Clinical diagnosis method and device for disease

Abstract

This invention discloses a clinical diagnostic method and apparatus for diseases, belonging to the field of disease diagnosis technology. The method can adaptively adjust the diagnostic path according to the specific circumstances of a case. It assesses the complexity of the condition through an intelligent agent and selects the most appropriate diagnostic procedure accordingly, thereby improving the accuracy and efficiency of diagnosis. By constructing a multi-agent system to simulate expert collaboration across different fields, it can integrate the opinions of multiple virtual experts, ensuring the comprehensiveness and reliability of diagnostic results and reducing the risk of misdiagnosis that may arise from a single model. The diagnostic process strictly follows clinical guidelines, and through multi-agent debate and self-correction procedures, it ensures that each diagnostic step is based on evidence, enhancing the interpretability of diagnostic results. This allows medical professionals to understand and trust the diagnostic output, facilitating its application in real-world clinical settings.

Description

Technical Field

[0001] This invention relates to the field of disease diagnosis technology, and in particular to a clinical diagnostic method and apparatus for diseases based on hybrid guidelines and data-driven adaptation. Background Technology

[0002] Liver disease, a serious global health issue, poses a significant challenge to human health due to its high incidence and mortality rates. Liver diseases are diverse, with complex and variable disease progressions exhibiting significant heterogeneity and dynamic changes, making the timing of diagnosis and treatment a major challenge. The complex clinical manifestations of liver disease require physicians to possess profound professional knowledge and extensive practical experience. In reality, high-level liver disease specialists are relatively scarce and unevenly distributed, making it difficult for many regions to obtain timely and accurate diagnostic services. Furthermore, traditional diagnostic methods often rely on the experience and judgment of individual physicians, making them susceptible to subjective influences and lacking standardization and reproducibility. Therefore, methods for applying intelligent liver disease diagnostic models to clinical liver disease analysis must meet the following conditions:

[0003] 1. Standardized Diagnostic Pathways. The model should follow clinical guidelines and best practices to construct standardized diagnostic pathways, ensuring that each step of the diagnosis has clear basis and procedure, thereby improving the consistency and reliability of diagnosis, while facilitating learning and application by medical personnel.

[0004] 2. Adaptive disease management capability. Given the diversity and complexity of liver diseases, the model needs to possess adaptive disease management capability, meaning it should be able to dynamically adjust diagnostic strategies based on the patient's specific situation. This means it should be able to make appropriate diagnostic responses regardless of the severity or urgency of the disease or the different accompanying symptoms, thereby improving the flexibility and specificity of diagnosis.

[0005] 3. Comprehensive Analysis and Interpretability. To enhance the reliability of diagnostic results, the model must be able to analyze all possible types of liver disease to the greatest extent possible and provide clear diagnostic evidence, including but not limited to etiological analysis, explanations of pathophysiological mechanisms, and rationale for treatment recommendations. It must ensure that every diagnostic conclusion is supported by sufficient evidence and presented to healthcare professionals in an easily understandable manner, making it highly interpretable and practical in actual clinical work.

[0006] Therefore, establishing an objective and reliable diagnostic model for liver diseases to improve the accuracy of comprehensive diagnosis of complex liver diseases is a pressing technical problem that needs to be solved by those skilled in the art, given the seriousness of medical diagnosis. This model must be transparent and interpretable, enabling healthcare professionals to understand and trust the diagnostic process and results for application in clinical practice. Summary of the Invention

[0007] In view of the above problems, the present invention provides a clinical diagnostic method and apparatus for overcoming or at least partially solving the above problems. This method, through in-depth analysis of clinical liver disease data, constructs a model capable of comprehensively analyzing all possible disease types and providing diagnostic criteria, ensuring that the diagnostic results have high interpretability and practicality, thereby improving the accuracy and efficiency of diagnosing complex liver diseases, and enhancing clinicians' confidence in the diagnostic results.

[0008] This invention provides the following solution:

[0009] A clinical diagnostic method for a disease, comprising:

[0010] Obtain a multi-dimensional feature data set related to the patient's disease, wherein the feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations;

[0011] The disease is classified using a classification agent combined with the feature data set to obtain a classification result, which includes any one of simple, intermediate, and complex classification results.

[0012] Based on the classification results, a combination of diagnostic agents is determined to generate diagnostic opinions, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agents include any one or a combination of several of the following: general knowledge agents, digital expert agents, fine-tuning agents, endocrinology expert agents, disease specialist agents, imaging, radiology, and laboratory physician agents.

[0013] The prediction results are analyzed using a prediction result analysis model to obtain the target diagnostic result for the patient's condition;

[0014] The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

[0015] Preferably: the classification agent classifies the illness using a complexity scoring function to obtain the classification result, wherein the complexity scoring function is expressed by the following formula:

[0016] Complexity(E) = f(e1,e2,…,e n )

[0017] In the formula: E = [e1, e2, ..., e n ] represents the feature data set, and f represents the scoring function derived by learning from a large number of case data.

[0018] Preferably, the classification result is represented by the following formula:

[0019]

[0020] In the formula: T1 and T2 represent the thresholds automatically learned from historical case data.

[0021] Preferably, the classification result includes simple, and the diagnostic agent combination includes the general knowledge agent;

[0022] The classification results include intermediate level, and the diagnostic agent combination includes the digital expert agent, the general knowledge agent, and the fine-tuning agent;

[0023] The classification results include complexity, and the diagnostic agent combination includes the general knowledge agent, the digital expert agent, the fine-tuning agent, the endocrinology expert agent, the disease specialist agent, and the imaging, radiology, and laboratory physician agents.

[0024] Preferably, the disease includes liver disease. By analyzing the liver disease diagnosis and treatment guidelines in conjunction with historical liver disease datasets, the logical thinking used in the diagnosis process is transformed into a digital reasoning structure, thereby constructing and obtaining the digital expert intelligent agent.

[0025] The digital expert agent is used to initially determine the category of liver disease based on the duration of the disease; based on the specific disease category, it assesses changes in the patient's liver function and calculates classification indicators.

[0026] The following formula is used to determine the likelihood of hepatocellular damage and cholestasis:

[0027]

[0028] In the formula: ALT represents the content of alanine aminotransferase in serum;

[0029] During the self-optimization process of the logic chain, the digital expert agent optimizes based on new case data D. new The logic chain is dynamically adjusted; this process can be represented by the following formula:

[0030] L optimized =L previous +ΔL(D new )

[0031] In the formula, ΔL represents the adjustment amount based on the new data.

[0032] Preferably, the prediction result analysis model further includes a cue vector and noise perturbation; the cue vector includes a mask vector generated based on historical data, and the expression after combining it with the input matrix Z is:

[0033] Z′=Z+αp+βn

[0034] Where n represents a noise vector that follows a Gaussian distribution, and α and β are weights for adjusting the intensity of the cue and the noise.

[0035] Preferably, the prediction result analysis model further includes a secondary learner based on an autoencoder, wherein the secondary learner extracts the latent representation of the prediction value matrix through encoding and decoding processes to obtain a refined comprehensive output;

[0036] The autoencoder is used to compress the input matrix into a low-dimensional representation H, as expressed by the following equation:

[0037]

[0038] In the formula: W e It is the encoder's weight matrix, b e σ is the bias term, and σ(·) is the activation function, where σ represents the ReLU activation function.

[0039] Preferably, the prediction result analysis model further includes a decoder, which is used to restore the low-dimensional representation H and generate a preliminary diagnostic result, expressed by the following formula:

[0040]

[0041] In the formula: W d Let b represent the weight matrix of the decoder. d This indicates the bias term.

[0042] Preferably, the prediction result analysis model further includes an integrated intelligent agent. E The integrated intelligent agent E Used to integrate the preliminary diagnostic results The preliminary diagnosis was verified against the patient's medical record data X and in accordance with authoritative liver disease guidelines G. The rationality of this is determined in order to obtain the final target diagnostic result.

[0043] A clinical diagnostic device for diseases, used to perform the above-described clinical diagnostic method for diseases, the device comprising:

[0044] The feature data set acquisition unit is used to acquire a multi-dimensional feature data set related to the patient's disease. The feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations.

[0045] The disease classification unit is used to classify the disease by combining the feature data set with a classification agent to obtain a classification result, which includes any one of simple, intermediate and complex.

[0046] The diagnostic agent prediction unit is used to determine a combination of diagnostic agents to generate diagnostic opinions based on the classification results, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agent includes any one or a combination of several of the following: general knowledge agent, digital expert agent, fine-tuning agent, endocrinology expert agent, disease specialist agent, imaging, radiology, and laboratory physician agent.

[0047] The target diagnosis result analysis unit is used to analyze the prediction results using the prediction result analysis model to obtain the target diagnosis result of the patient's condition;

[0048] The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

[0049] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0050] This application provides a clinical diagnostic method and apparatus for diseases. The method adaptively adjusts the diagnostic path based on the specific circumstances of each case. It assesses the complexity of the condition through an intelligent agent and selects the most appropriate diagnostic procedure accordingly, thereby improving the accuracy and efficiency of diagnosis. By constructing a multi-agent system to simulate collaboration among experts from different fields, the opinions of multiple virtual experts can be integrated, ensuring the comprehensiveness and reliability of the diagnostic results and reducing the risk of misdiagnosis that may arise from a single model. The diagnostic process strictly follows clinical guidelines, and multi-agent debate and self-correction procedures ensure that each diagnostic step is based on evidence, enhancing the interpretability of the diagnostic results. This allows medical professionals to understand and trust the diagnostic output, facilitating its application in real-world clinical settings.

[0051] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0053] Figure 1 This is a flowchart of a clinical diagnostic method for diseases provided in an embodiment of the present invention;

[0054] Figure 2 This is a flowchart illustrating the clinical diagnostic method for liver diseases provided in this embodiment of the invention.

[0055] Figure 3 This is a schematic diagram of the clinical diagnostic device for diseases provided in an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of a clinical diagnostic device for diseases provided in an embodiment of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0058] See Figure 1 This invention provides a clinical diagnostic method for a disease, such as... Figure 1 As shown, the method may include:

[0059] S101: Obtain a multi-dimensional feature data set related to the patient's disease, wherein the feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations;

[0060] S102: The classification agent is used in conjunction with the feature data set to classify the illness and obtain a classification result, wherein the classification result includes any one of simple, intermediate, and complex; further, in this embodiment, the classification agent can be provided to classify the illness using a complexity scoring function to obtain a classification result, wherein the complexity scoring function is expressed by the following formula:

[0061] Complexity(E) = f(e1,e2,…,e n )

[0062] In the formula: E = [e1, e2, ..., e n ] represents the feature data set, and f represents the scoring function derived by learning from a large number of case data.

[0063] The classification result is expressed by the following formula:

[0064]

[0065] In the formula: T1 and T2 represent the thresholds automatically learned from historical case data.

[0066] S103: Determine a combination of diagnostic agents to generate diagnostic opinions based on the classification results, so that the combination of diagnostic agents can output a prediction result for the patient's condition; the diagnostic agents include any one or a combination of several of the following: general knowledge agents, digital expert agents, fine-tuning agents, endocrinology expert agents, disease specialist agents, imaging, radiology, and laboratory physician agents.

[0067] In specific implementation, the embodiments of this application may provide that the classification result includes simplicity, and the diagnostic agent combination includes the general knowledge agent;

[0068] The classification results include intermediate level, and the diagnostic agent combination includes the digital expert agent, the general knowledge agent, and the fine-tuning agent;

[0069] The classification results include complexity, and the diagnostic agent combination includes the general knowledge agent, the digital expert agent, the fine-tuning agent, the endocrinology expert agent, the disease specialist agent, and the imaging, radiology, and laboratory physician agents.

[0070] S104: Analyze the prediction results using the prediction result analysis model to obtain the target diagnostic result of the patient's condition;

[0071] The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

[0072] In practical applications, each intelligent agent can be pre-constructed according to the type of disease actually being diagnosed. The construction method can adopt the existing intelligent agent construction method. For example, in one implementation, the embodiment of this application can provide that the disease includes liver disease. By combining historical liver disease datasets and parsing liver disease diagnosis and treatment guidelines, the logical thinking used in the diagnosis process is transformed into a digital reasoning structure, thereby constructing and obtaining the digital expert intelligent agent.

[0073] The digital expert agent is used to initially determine the category of liver disease based on the duration of the disease; based on the specific disease category, it assesses changes in the patient's liver function and calculates classification indicators.

[0074] The following formula is used to determine the likelihood of hepatocellular damage and cholestasis:

[0075]

[0076] In the formula: ALT represents the content of alanine aminotransferase in serum;

[0077] During the self-optimization process of the logic chain, the digital expert agent optimizes based on new case data D. new The logic chain is dynamically adjusted; this process can be represented by the following formula:

[0078] L optimized =L previous +ΔL(D new )

[0079] In the formula, ΔL represents the adjustment amount based on the new data.

[0080] To further improve the accuracy of diagnosing moderate and complex medical records, embodiments of this application may also provide that the prediction result analysis model further includes a cue vector and noise perturbation; the cue vector includes a mask vector generated based on historical data, and the expression after combining it with the input matrix Z is:

[0081] Z′=Z+αp+βn

[0082] Where n represents a noise vector that follows a Gaussian distribution, and α and β are weights for adjusting the intensity of the cue and the noise.

[0083] The prediction result analysis model also includes a secondary learner based on an autoencoder. The secondary learner extracts the latent representation of the prediction value matrix through encoding and decoding processes to obtain a refined comprehensive output.

[0084] The autoencoder is used to compress the input matrix into a low-dimensional representation H, as expressed by the following equation:

[0085]

[0086] In the formula: W e It is the encoder's weight matrix, b e σ is the bias term, and σ(·) is the activation function, where σ represents the ReLU activation function.

[0087] The prediction result analysis model also includes a decoder, which is used to restore the low-dimensional representation H and generate a preliminary diagnostic result, expressed by the following formula:

[0088]

[0089] In the formula: W d Let b represent the weight matrix of the decoder. d This indicates the bias term.

[0090] The prediction result analysis model also includes an integrated intelligent agent. E The integrated intelligent agent E Used to integrate the preliminary diagnostic results The preliminary diagnosis was verified against the patient's medical record data X and in accordance with authoritative liver disease guidelines G. The rationality of this is determined in order to obtain the final target diagnostic result.

[0091] The clinical diagnostic method for diseases provided in this application improves the accuracy and reliability of diagnosis by combining ensemble learning technology with large model generation technology and employing a digital clinical diagnostic pathway and dynamic collaboration among hybrid experts. Specifically, the method first automates the analysis of the hierarchical decomposition of disease guidelines, transforming the logical thinking used in the diagnostic process into a digital reasoning structure; second, it adaptively classifies cases according to the severity of the condition; and finally, it establishes a personalized diagnostic process through dynamic collaboration among hybrid experts and learns normalization processing through a self-supervised ensemble learning strategy, thereby generating a final comprehensive diagnostic result.

[0092] The following section provides a detailed introduction to the clinical diagnostic methods for liver diseases provided in this application, using their application in liver disease diagnosis as an example.

[0093] This application proposes an adaptive intelligent diagnostic model for liver diseases, Hepat iGLM, which achieves the function of matching the optimal diagnostic path based on the complexity of medical records, significantly improving the accuracy and efficiency of diagnosis. By constructing a digital expert agent and combining adaptive disease classification with hybrid expert collaborative analysis, Hepat iGLM simulates the clinical diagnostic process, standardizes diagnostic steps, and, through a hybrid expert autoencoder ensemble learning strategy, mines the semantic relationships between labels, enhancing the model's generalization performance on unlabeled data. This model, by combining traditional machine learning with a large language model, achieves more stable diagnostic results, especially in handling complex or uncertain cases, effectively reducing bias and noise in the diagnostic process, improving the reliability of diagnostic results and the model's generalization ability.

[0094] In practice:

[0095] An adaptive clinical diagnostic method for liver diseases based on hybrid guidelines and data-driven approaches is proposed. This method integrates standardized diagnostic and treatment logic with deep learning technology and adopts adaptive case classification and multi-agent collaborative strategies, aiming to improve the accuracy and consistency of diagnosis.

[0096] Specifically, this method first constructs a digital expert agent by automatically parsing liver disease diagnosis and treatment guidelines, thereby digitizing the standard diagnostic logic;

[0097] Next, the cases are dynamically classified based on a data-driven disease feature learning algorithm;

[0098] Finally, through collaborative analysis and ensemble learning strategies among hybrid expert agents, the diagnostic results are ensured to be validated in multiple dimensions and to meet clinical diagnostic and treatment standards, thereby yielding reliable and personalized final diagnostic recommendations.

[0099] like Figure 2 As shown, this adaptive clinical diagnostic approach for liver disease based on a hybrid guideline and data-driven approach includes the following steps:

[0100] Step 1: First, read in the electronic medical record dataset containing patient medical history, imaging examinations, biochemical tests, physical examination information, etc. This data is then segmented and statistically analyzed for subsequent processing. Specifically, given a patient's electronic medical record, the dataset is cleaned and integrated from the patient's medical history, chief complaints, imaging reports, physical examination results, and laboratory test information to form a summary report, thus constructing the liver disease dataset.

[0101] Step Two: Construction of the Digital Expert Agent. The method involves automatically parsing liver disease diagnosis and treatment guidelines, transforming the logical thinking within into a digital reasoning structure, extracting key diagnostic features, and forming logical chains to guide the decision-making process. This stage also includes generating preliminary logical chains and continuous optimization and iteration using extensive clinical data, enabling the agent to adapt to varying levels of disease complexity, thereby improving diagnostic accuracy and coverage. Ultimately, a digital expert agent integrating standardized diagnostic and treatment logic is constructed.

[0102] In practice, by automatically analyzing liver disease diagnosis and treatment guidelines, the logical thinking used in the diagnosis process is transformed into a digital reasoning structure, thereby constructing a digital expert intelligent agent.

[0103] In the initial generation of the logical chain, the disease category is initially determined based on the duration of the liver disease course. Next, according to the specific disease category, such as space-occupying or diffuse liver disease, the agent automatically assesses the patient's liver function changes and calculates classification indicators, using the following formula to determine the likelihood of hepatocellular damage and cholestasis.

[0104]

[0105] When the classification value is greater than 5, elevated ALT levels are the primary indicator, suggesting liver cell damage. The expert AI will automatically conduct more detailed examinations, such as viral infections, immune system abnormalities, and metal metabolism disorders. Conversely, when the classification value is less than 5, it suggests cholestasis or biliary tract disease. The AI ​​will then guide further imaging examinations (such as ultrasound and CT scans) and make corresponding diagnostic inferences based on the results.

[0106] During the self-optimization process of the logic chain, the agent optimizes based on the new case data D. new The logical chain is dynamically adjusted. This process can be represented as:

[0107] L optimized =L previous +ΔL(D new )

[0108] Where ΔL represents the adjustment amount based on the new data.

[0109] Step 3: Adaptive Disease Classification. The method uses a specially designed agent to initially assess the complexity of cases. This agent mimics the role of a junior physician, analyzing multi-dimensional characteristic data of patients, such as medical history, clinical symptoms, laboratory test results, and imaging examinations. Based on these characteristics, it establishes a self-iterative complexity scoring function to automatically classify the disease into three levels: simple, moderate, and complex, thereby guiding the subsequent diagnostic process.

[0110] In practice, an intelligent agent is designed to analyze multi-dimensional patient data and establish a self-iterative complexity scoring function to automatically classify the condition into three levels to guide the diagnostic process.

[0111] The model first collects multi-dimensional patient characteristic data, including medical history, clinical symptoms, laboratory test results, and imaging examinations. These features can be represented by a feature set E = [e1, e2, ..., e...]. n ], where n is the total number of features.

[0112] The classification agent establishes a complexity scoring function by analyzing combinations of different features:

[0113] Complexity(E) = f(e1,e2,…,e n )

[0114] Here, f is a scoring function derived from learning from a large amount of case data, which aims to comprehensively consider the impact of various features on the complexity of the disease.

[0115] Ultimately, the classification agent categorizes the illness into three levels: simple, moderate, and complex, using the following formula:

[0116]

[0117] T1 and T2 are thresholds automatically learned from historical case data, used to distinguish between conditions of different complexities.

[0118] Step 4: Hybrid Expert Collaborative Analysis. In hybrid expert collaborative analysis, appropriate combinations of expert agents are intelligently selected based on the complexity of different cases. Simple cases are quickly diagnosed by general knowledge agents, moderate cases are analyzed jointly by digital experts, general knowledge agents, and fine-tuned agents, while complex cases introduce more agents specializing in handling multiple abnormal indicators, complex medical histories, and inconsistent test results. Finally, an autoencoder ensemble learning strategy is used to synthesize multiple diagnostic opinions to arrive at a final diagnostic conclusion.

[0119] In practice, different combinations of expert agents are intelligently selected based on the complexity of the case, and the diagnostic opinions are synthesized through an autoencoder ensemble learning strategy to arrive at the final diagnostic conclusion.

[0120] The diagnosis of liver diseases is often highly complex due to the diversity of disease course, symptoms, and examination results. Therefore, different agent collaboration schemes were adopted for cases X of varying difficulty. The pre-diagnosis result `pre` generated by each agent can be represented as:

[0121] pre = Agent(X)

[0122] For simple medical records with clearly defined characteristics, embodiments of this application provide a general knowledge agent. G This method can quickly provide accurate diagnostic results through a series of simple prompts. It can quickly handle most routine cases, saving diagnostic time and resources and avoiding lengthy analysis processes.

[0123] For moderate cases with ambiguous symptoms, embodiments of this application provide a digital expert intelligent agent. D General knowledge agent G and fine-tuning the agent F A diagnostic framework based on joint analysis.

[0124] Among them, digital expert intelligent agent D It simulates the reasoning process of clinical experts and has a strong ability to interpret medical knowledge. It is responsible for conducting in-depth analysis of difficult symptoms based on the latest medical standards to ensure that the diagnosis conforms to clinical norms.

[0125] General knowledge agent G It retains the original knowledge system within the large language model, and is responsible for providing basic opinions to ensure that fundamental issues are not overlooked and to compensate for the biases that may result from overly specialized diagnoses.

[0126] Fine-tuning the agent F After fine-tuning based on clinical data related to liver disease, it has extremely high diagnostic specialization for liver disease. It can more sensitively capture abnormalities related to liver disease and reduce the possibility of misdiagnosis or overlooking key indicators.

[0127] For complex cases, embodiments of this application provide a multi-agent collaboration mechanism. In addition to the three agents used in moderate cases, embodiments of this application, taking into account the characteristics of complex liver diseases, additionally add an agent. M Agent Co and Agent I They address the challenges of multiple abnormal indicators, complex symptoms and medical history, and inconsistent test results.

[0128] Agent M Simulates experts in the field of endocrinology, specializing in addressing abnormalities in multiple biomarkers and laboratory indicators, and combining the dynamic change patterns among these indicators to identify potential systemic diseases or multi-system complications.

[0129] Agent Co Simulate a liver specialist to extract key information from long-term medical history and combine it with the development of current symptoms to provide diagnostic support for potential chronic diseases.

[0130] Agent I This system simulates the work of radiologists, radiologists, and laboratory physicians, performing consistency checks and comprehensive analysis on data from various sources, including imaging and laboratory tests. It bridges discrepancies between data points, provides a more comprehensive interpretation, and avoids misdiagnosis caused by a single piece of data.

[0131] To improve the accuracy and robustness of the prediction result analysis model in the diagnosis of moderate and complex cases, this application constructs a self-supervised secondary learner and designs an autoencoder ensemble learning strategy to comprehensively analyze the prediction results of multiple agents. In the ensemble framework, the outputs of multiple agent models are regarded as the prediction values ​​of the base learner, forming the input matrix:

[0132] Z ij =pre ij

[0133] Among them, Z ij This represents the prediction value of the i-th agent for the j-th sample.

[0134] To improve the robustness of the prediction results, this embodiment introduces a cue vector p and a noise perturbation n into the model. The cue vector is a mask vector generated based on historical data, and its expression after combining with the input matrix Z is:

[0135] Z′=Z+αp+βn

[0136] Here, n is a noise vector following a Gaussian distribution, and α and β are weights for adjusting the intensity of the cue and noise. Through this adjustment, the model can maintain greater stability in the face of data fluctuations.

[0137] To comprehensively normalize these predictions, this application employs an autoencoder-based secondary learner. Through encoding and decoding processes, it extracts the latent representation of the prediction matrix, yielding a refined comprehensive output. The encoder compresses the input matrix into a low-dimensional representation H, which can be expressed as:

[0138] H=σ(W e Z′+b e )

[0139] Among them, W e It is the encoder's weight matrix, b e σ is the bias term, and σ(·) is the activation function.

[0140] The decoder reconstructs the latent representation H and generates the final prediction output:

[0141]

[0142] Among them, W d These are the weight matrices of the decoder, b d It is a bias term.

[0143] Based on the ensemble learning framework, this application embodiment designs an integrated intelligent agent. E Its core task is to integrate diagnostic conclusions. The diagnosis was verified against the patient's medical record data (X) and the authoritative liver disease guideline (G). The rationality of the decision is determined to arrive at the final decision result.

[0144] In summary, the clinical diagnostic method for diseases provided in this application can adaptively adjust the diagnostic path according to the specific circumstances of the case. It assesses the complexity of the condition through an intelligent agent and selects the most appropriate diagnostic procedure accordingly, thereby improving the accuracy and efficiency of diagnosis. By constructing a multi-agent system to simulate expert collaboration across different fields, it can integrate the opinions of multiple virtual experts, ensuring the comprehensiveness and reliability of the diagnostic results and reducing the risk of misdiagnosis that may arise from a single model. The diagnostic process strictly follows clinical guidelines, and through multi-agent debate and self-correction procedures, it ensures that each diagnostic step is based on evidence, enhancing the interpretability of the diagnostic results. This allows medical professionals to understand and trust the diagnostic output, facilitating its application in real-world clinical settings.

[0145] See Figure 3 This application embodiment can also provide a clinical diagnostic device for diseases, such as... Figure 3 As shown, the device for performing the above-described clinical diagnostic method for the disease may include:

[0146] The feature data set acquisition unit 301 is used to acquire a multi-dimensional feature data set related to the patient's disease, wherein the feature data set includes at least medical history, clinical symptoms, laboratory test results and imaging examinations.

[0147] The disease classification unit 302 is used to classify the disease by combining the feature data set with a classification agent to obtain a classification result, wherein the classification result includes any one of simple, intermediate and complex.

[0148] The diagnostic agent prediction unit 303 is used to determine a combination of diagnostic agents to generate diagnostic opinions based on the classification results, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agent includes any one or a combination of several of the following: general knowledge agent, digital expert agent, fine-tuning agent, endocrinology expert agent, disease specialist agent, imaging, radiology, and laboratory physician agent.

[0149] The target diagnosis result analysis unit 304 is used to analyze the prediction results using the prediction result analysis model to obtain the target diagnosis result of the patient's condition;

[0150] The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

[0151] This application embodiment can also provide a clinical diagnostic device for a disease, the device including a processor and a memory:

[0152] The memory is used to store program code and transmit the program code to the processor;

[0153] The processor is used to execute the steps of the above-described clinical diagnostic method for diseases according to the instructions in the program code.

[0154] like Figure 4 As shown in the embodiment of this application, a clinical diagnostic device for diseases may include: a processor 10, a memory 11, a communication interface 12, and a communication bus 13. The processor 10, the memory 11, and the communication interface 12 all communicate with each other through the communication bus 13.

[0155] In the embodiments of this application, the processor 10 may be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic devices.

[0156] The processor 10 can call programs stored in the memory 11. Specifically, the processor 10 can execute operations in the embodiments of the clinical diagnostic method for diseases.

[0157] The memory 11 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment, the memory 11 stores at least a program for implementing the following functions:

[0158] Obtain a multi-dimensional feature data set related to the patient's disease, wherein the feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations;

[0159] The disease is classified using a classification agent combined with the feature data set to obtain a classification result, which includes any one of simple, intermediate, and complex classification results.

[0160] Based on the classification results, a combination of diagnostic agents is determined to generate diagnostic opinions, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agents include any one or a combination of several of the following: general knowledge agents, digital expert agents, fine-tuning agents, endocrinology expert agents, disease specialist agents, imaging, radiology, and laboratory physician agents.

[0161] The prediction results are analyzed using a prediction result analysis model to obtain the target diagnostic result for the patient's condition;

[0162] The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

[0163] In one possible implementation, the memory 11 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function (such as file creation or data read / write). The data storage area may store data created during use, such as initialization data.

[0164] In addition, memory 11 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0165] Communication interface 12 can be an interface for the communication module, used to connect with other devices or systems.

[0166] Of course, it should be noted that, Figure 4 The structure shown does not constitute a limitation on the clinical diagnostic device for diseases in the embodiments of this application. In practical applications, the clinical diagnostic device for diseases may include more than Figure 4 More or fewer components as shown, or combinations of certain components.

[0167] This application embodiment may also provide a computer-readable storage medium for storing program code for performing the steps of the above-described clinical diagnostic method for diseases.

[0168] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0169] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0170] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0171] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A clinical diagnostic method for a disease, characterized in that, include: Obtain a multi-dimensional feature data set related to the patient's disease, wherein the feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations; The disease is classified using a classification agent combined with the feature data set to obtain a classification result, which includes any one of simple, intermediate, and complex classification results. Based on the classification results, a combination of diagnostic agents is determined to generate diagnostic opinions, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agents include any one or a combination of several of the following: general knowledge agents, digital expert agents, fine-tuning agents, endocrinology expert agents, disease specialist agents, imaging, radiology, and laboratory physician agents. The prediction results are analyzed using a prediction result analysis model to obtain the target diagnostic result for the patient's condition; The prediction result analysis model includes a self-supervised secondary learner and an autoencoder ensemble learning strategy for comprehensively analyzing the prediction results output by each agent. The diseases mentioned include liver diseases. By combining historical liver disease datasets and analyzing liver disease diagnosis and treatment guidelines, the logical thinking used in the diagnosis process is transformed into a digital reasoning structure, thereby constructing and obtaining the digital expert intelligent agent. The digital expert agent is used to initially determine the category of liver disease based on the duration of the disease; based on the specific disease category, it assesses changes in the patient's liver function and calculates classification indicators. The following formula is used to determine the likelihood of hepatocellular damage and cholestasis: In the formula: This indicates the level of alanine aminotransferase in serum; During the self-optimization process of the logic chain, the digital expert agent adjusts its logic based on new case data. The logic chain is dynamically adjusted; this process can be represented by the following formula: In the formula, This indicates the amount of adjustment based on the new data; The prediction result analysis model also includes a cue vector and noise perturbation; the cue vector includes a mask vector generated based on historical data, and an input matrix. The combined expression is: in, This represents a noise vector that follows a Gaussian distribution. and It adjusts the weights of the prompts and noise intensity.

2. The clinical diagnostic method for a disease according to claim 1, characterized in that, The classification agent obtains the classification result by classifying the illness using a complexity scoring function, which is expressed by the following formula: In the formula: Represents a set of feature data. This represents a scoring function derived by learning from a large amount of case data.

3. The clinical diagnostic method for a disease according to claim 2, characterized in that, The classification result is expressed by the following formula: In the formula: and This indicates a threshold that is automatically learned based on historical case data.

4. The clinical diagnostic method for a disease according to claim 1, characterized in that, The classification result includes simple, and the diagnostic agent combination includes the general knowledge agent; The classification results include intermediate level, and the diagnostic agent combination includes the digital expert agent, the general knowledge agent, and the fine-tuning agent; The classification results include complexity, and the diagnostic agent combination includes the general knowledge agent, the digital expert agent, the fine-tuning agent, the endocrinology expert agent, the disease specialist agent, and the imaging, radiology, and laboratory physician agents.

5. The clinical diagnostic method for a disease according to claim 1, characterized in that, The prediction result analysis model also includes a secondary learner based on an autoencoder. The secondary learner extracts the latent representation of the prediction value matrix through encoding and decoding processes to obtain a refined comprehensive output. The autoencoder is used to compress the input matrix into a low-dimensional representation. It is expressed by the following formula: In the formula: It is the encoder's weight matrix. It is a bias term. For activation function, This represents the activation function ReLU.

6. The clinical diagnostic method for a disease according to claim 5, characterized in that, The prediction result analysis model also includes a decoder, which is used to process the low-dimensional representation. The preliminary diagnostic results are restored and generated, as expressed by the following formula: In the formula: This represents the weight matrix of the decoder. This indicates the bias term.

7. The clinical diagnostic method for a disease according to claim 6, characterized in that, The prediction result analysis model also includes an integrated intelligent agent. The integrated intelligent agent Used to integrate the preliminary diagnostic results With patient medical record data Based on authoritative liver disease guidelines Verify the preliminary diagnostic conclusion The rationality of this is determined in order to obtain the final target diagnostic result. .

8. A clinical diagnostic device for a disease, characterized in that, The apparatus for performing the clinical diagnostic method for a disease according to any one of claims 1-7, the apparatus comprising: The feature data set acquisition unit is used to acquire a multi-dimensional feature data set related to the patient's disease. The feature data set includes at least medical history, clinical symptoms, laboratory test results, and imaging examinations. The disease classification unit is used to classify the disease by combining the feature data set with a classification agent to obtain a classification result, which includes any one of simple, intermediate and complex. The diagnostic agent prediction unit is used to determine a combination of diagnostic agents to generate diagnostic opinions based on the classification results, so that the combination of diagnostic agents can output a diagnostic prediction result for the patient's condition; the diagnostic agent includes any one or a combination of several of the following: general knowledge agent, digital expert agent, fine-tuning agent, endocrinology expert agent, disease specialist agent, imaging, radiology, and laboratory physician agent. The target diagnosis result analysis unit is used to analyze the prediction results using the prediction result analysis model to obtain the target diagnosis result of the patient's condition; The prediction result analysis model includes a self-supervised secondary learner for comprehensively analyzing the prediction results output by each agent and an autoencoder ensemble learning strategy.

Images