Structured judgment-based multi-stage medical decision method, system, and storage medium

Abstract

The application provides a multi-stage medical decision-making method and system based on structured judgment and a storage medium, and aims to solve the problems that the existing medical decision-making support system is difficult to effectively utilize expert experience, the decision-making ability in gray area scenarios is insufficient, and there is a lack of safety verification and explainable mechanism. The method comprises the following steps: converting expert judgment information into a structured judgment node; performing a multi-stage medical decision-making process, verifying the integrity and timeliness of patient data, generating explainable decision-making basis, and iteratively updating node weights and versions according to clinical adoption. The application structures the implicit knowledge of experts, realizes safe, explainable, traceable and iteratively optimized clinical decision-making support in the gray area scenarios where evidence-based medical guidelines are not applicable or conflict.

Description

Technical Field

[0001] This invention relates to the field of medical information processing and clinical decision support technology, specifically to a multi-stage medical decision-making method, system, and storage medium based on structured judgment. Background Technology

[0002] In modern clinical practice, for complex comorbidities, individualized treatment, and other clinical gray areas where evidence-based medicine guidelines are not clearly defined or where there are conflicting provisions, decisions are usually made based on the experience and judgment of senior medical experts. However, expert experience is often tacit knowledge, which is difficult to systematically organize and reuse.

[0003] Existing medical decision support systems suffer from the following main shortcomings: rule-based systems based on evidence-based medicine guidelines are only applicable to standardized diagnosis and treatment scenarios and cannot provide effective decision support in gray-area scenarios; data-driven systems based on artificial intelligence operate in a black-box mode, lacking interpretability and traceability of decision logic, and making it difficult to clearly define the boundaries of decision responsibility; general decision frameworks can only optimize the execution of existing processes, cannot proactively identify failure scenarios in standard processes, and have not established a mechanism to transform unstructured expert experience into computer-executable, verifiable, and traceable structured decision-making. These problems result in insufficient decision security, interpretability, and knowledge iteration capabilities in clinical gray-area scenarios.

[0004] A patent search revealed an invention patent with publication number CN121011339A, which discloses a medical auxiliary diagnosis method and system based on temporal and semantic weighting, belonging to the field of medical information processing technology. The method involves constructing patient medical record data from pre-processed current and historical medical record information, inputting it into a pre-trained medical language model to generate a preliminary diagnosis. The patient medical record data vector sequence is then weighted using a time decay function and content relevance weighting to obtain a fused medical record vector, which is concatenated with the preliminary diagnosis to form a query vector. Based on the query vector, candidate segments are retrieved from a clinical guideline knowledge base. Semantic evidence scores and coverage scores are calculated based on the patient medical record data and the tokens of the candidate segments. Combining the prior scores of the candidate segments' metadata, the ranking probability of the candidate segments is obtained, and target guideline segments are selected. The preliminary diagnosis and target guideline segments are then fused to generate the final auxiliary diagnosis and treatment information. However, this patent lacks data integrity and timeliness verification, resulting in insufficient decision security. It also fails to generate a traceable causal reasoning chain, leading to weak interpretability. Furthermore, it lacks a node weight iteration and version update mechanism, hindering closed-loop optimization of knowledge.

[0005] In summary, given the lack of a technical mechanism in the existing systems to transform expert experience into a computable data structure and participate in the automated decision-making process, there is an urgent need for a multi-stage medical decision-making method, system, and storage medium that can transform expert experience into a computable data structure and participate in the automated decision-making process. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a multi-stage medical decision-making method, system, and storage medium based on structured judgment.

[0007] A multi-stage medical decision-making method based on structured judgment, provided by the present invention, includes the following steps:

[0008] Step S1: Collect medical expert judgment information, transform the medical expert judgment information into structured judgment node data objects containing field structures and store them in the knowledge base. Each structured judgment node includes disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action, initial judgment weight, node ID, node version number, data time window, and decision priority.

[0009] Step S2: Obtain patient data, which includes disease diagnosis codes and physiological indicators. Match the physiological indicators with preset evidence-based medicine guidelines. If the matching fails, it is determined to be a gray zone decision scenario, and proceed to step S3; otherwise, terminate the process. Step S3: Read the disease diagnosis code, retrieve the structured judgment nodes in the knowledge base that match the disease diagnosis code, and record the successfully matched structured judgment nodes as candidate nodes; for each candidate node, verify whether the physiological indicators meet the applicable boundary conditions of the candidate node, construct the candidate node set of the candidate nodes that pass the verification, and proceed to step S4; if all candidate nodes fail the verification, output a warning message and terminate the process. Step S4: Obtain the logical judgment threshold of each candidate node in the candidate node set and compare it with the physiological indicators one by one; record the candidate nodes whose physiological indicators meet the logical judgment threshold as target nodes, and generate a structured causal reasoning chain based on the content of the target nodes; if the physiological indicators do not meet all logical judgment thresholds, terminate the process. Step S5: Read the causal reasoning chain, generate decision information according to the preset template format, and output it to the user interface.

[0010] Preferably, in step S1, the structured judgment node further includes a node ID and a version number; the disease diagnosis code is the International Classification of Diseases (ICD) code, used to identify the disease type to which the structured judgment node is applicable; the clinical variables are used to describe the patient's current status indicators; the applicable boundary conditions are the timeliness and completeness conditions of the clinical variables, including the longest data collection time and the list of mandatory variables; the logical judgment threshold is the physiological indicator threshold range of the clinical variables; the output action is used to define recommended or avoided treatment measures; the judgment weight is used to represent the priority of the structured judgment node; each structured judgment node is stored in the form of a data record.

[0011] Preferably, in step S2, the evidence-based medicine guideline clauses corresponding to each clinical scenario are pre-converted into machine-readable logical expressions, and the applicable indicator ranges of each evidence-based medicine guideline clause are stored in the rule engine. During matching, the physiological indicators are compared one by one with the logical expressions in the rule engine: If a matching evidence-based medicine guideline clause exists, the number of matched clauses is further determined: if only a single evidence-based medicine guideline clause is matched, the single evidence-based medicine guideline clause is executed and the process ends; if multiple evidence-based medicine guideline clauses are matched, the logical consistency of the multiple evidence-based medicine guideline clauses is checked: if there is no contradiction between the clauses, one of them is selected for execution and the process ends; if there is a contradiction between the clauses, it is determined to be a gray zone decision scenario and proceeds to step S3. If no matching evidence-based medicine guidelines are found, the scenario is classified as a gray zone decision-making scenario, and the process proceeds to step S3.

[0012] Preferably, step S3 includes the following sub-steps: Step S3.1: Read the disease diagnosis code, and use the disease diagnosis code as the query condition to retrieve all structured judgment nodes in the knowledge base that match the disease diagnosis code. Record the successfully matched structured judgment nodes as candidate nodes. Step S3.2: Obtain the applicable boundary conditions for each candidate node and verify them one by one with the physiological indicators of the patient data. The verification includes: verification of the completeness of required variables and verification of the timeliness of variable data. Step S3.3: If all candidate nodes fail the verification, output a warning message and terminate the process; if there is at least one candidate node that passes the verification, construct a candidate node set from the candidate nodes that pass the verification and execute step S4.

[0013] Preferably, in step S4, the causal reasoning chain is formatted as follows: [clinical variable] satisfies [logical judgment threshold]; trigger expert judgment node [node ID]; execute [output decision action].

[0014] Preferably, in step S4, if there are multiple target nodes, all target nodes are sorted from high to low according to their judgment weights to form a priority recommendation list. Based on the content of the first node in the priority recommendation list, a causal inference chain is generated.

[0015] Preferably, the multi-stage medical decision-making approach also includes: Step S6: Calculate the adoption rate and clinical outcome indicators of the target nodes corresponding to the causal inference chain, adjust the judgment weight of the target nodes and update the version number based on the statistical results.

[0016] Preferably, in step S6, the adoption rate is the actual number of adoptions divided by the total number of recommendations. By periodically calculating the clinical outcome indicators of the target node, if the adoption rate is lower than the preset adoption threshold, or the clinical outcome indicators do not reach the preset indicator threshold, the judgment weight of the target node is multiplied by the preset decay coefficient and updated to the knowledge base, and the version number is updated at the same time. If the clinical outcome indicator reaches the preset threshold and the adoption rate is not lower than the preset adoption threshold, the judgment weight of the target node will be increased by a preset fixed value and updated to the knowledge base. At the same time, the version number will be updated to form a traceable record.

[0017] This invention also provides a multi-stage medical decision-making system based on structured judgment, employing the aforementioned multi-stage medical decision-making method based on structured judgment, including: Module M1 collects medical expert judgment information, transforms the medical expert judgment information into structured judgment nodes and stores them in the knowledge base. Each structured judgment node includes a disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action and initial judgment weight. Module M2 acquires patient data, including disease diagnosis codes and physiological indicators. It matches the physiological indicators with preset evidence-based medicine guidelines. If the match fails, it is determined to be a gray zone decision scenario and proceeds to step S3; otherwise, the process is terminated. Module M3 reads the disease diagnosis code, retrieves the structured judgment nodes in the knowledge base that match the disease diagnosis code, and records the successfully matched structured judgment nodes as candidate nodes; for each candidate node, it verifies whether the physiological indicators meet the applicable boundary conditions of the candidate node, constructs the candidate nodes that pass the verification into a candidate node set, and proceeds to step S4; if all candidate nodes fail the verification, it outputs a warning message and terminates the process. Module M4 obtains the logical judgment threshold of each candidate node in the candidate node set and compares it with the physiological indicators one by one; it records the candidate nodes whose physiological indicators meet the logical judgment thresholds as target nodes and generates a causal inference chain based on the content of the target nodes; if the physiological indicators do not meet all the logical judgment thresholds, the process is terminated. Module M5 reads the causal reasoning chain, generates decision information according to a preset template format, and outputs it to the user interface.

[0018] Preferably, the multi-stage medical decision-making system further includes: Module M6 calculates the adoption rate and clinical outcome indicators of the target nodes corresponding to the causal inference chain, adjusts the judgment weight of the target nodes based on the statistical results, and updates the version number.

[0019] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described multi-stage medical decision-making method based on structured judgment.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention transforms the experience of medical experts into structured judgment nodes containing fields such as disease diagnosis codes, clinical variables, logical judgment thresholds, and applicable boundary conditions, thereby realizing the structuring, storage, and reusability of expert tacit knowledge and solving the problem of the difficulty in inheriting and scaling up the application of expert experience.

[0021] 2. This invention improves the safety and rigor of medical decision-making in gray-zone scenarios by implementing a multi-stage decision-making process, including evidence-based medicine guideline matching, gray-zone scenario determination, data integrity and timeliness verification, and logical threshold judgment. When the data is incomplete, expired, or does not meet the judgment conditions, it outputs warnings or terminates the process, thereby improving the safety and rigor of medical decision-making in gray-zone scenarios.

[0022] 3. This invention generates a standardized causal reasoning chain, which enables decision outputs to have clear variable basis, node source and execution action, making the decision process traceable and explainable, thus solving the problems of traditional decision black box that is unexplainable and unclear responsibility boundaries.

[0023] 4. This invention dynamically adjusts the judgment weight and updates the version number by statistically analyzing the node adoption rate and clinical outcome indicators, forming a closed-loop feedback iteration mechanism. This allows the decision-making rules to be continuously optimized with clinical practice, improving the accuracy and applicability of the decision-making rules. Attached Figure Description

[0024] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a multi-stage medical decision-making method based on structured judgment, as described in an embodiment of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0026] This invention discloses a multi-stage medical decision-making method, system, and storage medium based on structured judgment, aiming to address the problems of existing medical decision support systems, such as difficulty in effectively utilizing expert experience, insufficient decision-making capabilities in gray-area scenarios, and lack of security verification and interpretability mechanisms. The method includes: transforming expert judgment information into a structured judgment node data structure; executing a multi-stage medical decision-making process; verifying the completeness and timeliness of patient data; generating interpretable decision-making basis; and dynamically updating node weights based on clinical adoption and clinical outcome indicators, generating version records. This invention structures tacit expert knowledge, achieving secure, interpretable, traceable, and iteratively optimizable clinical decision support in gray-area scenarios where evidence-based medicine guidelines are inapplicable or conflicting.

[0027] Example 1: Figure 1 This is a flowchart illustrating a multi-stage medical decision-making method based on structured judgment, as described in an embodiment of the present invention.

[0028] like Figure 1 As shown, this embodiment provides a multi-stage medical decision-making method based on structured judgment. Taking the gray area scenario of chronic kidney disease complicated with heart failure as an example, it includes the following steps: Step S1: Collect medical expert judgment information, transform the medical expert judgment information into structured judgment nodes and store them in the knowledge base. Each structured judgment node includes disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action and initial judgment weight.

[0029] Specifically, each structured judgment node also includes a node ID and version number; the disease diagnosis code is the International Classification of Diseases (ICD) code, used to identify the disease type to which the structured judgment node applies; clinical variables are used to describe the patient's current status indicators; applicable boundary conditions are the timeliness and completeness conditions of the clinical variables, including the longest data collection time and the list of required variables; logical judgment thresholds are the physiological indicator threshold ranges of the clinical variables; output actions are used to define recommended or avoided treatment measures; judgment weights are used to represent the priority of the structured judgment node; each structured judgment node is stored in the form of a data record.

[0030] In this embodiment, the judgment information from medical experts in cardiology and nephrology is obtained: for patients with creatinine clearance <30 ml / min and serum potassium ≤5.0 mmol / L, the use of high-potassium risk drug A should be avoided, and drug B, which is not metabolized by the kidneys, should be given priority. This is converted into a structured judgment node, including: a unique node ID of CKD_HF_001, a disease diagnosis code of ICD-10 code [N18.4, I50.9], indicating that it is applicable to patients with stage 4 chronic kidney disease complicated with heart failure; clinical variables [creatinine clearance, serum potassium]; logical judgment threshold of {creatinine clearance <30, serum potassium ≤5.0}; applicable boundary conditions of {serum potassium timeliness: 48h, required variables: [creatinine clearance, serum potassium]}, indicating that serum potassium data must be within 48 hours and both creatinine clearance and serum potassium are required fields; output decision action: recommend drug B, prohibit drug A; initial judgment weight is set to 1.0; version number is V1.0. The above fields are stored in the knowledge base as data records.

[0031] Step S2: Obtain patient data, which includes disease diagnosis codes and physiological indicators. Match the patient data with preset evidence-based medicine guidelines. If the match fails, it is determined to be a gray zone decision scenario and proceeds to step S3; otherwise, the process ends.

[0032] The evidence-based medicine guidelines for each clinical scenario are pre-converted into machine-readable logical expressions, and the applicable indicator ranges of each guideline clause are stored in the rule engine. During matching, the physiological indicators are compared one by one with the logical expressions in the rule engine. If a matching evidence-based medicine guideline clause exists, the number of matched clauses is further determined: if only a single evidence-based medicine guideline clause is matched, the single evidence-based medicine guideline clause is executed, and the process ends. If multiple evidence-based medicine guideline clauses are matched, a logical consistency check is performed on the multiple evidence-based medicine guideline clauses: if there is no contradiction between the clauses, one of them is selected for execution and the process ends; if there is a contradiction between the clauses, it is determined to be a gray zone decision scenario and proceeds to step S3. If no matching evidence-based medicine guidelines are found, the scenario is classified as a gray zone decision-making scenario, and the process proceeds to step S3.

[0033] In this embodiment, patient data was obtained as follows: Patient Zhang XX, disease diagnosis code ICD-10 is N18.4 (stage 4 chronic kidney disease), complication diagnosis is I50.9 (heart failure). Physiological indicators: creatinine clearance rate = 25 ml / min, serum potassium = 5.2 mmol / L (latest).

[0034] The evidence-based medicine guidelines for chronic kidney disease, when translated into logical expressions and applicable index ranges, are as follows: Drug A is suitable for patients with creatinine clearance ≥30 ml / min.

[0035] The physiological indicators were matched against all logical expressions. The patient's blood flow of 25 ml / min did not match the criteria in the evidence-based medicine guidelines. Therefore, the patient was determined to be in the "gray zone decision-making scenario where the guidelines do not explicitly specify" and proceeded to step S3.

[0036] Step S3: Read the disease diagnosis code, retrieve the structured judgment node in the knowledge base that matches the disease diagnosis code, and record the successfully matched structured judgment node as a candidate node; for each candidate node, verify whether the physiological indicators meet the applicable boundary conditions of the candidate node, construct the candidate node set of the candidate nodes that pass the verification, and proceed to step S4; if all candidate nodes fail the verification, output a warning message and terminate the process.

[0037] Step S3 includes the following sub-steps: Step S3.1: Read the disease diagnosis code, and use the disease diagnosis code as the query condition to retrieve all structured judgment nodes in the knowledge base that match the disease diagnosis code. Record the successfully matched structured judgment nodes as candidate nodes. Step S3.2: Obtain the applicable boundary conditions for each candidate node and verify them one by one with the physiological indicators of the patient data. The verification includes: verification of the completeness of required variables (i.e., checking whether the patient data contains all the required variables required by the node) and verification of the timeliness of variable data (i.e., checking whether the collection time of each physiological indicator is within the maximum allowable collection time specified by the node). Step S3.3: If all candidate nodes fail the verification, output a warning message and terminate the process; if there is at least one candidate node that passes the verification, construct a candidate node set from the candidate nodes that pass the verification and execute step S4.

[0038] In this embodiment, the ICD-10 disease diagnosis code of the patient data is read as [N18.4, I50.9]. Using this as a query condition, the structured judgment node CKD_HF_001, which contains both N18.4 and I50.9, is retrieved and recorded as a candidate node. The applicable boundary conditions of this candidate node are obtained and verified against the patient data. Required variable integrity verification: Checking whether the patient data includes creatinine clearance rate and serum potassium, the verification passes. Variable data timeliness verification: Checking the collection time of the serum potassium data, it is found that the patient's serum potassium data was collected 7 days ago, exceeding the 48-hour requirement of the applicable boundary conditions, and the verification fails.

[0039] Since the verification of the only candidate node failed, it was determined that there were no successfully verified candidate nodes. A security warning was displayed in the user interface pop-up window: "[Security Verification Failed] Blood potassium data is outdated (>48h), unable to proceed to the decision-making stage. Please review and update the blood potassium data, then click the 'Re-decide' button to re-initiate the process." The current decision-making process was terminated.

[0040] After seeing the prompt, the doctor ordered a repeat blood potassium test for the patient. 30 minutes later, the laboratory uploaded the new results: blood potassium = 4.5 mmol / L, collection time < 1 hour. The doctor clicked the "Re-decision" button and started from step S2.

[0041] Candidate node CKD_HF_001 was retrieved again and verified: the completeness of required variables passed the verification; the timeliness of variable data showed that the blood potassium data was collected within 30 minutes, meeting the ≤48h requirement, and the verification passed. The verified candidate node CKD_HF_001 was constructed into a candidate node set, and step S4 was executed.

[0042] Step S4: Obtain the logical judgment threshold of each candidate node in the candidate node set and compare it with the physiological indicators one by one; record the candidate nodes whose physiological indicators meet the logical judgment thresholds as target nodes, and generate a causal reasoning chain based on the content of the target nodes; if the physiological indicators do not meet all logical judgment thresholds, the process ends.

[0043] In this embodiment, the causal reasoning chain includes variable nodes, judgment nodes, and decision action nodes. Variable nodes, judgment nodes, and decision action nodes are data structure node types that constitute the causal reasoning chain, representing medical variable inputs, logical judgment rules, and decision output actions, respectively. Their meanings are as follows: Variable nodes represent input variables or clinical indicators involved in the medical decision-making process, such as physiological indicators, test results, or symptom information. These nodes provide the basic data source for the causal reasoning chain. Judgment nodes perform logical judgments on variable nodes, based on preset logical judgment thresholds or rules. For example, they compare the physiological indicators corresponding to the variable node with the logical judgment threshold to determine whether the triggering condition is met. Decision action nodes trigger corresponding medical decision outputs when the judgment node meets preset conditions, such as generating medical prompts, risk assessment results, or intervention recommendations.

[0044] Specifically, the causal inference chain is formatted as follows: [clinical variable] satisfies [logical judgment threshold]; triggers expert judgment node [node ID]; executes [output decision action].

[0045] Furthermore, if there are multiple target nodes, all target nodes are sorted from high to low according to their judgment weights to form a priority recommendation list. Based on the content of the first node in the priority recommendation list, a causal inference chain is generated.

[0046] In this embodiment, physiological indicators are verified against the logical judgment thresholds of candidate node CKD_HF_001: creatinine clearance rate of 25 ml / min meets the condition <30, and serum potassium of 4.5 mmol / L meets the condition ≤5.0. The corresponding candidate node CKD_HF_001 is recorded as the target node. A causal reasoning chain is generated based on the content of the target node: creatinine clearance rate meets <30, and serum potassium meets ≤5.0; expert judgment node CKD_HF_001 is triggered; drug B is recommended, drug A is prohibited, and step S5 is executed.

[0047] Step S5: Read the causal reasoning chain, generate decision information according to the preset template format, and output it to the user interface.

[0048] In this embodiment, the causal reasoning chain is read, and decision information including recommended actions, justifications, risk warnings, and source nodes is generated according to a preset template format: the recommended option is to use drug B, the prohibited option is to prohibit drug A and warn of high potassium risk, the decision basis is a creatinine clearance rate of 25 ml / min less than 30 and a serum potassium level of 4.5 mmol / L less than or equal to 5.0, which conforms to the rule of knowledge base node CKD_HF_001, there are no alternative options, and the source is V1.0 of expert judgment node CKD_HF_001. The above information is then output to the user interface.

[0049] Step S6: Calculate the adoption rate and clinical outcome indicators of the target nodes corresponding to the causal inference chain, adjust the judgment weight of the target nodes and update the version number based on the statistical results.

[0050] Specifically, the adoption rate is the actual number of adoptions divided by the total number of recommendations. By periodically calculating the clinical outcome indicators of the target node, if the adoption rate is lower than the preset adoption threshold, or the clinical outcome indicators do not reach the preset indicator threshold, the judgment weight of the target node is multiplied by the preset decay coefficient and updated to the knowledge base, and the version number is updated at the same time. If the clinical outcome indicator reaches the preset threshold and the adoption rate is not lower than the preset adoption threshold, the judgment weight of the target node will be increased by a preset fixed value and updated to the knowledge base. At the same time, the version number will be updated to form a traceable record.

[0051] In this embodiment, the target node CKD_HF_001 was recommended 10 times, and the doctor adopted it 8 times, resulting in an adoption rate of 80%. By tracking the subsequent serum potassium data of the 8 patients who adopted the recommendations, it was found that 7 patients had serum potassium ≤5.5mmol / L for 7 consecutive days after using drug B, and the clinical outcome indicator was determined to be met (the preset indicator threshold is set to serum potassium ≤5.5mmol / L for 7 consecutive days). Furthermore, the adoption rate of 80% ≥ the adoption rate threshold (set to 60%), so the judgment weight of the target node CKD_HF_001 is increased by a preset fixed value (0.1), updated to 1.1, and the judgment weight of the target node in the knowledge base is updated through SQL statements and the version number is increased to V1.1.

[0052] Example 2: The present invention also provides a multi-stage medical decision-making system based on structured judgment. The multi-stage medical decision-making system based on structured judgment can be implemented by executing the process steps of the multi-stage medical decision-making method based on structured judgment. That is, those skilled in the art can understand the multi-stage medical decision-making method based on structured judgment as a preferred implementation of the multi-stage medical decision-making system based on structured judgment.

[0053] Specifically, this multi-stage medical decision-making system based on structured judgment includes: Module M1 collects medical expert judgment information, transforms the medical expert judgment information into structured judgment nodes and stores them in the knowledge base. Each structured judgment node includes a disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action and initial judgment weight. Module M2 acquires patient data, including disease diagnosis codes and physiological indicators. It matches the physiological indicators with preset evidence-based medicine guidelines. If the match fails, it is determined to be a gray zone decision scenario and proceeds to step S3; otherwise, the process is terminated. Module M3 reads the disease diagnosis code, retrieves the structured judgment nodes in the knowledge base that match the disease diagnosis code, and records the successfully matched structured judgment nodes as candidate nodes; for each candidate node, it verifies whether the physiological indicators meet the applicable boundary conditions of the candidate node, constructs the candidate nodes that pass the verification into a candidate node set, and proceeds to step S4; if all candidate nodes fail the verification, it outputs a warning message and terminates the process. Specifically, module M3 includes the following sub-modules: Module M3.1 reads the disease diagnosis code, retrieves all structured judgment nodes in the knowledge base that match the disease diagnosis code using the disease diagnosis code as the query condition, and records the successfully matched structured judgment nodes as candidate nodes. Module M3.2 obtains the applicable boundary conditions for each candidate node and verifies them one by one with the physiological indicators of the patient data. The verification includes: verification of the completeness of required variables (i.e., checking whether the patient data contains all the required variables required by the node) and verification of the timeliness of variable data (i.e., checking whether the collection time of each physiological indicator is within the maximum allowable collection time specified by the node). Module M3.3: If all candidate nodes fail the verification, a warning message will be output and the process will be terminated; if at least one candidate node passes the verification, the candidate nodes that pass the verification will be constructed into a candidate node set and module M4 will be executed.

[0054] Module M4 obtains the logical judgment threshold of each candidate node in the candidate node set and compares it with the physiological indicators one by one; it records the candidate nodes whose physiological indicators meet the logical judgment thresholds as target nodes and generates a causal inference chain based on the content of the target nodes; if the physiological indicators do not meet all the logical judgment thresholds, the process is terminated. Module M5 reads the causal reasoning chain, generates decision information according to a preset template format, and outputs it to the user interface.

[0055] Module M6 calculates the adoption rate and clinical outcome indicators of the target nodes corresponding to the causal inference chain, adjusts the judgment weight of the target nodes based on the statistical results, and updates the version number.

[0056] Example 3: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a multi-stage medical decision-making method based on structured judgment, as described in Embodiment 1.

[0057] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0058] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A multi-stage medical decision-making method based on structured judgment, characterized in that, Includes the following steps: Step S1: Collect medical expert judgment information, convert the medical expert judgment information into structured judgment nodes and store them in the knowledge base. Each structured judgment node includes a disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action and initial judgment weight. Step S2: Obtain patient data, which includes disease diagnosis codes and physiological indicators. Match the physiological indicators with preset evidence-based medicine guidelines. If the matching fails, it is determined to be a gray zone decision scenario, and proceed to step S3; otherwise, terminate the process. Step S3: Read the disease diagnosis code, retrieve the structured judgment nodes in the knowledge base that match the disease diagnosis code, and record the successfully matched structured judgment nodes as candidate nodes; for each candidate node, verify whether the physiological indicators meet the applicable boundary conditions of the candidate node, construct the candidate node set of the verified candidate nodes, and proceed to step S4. If all candidate nodes fail the verification, a warning message will be output and the process will be terminated. Step S4: Obtain the logical judgment threshold of each candidate node in the candidate node set and compare it with the physiological indicators one by one; Candidate nodes whose physiological indicators meet the logical judgment threshold are recorded as target nodes, and a causal reasoning chain is generated based on the content of the target nodes. If the physiological indicators do not meet all logical judgment thresholds, the process will terminate. Step S5: Read the causal reasoning chain, generate decision information according to the preset template format, and output it to the user interface.

2. The multi-stage medical decision-making method based on structured judgment according to claim 1, characterized in that, In step S1, the structured judgment node further includes a node ID and a version number; the disease diagnosis code is the International Classification of Diseases (ICD) code, used to identify the disease type to which the structured judgment node is applicable; the clinical variables are used to describe the patient's current status indicators; the applicable boundary conditions are the timeliness and completeness conditions of the clinical variables, including the longest data collection time and the list of mandatory variables; the logical judgment threshold is the physiological indicator threshold range of the clinical variables; the output action is used to define recommended or avoided treatment measures; the judgment weight is used to represent the priority of the structured judgment node; each structured judgment node is stored in the form of a data record.

3. The multi-stage medical decision-making method based on structured judgment according to claim 2, characterized in that, In step S2, the evidence-based medicine guidelines corresponding to each clinical scenario are pre-converted into machine-readable logical expressions, and the applicable indicator range of each evidence-based medicine guideline clause is stored in the rule engine. During matching, the physiological indicators are compared one by one with the logical expressions in the rule engine: If a matching evidence-based medicine guideline clause exists, the number of matched clauses is further determined: if only a single evidence-based medicine guideline clause is matched, the single evidence-based medicine guideline clause is executed and the process ends; if multiple evidence-based medicine guideline clauses are matched, the logical consistency of the multiple evidence-based medicine guideline clauses is checked: if there is no contradiction between the clauses, one of them is selected for execution and the process ends; if there is a contradiction between the clauses, it is determined to be a gray zone decision scenario and proceeds to step S3. If no matching evidence-based medicine guidelines are found, the scenario is classified as a gray zone decision-making scenario, and the process proceeds to step S3.

4. The multi-stage medical decision-making method based on structured judgment according to claim 3, characterized in that, Step S3 includes the following sub-steps: Step S3.1: Read the disease diagnosis code, use the disease diagnosis code as the query condition to retrieve all structured judgment nodes in the knowledge base that match the disease diagnosis code, and record the successfully matched structured judgment nodes as candidate nodes; Step S3.2: Obtain the applicable boundary conditions for each candidate node and verify them one by one with the physiological indicators of the patient data. The verification includes: verification of the completeness of required variables and verification of the timeliness of variable data. Step S3.3: If all candidate nodes fail the verification, output a warning message and terminate the process; if there is at least one candidate node that passes the verification, construct a candidate node set from the candidate nodes that pass the verification and execute step S4.

5. The multi-stage medical decision-making method based on structured judgment according to claim 4, characterized in that, In step S4, the causal reasoning chain is formatted as follows: [clinical variable] satisfies [logical judgment threshold]; triggers expert judgment node [node ID]; executes [output decision action].

6. The multi-stage medical decision-making method based on structured judgment according to claim 5, characterized in that, In step S4, if there are multiple target nodes, all target nodes are sorted from high to low according to their judgment weights to form a priority recommendation list. Based on the content of the first node in the priority recommendation list, a causal inference chain is generated.

7. The multi-stage medical decision-making method based on structured judgment according to claim 1, characterized in that, The multi-stage medical decision-making approach also includes: Step S6: Calculate the adoption rate and clinical outcome indicators of the target node corresponding to the causal reasoning chain, adjust the judgment weight of the target node and update the version number based on the statistical results.

8. The multi-stage medical decision-making method based on structured judgment according to claim 7, characterized in that, In step S6, the adoption rate is the actual number of adoptions divided by the total number of recommendations. By periodically calculating the clinical outcome indicators of the target node, if the adoption rate is lower than the preset adoption threshold, or the clinical outcome indicators do not reach the preset indicator threshold, the judgment weight of the target node is multiplied by the preset decay coefficient and updated to the knowledge base, and the version number is updated at the same time. If the clinical outcome indicator reaches the preset indicator threshold and the adoption rate is not lower than the preset adoption threshold, the judgment weight of the target node is increased by a preset fixed value and updated to the knowledge base, while the version number is updated to form a traceable record.

9. A multi-stage medical decision-making system based on structured judgment, employing the multi-stage medical decision-making method based on structured judgment as described in any one of claims 1-8, characterized in that, include: Module M1 collects medical expert judgment information, transforms the medical expert judgment information into structured judgment nodes and stores them in the knowledge base. Each structured judgment node includes a disease diagnosis code, clinical variables, logical judgment threshold, applicable boundary conditions, output decision action and initial judgment weight. Module M2 acquires patient data, which includes disease diagnosis codes and physiological indicators. The physiological indicators are matched with preset evidence-based medicine guidelines. If the matching fails, it is determined to be a gray zone decision scenario and proceeds to step S3; otherwise, the process is terminated. Module M3 reads the disease diagnosis code, retrieves the structured judgment nodes in the knowledge base that match the disease diagnosis code, and records the successfully matched structured judgment nodes as candidate nodes; for each candidate node, verifies whether the physiological indicators meet the applicable boundary conditions of the candidate node, constructs the candidate nodes that pass the verification into a candidate node set, and proceeds to step S4; If all candidate nodes fail the verification, a warning message will be output and the process will be terminated. Module M4 obtains the logical judgment threshold of each candidate node in the candidate node set and compares it with the physiological indicators one by one; Candidate nodes whose physiological indicators meet the logical judgment threshold are recorded as target nodes, and a causal reasoning chain is generated based on the content of the target nodes. If the physiological indicators do not meet all logical judgment thresholds, the process will terminate. Module M5 reads the causal reasoning chain, generates decision information according to a preset template format, and outputs it to the user interface; Module M6 calculates the adoption rate and clinical outcome indicators of the target nodes corresponding to the causal reasoning chain, adjusts the judgment weight of the target nodes based on the statistical results, and updates the version number.

10. A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method of any one of claims 1-8.

Images