization ability of a diagnostic system based on multi-dimensional performance indicators; S62: Set up multiple groups of randomized controlled cases, conduct long-term follow-up verification of multiple groups of randomized controlled cases, and perform cost-benefit analysis; S63: Track user information corresponding to pathological samples in real time, obtain treatment data and prognostic diagnosis data from user information, compare the prognostic diagnosis results with the system's diagnosis results, if they are consistent, use the cases for positive model reinforcement and self-evolution, if they are inconsistent, automatically screen and issue warnings and drive model error correction and optimization.

[0013] A diagnostic system for an AI-assisted pathological sample diagnosis method includes a data access layer, a multi-scale feature extraction network, a Bayesian neural network, a cross-modal attention fusion module, a knowledge graph reasoning module, an incremental learning module, and a visualization module.

[0014] Therefore, the AI-assisted pathological sample diagnosis method and system described above have the following beneficial effects: (1) This scheme combines a multi-scale feature extraction network with an attention mechanism and a graph neural network, which can efficiently identify the microstructure in pathological images and realize multi-resolution feature representation, significantly improving the ability to capture and express complex pathological information. (2) This scheme introduces a Bayesian neural network to dynamically quantify the uncertainty of pathological samples, models the uncertainty of the model and the data respectively, and integrates them into a comprehensive score, which effectively improves the robustness and credibility of the diagnostic results; (3) This scheme achieves efficient collaborative expression and knowledge fusion of multi-source heterogeneous data through cross-modal attention fusion, knowledge graph reasoning and incremental learning modules, and presents the diagnostic process transparently in combination with visualization module, thereby enhancing the interpretability and clinical applicability of the system.

[0015] The method of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is a flowchart of an AI-assisted pathological sample diagnosis method according to the present invention; Figure 2 This is a structural diagram of a diagnostic system for an AI-assisted pathological sample diagnosis method according to the present invention. Detailed Implementation

[0017] The method of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] Unless otherwise defined, the methodological or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0019] The terms "comprising" or "including" as used in this invention mean that the element preceding the term encompasses the element listed after the term, and do not exclude the possibility of encompassing other elements. Terms such as "inner," "outer," "upper," and "lower" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. When the absolute position of the described object changes, the relative positional relationship may also change accordingly. In this invention, unless otherwise explicitly specified and limited, the term "attached" and similar terms should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can refer to a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication of two elements or the interaction relationship between two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0020] Example like Figure 1 As shown, this invention provides an AI-assisted method for diagnosing pathological samples, comprising the following steps: S1: Obtain pathological samples from the digital pathology image library and obtain user information corresponding to the pathological samples from the clinical information database through the data access layer; S2: Multi-scale attention feature extraction and representation learning of pathological samples are performed using a multi-scale feature extraction network; Step S2 specifically includes the following steps: S21: The microstructure of pathological samples is identified through an attention mechanism CNN, and multi-resolution pyramid feature extraction is performed to obtain multi-scale feature representations. Multi-scale feature representation Specifically set as follows: ; in, This indicates the input pathological sample. Indicates the number of pyramid levels. Indicates the first The feature transformation function of the layer, Indicates the first The graph neural network modeling function of the layer, Indicates the first Feature extraction function of layer Indicates the first Parameters of the layer feature extraction network, Indicates the first Parameters of the layer feature transformation network; S22: Perform feature weighting for the attention mechanism, and calculate the... Layer Attention weight of each region , No. Layer Attention weight of each region Specifically set as follows: ; in, Represents an exponential function. The weight vector representing the attention mechanism. This represents the hyperbolic tangent activation function. The transformation matrix representing the attention mechanism. Indicates the first Layer Feature vectors of each region This indicates the bias term.

[0021] S3: Quantify the dynamic uncertainty of pathological samples using a Bayesian neural network; Step S3 specifically includes the following steps: S31: Predict pathological samples using a Bayesian neural network and quantify model uncertainty. Specifically set as follows: ; in, This represents the output value of the Bayesian neural network. Represents the parameters of a Bayesian neural network. This indicates the probability of the output value. This represents the model's uncertainty and prediction variance. Indicates the number of samples. Indicates the first The parameters corresponding to the next sampling , express The mean of the predictions; In step S31, the prediction result Specifically set as follows: ; in, Represents the training data set. This represents the posterior probability distribution of the Bayesian neural network.

[0022] S32: Predict pathological samples using a Bayesian neural network and quantify data uncertainty. Specifically set as follows: ; in, This indicates the variance in the prediction of data uncertainty. S33: Calculate the dynamic uncertainty by comprehensively evaluating the uncertainty score. Dynamic uncertainty Specifically set as follows: ; ; in, Indicating model uncertainty The weighting coefficients, Indicating data uncertainty The weighting coefficients.

[0023] S4: Perform multimodal knowledge fusion; Step S4 specifically includes the following steps: S41: By using the cross-modal attention fusion module, pathological samples and corresponding user information are projected into a shared feature space to establish semantic associations between pathological samples and user information, enabling collaborative expression of heterogeneous data. S42: Based on the pathology knowledge base, clinical guidelines and historical research results, a hierarchical knowledge graph is constructed through the knowledge graph reasoning module. The hierarchical knowledge graph includes entity relationships, diagnostic rules and clinical pathways. S43: Store representative samples through an incremental learning module and dynamically evaluate the Bayesian neural network parameters corresponding to the representative samples. The importance of Bayesian neural network parameters based on importance ranking Strengthen constraints and monitor dynamic uncertainties. .

[0024] S5: Visualize the diagnostic decision-making process through the visualization module; Step S5 specifically includes the following steps: S51: Display the lesion areas in the pathological sample using a color heatmap, with lesion areas of high abnormality set to dark color and lesion areas of low abnormality set to light color; S52: Retrieve cases similar to the current case, display the pathological images and diagnostic results of the current case and similar cases side by side, and determine the accuracy of the diagnostic results of the current case; S53: The diagnostic reasoning process is unfolded in the form of a tree diagram, and a complete thinking chain is constructed based on the decision logic of each branch of the tree diagram.

[0025] S6: Evaluate and validate the diagnostic system based on the visualized output results.

[0026] Step S6 specifically includes the following steps: S61: Evaluate the diagnostic accuracy, clinical usability, robustness, and generalization ability of a diagnostic system based on multi-dimensional performance indicators; S62: Set up multiple groups of randomized controlled cases, conduct long-term follow-up verification of multiple groups of randomized controlled cases, and perform cost-benefit analysis; S63: Track user information corresponding to pathological samples in real time, obtain treatment data and prognostic data from user information, compare the prognostic results with the system's diagnostic results, if they match, use the case for positive model reinforcement and self-evolution, if they do not match, automatically screen and issue warnings and drive model error correction and optimization, so as to achieve closed-loop iteration and continuous improvement of AI diagnostic capabilities.

[0027] like Figure 2 As shown, a diagnostic system for an AI-assisted pathological sample diagnosis method includes a data access layer, a multi-scale feature extraction network, a Bayesian neural network, a cross-modal attention fusion module, a knowledge graph reasoning module, an incremental learning module, and a visualization module.

[0028] Therefore, the present invention employs the above-mentioned AI-assisted pathological sample diagnosis method and system, which significantly improves the accuracy, interpretability and clinical applicability of pathological sample diagnosis, while effectively reducing the subjectivity and uncertainty of manual slide reading.

[0029] Finally, it should be noted that the above embodiments are only used to illustrate the method of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the method of the present invention, and these modifications or equivalent substitutions should not cause the modified method to deviate from the spirit and scope of the method of the present invention.

Claims

1. An AI-assisted method for diagnosing pathological samples, characterized in that, Includes the following steps: S1: Obtain pathological samples from the digital pathology image library and obtain user information corresponding to the pathological samples from the clinical information database through the data access layer; S2: Multi-scale attention feature extraction and representation learning of pathological samples are performed using a multi-scale feature extraction network; S3: Quantify the dynamic uncertainty of pathological samples using a Bayesian neural network; S4: Perform multimodal knowledge fusion; S5: Visualize the diagnostic decision-making process through the visualization module; S6: Evaluate and validate the diagnostic system based on the visualized output results.

2. The AI-assisted pathological sample diagnosis method according to claim 1, characterized in that, Step S2 specifically includes the following steps: S21: The microstructure of pathological samples is identified through an attention mechanism CNN, and multi-resolution pyramid feature extraction is performed to obtain multi-scale feature representations. Multi-scale feature representation Specifically set as follows: ； in, This indicates the input pathological sample. Indicates the number of pyramid levels. Indicates the first The feature transformation function of the layer, Indicates the first The graph neural network modeling function of the layer, Indicates the first Feature extraction function of layer Indicates the first Parameters of the layer feature extraction network, Indicates the first Parameters of the layer feature transformation network; S22: Perform feature weighting for the attention mechanism, and calculate the... Layer Attention weight of each region , No. Layer Attention weight of each region Specifically set as follows: ； in, Represents an exponential function. The weight vector representing the attention mechanism. This represents the hyperbolic tangent activation function. The transformation matrix representing the attention mechanism. Indicates the first Layer Feature vectors of each region This indicates the bias term.

3. The AI-assisted pathological sample diagnosis method according to claim 2, characterized in that, Step S3 specifically includes the following steps: S31: Predict pathological samples using a Bayesian neural network and quantify model uncertainty. Specifically set as follows: ； in, This represents the output value of the Bayesian neural network. Represents the parameters of a Bayesian neural network. This indicates the probability of the output value. This represents the model's uncertainty and prediction variance. Indicates the number of samples. Indicates the first The parameters corresponding to the next sampling , express The mean of the predictions; S32: Predict pathological samples using a Bayesian neural network and quantify data uncertainty. Specifically set as follows: ； in, This indicates the variance in the prediction of data uncertainty. S33: Calculate the dynamic uncertainty by comprehensively evaluating the uncertainty score. Dynamic uncertainty Specifically set as follows: ； ； in, Indicating model uncertainty The weighting coefficients, Indicating data uncertainty The weighting coefficients.

4. The AI-assisted pathological sample diagnosis method according to claim 3, characterized in that, In step S31, the prediction result Specifically set as follows: ； in, Represents the training data set. This represents the posterior probability distribution of the Bayesian neural network.

5. The AI-assisted pathological sample diagnosis method according to claim 3, characterized in that, Step S4 specifically includes the following steps: S41: By using the cross-modal attention fusion module, pathological samples and corresponding user information are projected into a shared feature space to establish semantic associations between pathological samples and user information, enabling collaborative expression of heterogeneous data. S42: Based on the pathology knowledge base, clinical guidelines and historical research results, a hierarchical knowledge graph is constructed through the knowledge graph reasoning module. The hierarchical knowledge graph includes entity relationships, diagnostic rules and clinical pathways. S43: Store representative samples through an incremental learning module and dynamically evaluate the Bayesian neural network parameters corresponding to the representative samples. The importance of Bayesian neural network parameters based on importance ranking Strengthen constraints and monitor dynamic uncertainties. .

6. The AI-assisted pathological sample diagnosis method according to claim 1, characterized in that, Step S5 specifically includes the following steps: S51: Display the lesion areas in the pathological sample using a color heatmap, with lesion areas of high abnormality set to dark color and lesion areas of low abnormality set to light color; S52: Retrieve cases similar to the current case, display the pathological images and diagnostic results of the current case and similar cases side by side, and determine the accuracy of the diagnostic results of the current case; S53: The diagnostic reasoning process is unfolded in the form of a tree diagram, and a complete thinking chain is constructed based on the decision logic of each branch of the tree diagram.

7. The AI-assisted pathological sample diagnosis method according to claim 1, characterized in that, Step S6 specifically includes the following steps: S61: Evaluate the diagnostic accuracy, clinical usability, robustness, and generalization ability of a diagnostic system based on multi-dimensional performance indicators; S62: Set up multiple groups of randomized controlled cases, conduct long-term follow-up verification of multiple groups of randomized controlled cases, and perform cost-benefit analysis; S63: Track user information corresponding to pathological samples in real time, obtain treatment data and prognostic diagnosis data from user information, compare the prognostic diagnosis results with the system's diagnosis results, if they are consistent, use the cases for positive model reinforcement and self-evolution, if they are inconsistent, automatically screen and issue warnings and drive model error correction and optimization.

8. A diagnostic system for an AI-assisted pathological sample diagnosis method according to any one of claims 1-7, characterized in that, It includes a data access layer, a multi-scale feature extraction network, a Bayesian neural network, a cross-modal attention fusion module, a knowledge graph reasoning module, an incremental learning module, and a visualization module.

Images