A large model-based decision intelligence reasoning method and system

Abstract

The application discloses a decision-making intelligent reasoning method and system based on a large model, relates to the field of clinical medicine, and solves the problem that the existing clinical intelligent reasoning system has limited timeliness and accuracy in making reasoning decisions due to relying on specific patient characteristics inputted, and adopts the following scheme: S1, a clinical diagnosis and treatment database is built; S2, a patient diagnosis and treatment model is customized; S3, feature matching is performed; S4, optimization and rearrangement are performed; S5, a patient clinical feature model is customized; S6, secondary feature matching is performed; and S7, a case model decision is made; the medical clinical diagnosis and treatment decision-making intelligent reasoning method and system based on the large model can select and rearrange case models in the database according to patient physical sign characteristics, input new clinical features of a patient according to the demand of the patient, evaluate the rearranged case models in the case models according to the specific characteristics, and finally make a final decision by a doctor in charge of the patient.

Description

Technical Field

[0001] This invention relates to the field of clinical medical technology, specifically to a decision-making intelligent reasoning method and system based on a large model. Background Technology

[0002] The demand for clinical medical resources is surging, but medical personnel and related resources are still scarce. The quality and efficiency of diagnosis and treatment need to be improved. The traditional diagnosis and treatment model relies on doctors' experience and evidence-based medicine guidelines, which has problems such as low efficiency and insufficient objectivity and accuracy in judgment. In addition, due to the complexity of diseases and differences in doctors' experience, deviations in the implementation of guidelines are prone to occur, and even missed diagnoses and misdiagnoses, which increase the burden on patients and intensify doctor-patient conflicts. Although high-quality medical personnel training is being carried out, the training cycle is long and it is difficult to meet the current urgent diagnosis and treatment needs.

[0003] Intelligent clinical decision support systems are an effective solution to current clinical medical problems. These systems rely on expert knowledge and knowledge reasoning mechanisms to simulate and assist clinical decision-making. Their core advantages lie in their comprehensive knowledge reserves and efficient knowledge reasoning. The system can uniformly integrate and express standard treatment guidelines, enabling rapid knowledge retrieval, reasoning, and application in clinical decision-making: during the triage and assessment phase, rule-based reasoning accurately applies guidelines; during the treatment plan development phase, case model reasoning draws upon historical case models. Furthermore, the system is flexible and scalable, and can also serve as an auxiliary tool for training medical and nursing personnel.

[0004] In existing intelligent clinical decision support systems, patients' clinical manifestations are intelligently combined to assist the decision-making system. However, the characteristics of clinical manifestations are limited, and the values ​​of the detected parameters fluctuate to a certain extent. As a result, real-time clinical manifestations cannot be used as a stable basis for clinical decision support, making it difficult to guarantee the accuracy of intelligent decision support in clinical practice.

[0005] Therefore, we propose a decision-making intelligent reasoning method and system based on a large model. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a decision-making intelligent reasoning method and system based on a large model, which solves the problem that existing clinical intelligent reasoning systems rely on specific patient characteristics to make reasoning decisions, resulting in limited timeliness and consequently limited accuracy.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a decision-making intelligent reasoning method based on a large model, comprising the following steps:

[0008] S1: Build a clinical diagnosis and treatment database and prepare case models;

[0009] S2: Collect patient past information features, collect specific basic clinical information of patients and relevant features in patients' past medical records, and customize patient diagnosis and treatment models;

[0010] S3: Feature matching, retrieves basic clinical information and relevant features of the input patient from the clinical diagnosis and treatment database, executes the rule retrieval mechanism, and retrieves multiple similar case models from the clinical database;

[0011] S4: Optimize and rearrange. The case model rearrangement optimization mechanism is used as the execution mechanism to score and rearrange multiple similar case models retrieved, and output multiple optimized case models.

[0012] S5: Collect new clinical characteristics of patients, collect information on various indicators of patients' clinical examinations, types of medications and dosages of medications, and customize patients' clinical characteristic models;

[0013] S6: Secondary feature matching, using the collected new clinical features of patients as input variables, executes a data rule reasoning mechanism to select a suitable evaluation case model from multiple optimized case models;

[0014] S7: Case model decision-making: Based on the specific opinions of the patient's attending physician and the selected evaluation case models, the clinical diagnosis and treatment case model that best matches the patient's characteristics is chosen.

[0015] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model described in this invention, the construction of the clinical diagnosis and treatment database in step S1 includes two parts: data model manufacturing and data iterative updating. The data model manufacturing includes a data collection part, which takes the patient's own characteristics as the center, the types and amounts of medications used in the diagnosis and treatment process as clues, and prepares case models after the data is completely desensitized.

[0016] Among them, information on individual patients in each department is collected in accordance with standardized transmission protocols and in-hospital integrated engines, and the collected information is desensitized and prepared into specific case models.

[0017] The clinical diagnosis and treatment database is updated and iterated. This clinical diagnosis and treatment database is connected to the EMR / HIS system in various hospitals. After automatic desensitization, the feature information in the case models is compared and different case models are classified into different departmental databases.

[0018] In this process, after patient information is entered into the hospital's EMR / HIS system, the information is collected using a standardized transmission protocol and the hospital's integrated engine. After being desensitized, new case models are prepared and stored in the clinical diagnosis and treatment database according to the department category.

[0019] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model described in this invention, in step S2, the patient's past information is collected, and a patient diagnosis and treatment model is customized with the patient's own characteristics as the center and the characteristics during the diagnosis and treatment process as clues.

[0020] The patient is automatically categorized based on their own characteristics, and a patient diagnosis and treatment model is customized based on the patient's previous medical history and medication information when the patient is admitted to the hospital.

[0021] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model described in this invention, feature matching in step S3 refers to using the patient diagnosis and treatment model established in step S2 as a comparison file, executing a rule retrieval mechanism in the diagnosis and treatment database, and retrieving multiple similar case models.

[0022] In this process, features are extracted from the patient's diagnosis and treatment model. During the comparison, features from the clinical diagnosis and treatment database are extracted simultaneously. The matching degree of features in the patient's diagnosis and treatment model and the corresponding case model is compared according to the cosine similarity algorithm. This algorithm is used to retrieve multiple similar case models from the clinical diagnosis and treatment database.

[0023] The rule retrieval mechanism refers to taking the different specific features in the diagnosis and treatment model as input feature variables, comparing the input feature variables in the clinical diagnosis and treatment database, and filtering out similar case models from the past based on the matching degree of the input feature variables.

[0024] The rule retrieval mechanism is the cosine similarity algorithm. The input feature variables are the features extracted from the patient diagnosis and treatment model, while the corresponding features can be extracted from the case model. The algorithm compares the input and extracted features and uses the matching degree data to select similar case models from the clinical diagnosis and treatment database.

[0025] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model as described in this invention, the optimization and rearrangement in step S4 refers to optimizing and rearranging each selected past similar case model through a case model rearrangement optimization mechanism.

[0026] Among them, the optimization and rearrangement step can extract features from clinical data and compare these features in the clinical diagnosis and treatment database, thereby re-selecting a new batch of similar case models in the clinical diagnosis and treatment database;

[0027] The case model reordering optimization mechanism includes feature extraction, which extracts features from similar case models and uses the repeated features from multiple similar case models as input variables to feed them back into step S3. The rule enforcement retrieval mechanism therein uses these repeated features as input variables to re-select a batch of similar case models.

[0028] Among them, repeated feature extraction and subsequent feature comparison can filter out multiple features or high-probability covered features from massive database case model modules, trace back to the original case model, and plan the case model as a similar case model;

[0029] Then, all the selected past case models are subjected to multi-dimensional adaptation scores. After the scores are completed, they are initially sorted according to the rules. Then, the selected case models are optimized in local solutions. The optimization is carried out based on the differences in local features of the selected solutions and the recovery rate after treatment. After the optimization is completed, the optimized solutions are re-verified and re-sorted. The optimized case models are selected by score and sorted again according to the multi-dimensional adaptation scores.

[0030] The multi-dimensional adaptation score includes a clinical medical adaptation dimension, which includes disease feature matching degree, individual basic matching degree, contraindication matching degree, and treatment scenario matching degree. It also includes a protocol execution adaptation dimension, which includes protocol efficacy matching degree, adverse reaction matching degree, medical resource matching degree, and economic and compliance matching degree. After the scoring is completed, it is initially ranked according to rules, and then local optimization is performed. Gradient descent (GD / SGD) method is used to iteratively adjust local feature parameters until the target converges, thereby achieving local optimization of the case model. After optimization, the cases are ranked according to the scores.

[0031] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model described in this invention, the new clinical characteristics of the patient in step S5 refer to the characteristics generated after the patient is admitted to the hospital. The new clinical characteristics of the patient are taken as the center and the medication information is used as a clue to formulate the patient's clinical characteristic model.

[0032] Among them, the new clinical characteristics of patients are set when patients enter the corresponding department for examination and exhibit new clinical characteristics. The hospital's standardized transmission protocol and integrated engine work together to collect this information and then desensitize it to prepare a new clinical characteristic model of the patient.

[0033] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model as described in this invention, in step S6, the clinical feature model established in step S5 is used as the input variable, and the data rule reasoning mechanism is executed to select a suitable evaluation case model from the optimized case model.

[0034] The input variable is the new clinical feature model of the patient. The feature matching degree in the optimized case model is compared with the new clinical feature model of the patient, and the evaluation case model in the optimized case model is selected through the data rule reasoning mechanism.

[0035] Furthermore, the rule-based reasoning mechanism includes four steps: feature matching, condition judgment, conflict resolution, and optimal case model selection.

[0036] Feature matching refers to quantitatively comparing the features of the case model to be analyzed with the features of existing case models (through cosine similarity) to select candidate case models with similar features;

[0037] Conditional judgment refers to verifying whether the candidate case model meets the scenario and constraints of the case model to be analyzed, based on preset rules / diagnosis and treatment standards.

[0038] Conflict resolution refers to resolving conflicts and eliminating invalid case models when there are contradictions in features / solutions between candidate case models and case models to be analyzed, through methods such as rule priority and weight assignment.

[0039] The optimal case model refers to selecting the best / most suitable case model from the effective candidate case models after matching, judgment and elimination, based on similarity, fitness and effect priority.

[0040] As a preferred embodiment of the intelligent reasoning method for medical clinical diagnosis and treatment decision-making based on a large model as described in this invention, in step S7, external conditions are used as input variables, and the attending physician of the patient comprehensively evaluates the case model to make the final appropriate clinical diagnosis and treatment decision.

[0041] External factors include the degree of manifestation of the patient's specific characteristics and equipment. The patient's attending physician comprehensively assesses the patient's specific situation and selects the most suitable clinical diagnosis and treatment decision.

[0042] A medical clinical diagnosis and treatment decision-making intelligent reasoning system based on a large model, the system includes a clinical diagnosis and treatment database module, a feature matching module, an optimization and rearrangement module, a secondary feature matching module, and a case model decision-making module;

[0043] The clinical diagnosis and treatment database module uses a unified representation to store various clinical diagnosis and treatment knowledge.

[0044] The feature matching module takes the patient diagnosis and treatment model as input variable, executes the rule retrieval mechanism to output similar case models that match the features of new patients;

[0045] The optimization and rearrangement module uses the common features extracted from the similar case model as input variables, executes the case model rearrangement optimization mechanism, and outputs an optimized case model with optimized features.

[0046] The secondary feature matching module uses the patient's clinical feature model as input variable and executes a data rule reasoning mechanism to output a filtered and adapted evaluation case model.

[0047] The case model decision module uses the specific external environment as input variables, executes the reasoning mechanism of the patient's attending physician, and makes the final appropriate clinical diagnosis and treatment decision by the patient's attending physician.

[0048] This invention provides a decision-making intelligent reasoning method and system based on a large model. It has the following beneficial effects:

[0049] 1. This intelligent reasoning method and system for medical clinical diagnosis and treatment decision-making based on a large model, through the setting of a clinical diagnosis and treatment database module, a feature matching module, an optimization and rearrangement module, a secondary feature matching module, and a case model decision-making module, can select and rearrange case models in the database based on the patient's physical characteristics, and input new clinical characteristics of the patient according to their needs. The rearranged case model is evaluated in the case model based on the specific characteristics, ensuring the suitability of the case model with the patient's diagnosis and treatment information. Finally, the patient's attending physician makes the final decision based on the patient's physical condition, the patient's latest clinical characteristics, external conditions, and the optimized case model, ensuring the accuracy of the clinical diagnosis and treatment decision. This solves the problem that the timeliness of reasoning decisions made by existing clinical intelligent reasoning systems based on the input of specific patient characteristics is limited, which in turn leads to limited accuracy.

[0050] 2. This intelligent reasoning method and system for medical clinical diagnosis and treatment decision-making based on a large model, in the existing reasoning and decision-making process of clinical diagnosis and treatment, selects similar case models in the database through feature extraction and screening, and then performs secondary optimization screening of the selected similar case models with the case model re-ranking optimization mechanism. It also uses the multi-dimensional adaptation scoring mechanism to re-rank the optimized case models in terms of scoring, thereby ensuring the degree of fit between the selected and optimized case models and the specific clinical characteristics of the patients, and further ensuring the accuracy of the case models determined after subsequent reasoning. Attached Figure Description

[0051] Figure 1 This is a flowchart of the method of the present invention;

[0052] Figure 2 This is a system module diagram of the present invention;

[0053] Figure 3 This is a flowchart of the rule retrieval mechanism of the present invention;

[0054] Figure 4 This is a flowchart of the case model rearrangement optimization mechanism of the present invention;

[0055] Figure 5 This is a flowchart of the rule-based reasoning mechanism of the present invention. Detailed Implementation

[0056] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1-5 This invention provides a technical solution: a decision-making intelligent reasoning method based on a large model, which includes the following steps:

[0058] S1: Build a clinical diagnosis and treatment database and prepare case models;

[0059] S2: Collect patient past information features, collect specific basic clinical information of patients and relevant features in patients' past medical records, and customize patient diagnosis and treatment models;

[0060] S3: Feature matching, retrieves the basic clinical information and related features of the input patient from the clinical diagnosis and treatment database, executes the rule retrieval mechanism, and retrieves multiple similar case models from the clinical database;

[0061] S4: Optimize and rearrange. The case model rearrangement optimization mechanism is used as the execution mechanism to score and rearrange multiple similar case models retrieved, and output multiple optimized case models.

[0062] S5: Collect new clinical characteristics of patients, collect information on various indicators of patients' clinical examinations, types of medications and dosages of medications, and customize patients' clinical characteristic models;

[0063] S6: Secondary feature matching, using the collected new clinical features of patients as input variables, executes a data rule reasoning mechanism to select a suitable evaluation case model from multiple optimized case models;

[0064] S7: Case model decision-making: Based on the specific opinions of the patient's attending physician and the selected evaluation case models, the clinical diagnosis and treatment case model that best matches the patient's characteristics is chosen.

[0065] The intelligent reasoning method and system for medical clinical diagnosis and treatment decision-making based on a large model, through the setting of a clinical diagnosis and treatment database module, a feature matching module, an optimization and rearrangement module, a secondary feature matching module, and a case model decision-making module, can select and rearrange case models in the database based on the patient's physical characteristics. It can also input new clinical characteristics of the patient according to their needs, evaluate the rearranged case model in the case model based on the specific characteristics, and ensure the suitability of the case model with the patient's diagnosis and treatment information. Finally, the patient's attending physician makes the final decision based on the patient's physical condition, the patient's latest clinical characteristics, external conditions, and the optimized case model, ensuring the accuracy of the clinical diagnosis and treatment decision. This solves the problem that the timeliness of reasoning decisions made by existing clinical intelligent reasoning systems based on the input of specific patient characteristics is limited, which in turn leads to limited accuracy.

[0066] It should be noted that the construction of the clinical diagnosis and treatment database in step S1 includes two parts: data model creation and data iterative updates. Data model creation includes the data collection part, which focuses on the patient's own characteristics and uses the types and dosages of medications used in the diagnosis and treatment process as clues. Case models are prepared after the data is completely de-identified. Specifically, information on individual patients in each department is collected according to standardized transmission protocols and in-hospital integration engines, and the collected information is de-identified and prepared into specific case models.

[0067] Furthermore, the clinical diagnosis and treatment database is updated and iterated. This database is connected to the EMR / HIS system in various hospitals. After automatic desensitization, the feature information in the case models is compared, and different case models are classified into different departmental databases. Specifically, according to the hospital's EMR / HIS system, after the patient information is entered into the system, the information is collected by a standardized transmission protocol and the hospital's integrated engine. After desensitization, new case models are prepared, and the prepared case models are classified and stored in the clinical diagnosis and treatment database according to the department category.

[0068] It should be noted that in step S2, the patient's past information is collected, and the patient's diagnosis and treatment model is customized based on the patient's own characteristics and the characteristics during the diagnosis and treatment process. Specifically, the patient is automatically classified according to his or her own characteristics, and the patient's diagnosis and treatment model is customized based on the patient's previous medical history and medication information when the patient is admitted to the hospital.

[0069] It should be noted that feature matching in step S3 refers to using the patient diagnosis and treatment model established in step S2 as a comparison file, executing a rule-based retrieval mechanism in the diagnosis and treatment database to retrieve multiple similar case models. Specifically, features are extracted from the patient diagnosis and treatment model, and during comparison, features from the clinical diagnosis and treatment database being compared are simultaneously extracted. The matching degree of features in the patient diagnosis and treatment model and the corresponding case models is compared using a cosine similarity algorithm. This algorithm is used to retrieve multiple similar case models from the clinical diagnosis and treatment database.

[0070] It is further necessary to explain that the rule-based retrieval mechanism refers to using the different specific features in the diagnosis and treatment model as input feature variables, comparing the input feature variables with those in the clinical diagnosis and treatment database, and selecting similar case models from the past based on the matching degree of the input feature variables. This rule-based retrieval mechanism is the cosine similarity algorithm, where the input feature variables are features extracted from the patient diagnosis and treatment model, and corresponding features can be extracted from the case models. The algorithm compares and matches the input and extracted features, and uses the matching degree data to select similar case models from the clinical diagnosis and treatment database.

[0071] It should be noted that the optimization and reordering in step S4 refers to optimizing and reordering each of the selected past similar case models through the case model reordering optimization mechanism; the optimization and reordering step can extract features from clinical data and compare these features with features in the clinical diagnosis and treatment database, thereby re-selecting a new batch of similar case models in the clinical diagnosis and treatment database;

[0072] Furthermore, the case model reordering optimization mechanism includes feature extraction, which extracts features from similar case models and re-feeds the repeated features from multiple similar case models as input variables to step S3. The rule enforcement retrieval mechanism therein uses these repeated features as input variables to re-select a batch of similar case models. The repeated feature extraction and subsequent feature comparison can filter out multiple features or high-probability covered features from the massive database of case model modules, trace them back to the original case model, and classify the case model as a similar case model.

[0073] Furthermore, it needs to be explained that all selected past case models undergo multi-dimensional adaptation scoring. After scoring, they are initially ranked according to rules, and then the selected case models are optimized in local aspects based on the differences in local features and the recovery rate after treatment. After optimization, the optimized plans are re-verified and re-ranked, and then re-ranked according to the multi-dimensional adaptation scores to select the optimized case models. The multi-dimensional adaptation scores include clinical medical adaptation dimensions, including disease feature matching, individual baseline matching, contraindication matching, and treatment scenario matching, as well as plan execution adaptation dimensions, including plan efficacy matching, adverse reaction matching, medical resource matching, and economic and compliance matching. After scoring, they are initially ranked according to rules, and then local optimization is performed using gradient descent (GD / SGD) – iteratively adjusting local feature parameters until the target converges, achieving local optimization of the case models. After optimization, they are ranked according to the scores.

[0074] It should be noted that the new clinical characteristics of the patient in step S5 refer to the characteristics that the patient develops after being admitted to the hospital. The new clinical characteristics of the patient are taken as the center and the medication information is used as a clue to develop the patient's clinical characteristic model. Specifically, the new clinical characteristics of the patient are set as the new clinical characteristics that the patient exhibits when entering the corresponding department for examination. The standardized transmission protocol and integration engine within the hospital are used to collect this information, and after desensitization, a new patient clinical characteristic model is prepared.

[0075] It should be noted that in step S6, the clinical feature model established in step S5 is used as the input variable, and a data rule reasoning mechanism is executed to select a suitable evaluation case model from the optimized case model. Here, the input variable is the new clinical feature model of the patient. The feature matching degree of the new clinical feature model of the patient is compared with that of the optimized case model, and the evaluation case model in the optimized case model is selected through the data rule reasoning mechanism.

[0076] Furthermore, the rule-based reasoning mechanism includes four steps: feature matching, condition judgment, conflict resolution, and optimal case model selection.

[0077] Feature matching refers to quantitatively comparing the features of the case model to be analyzed with the features of existing case models (through cosine similarity) to select candidate case models with similar features.

[0078] Conditional judgment refers to verifying whether the candidate case model meets the scenario and constraints of the case model to be analyzed, based on preset rules / diagnosis and treatment standards.

[0079] Conflict resolution refers to resolving conflicts and eliminating invalid case models when there are contradictions in features / solutions between candidate case models and case models to be analyzed, through methods such as rule priority and weight assignment.

[0080] The optimal case model refers to selecting the best / most suitable case model from the effective candidate case models after matching, judgment and elimination, based on similarity, fitness and effect priority.

[0081] It should be noted that in step S7, external conditions are used as input variables, and the patient's attending physician comprehensively evaluates the case model to make the final appropriate clinical diagnosis and treatment decision. The external conditions include the degree of manifestation of the patient's specific characteristics, equipment, etc. The patient's attending physician comprehensively evaluates the patient's specific situation and selects the most appropriate clinical diagnosis and treatment decision for the patient.

[0082] A medical clinical diagnosis and treatment decision-making intelligent reasoning system based on a large model is characterized in that: the system includes a clinical diagnosis and treatment database module, a feature matching module, an optimization and rearrangement module, a secondary feature matching module, and a case model decision-making module;

[0083] The clinical diagnosis and treatment database module uses a unified representation to store various clinical diagnosis and treatment knowledge.

[0084] The feature matching module takes the patient diagnosis and treatment model as input variable, executes the rule retrieval mechanism to output similar case models that match the features of new patients;

[0085] The optimization and rearrangement module uses the common features extracted from the similar case model as input variables, executes the case model rearrangement optimization mechanism, and outputs an optimized case model with optimized features.

[0086] The secondary feature matching module uses the patient's clinical feature model as input variable and executes a data rule reasoning mechanism to output a filtered and adapted evaluation case model.

[0087] The case model decision module uses the specific external environment as input variables, executes the reasoning mechanism of the patient's attending physician, and makes the final appropriate clinical diagnosis and treatment decision by the patient's attending physician.

[0088] Furthermore, this intelligent reasoning method and system for medical clinical diagnosis and treatment decision-making based on a large model, in the existing reasoning and decision-making process of clinical diagnosis and treatment, selects similar case models in the database through feature extraction and screening, and then performs secondary optimization screening of the selected similar case models with the case model re-ranking optimization mechanism, and uses the multi-dimensional adaptation scoring mechanism to re-rank the optimized case models in terms of scoring, thereby ensuring the degree of fit between the selected and optimized case models and the specific clinical characteristics of the patients, and further ensuring the accuracy of the case models determined after subsequent reasoning.

[0089] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0090] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A decision-making intelligent reasoning method based on a large model, characterized in that: The method includes the following steps: S1: Build a clinical diagnosis and treatment database and prepare case models; S2: Collect patient information features, including basic clinical information and relevant features in the patient's past medical records, and customize the patient diagnosis and treatment model; S3: Feature matching, retrieves the basic clinical information and related features of the input patient from the clinical diagnosis and treatment database, executes the rule retrieval mechanism, and retrieves multiple similar case models from the clinical database; S4: Optimize and rearrange. The case model rearrangement optimization mechanism is used as the execution mechanism to score and rearrange multiple similar case models retrieved, and output multiple optimized case models. S5: Collect new clinical characteristics of patients, collect information on various indicators of patients' clinical examinations, types of medications and dosages of medications, and customize patients' clinical characteristic models; S6: Secondary feature matching, using the collected new clinical features of patients as input variables, executes a data rule reasoning mechanism to select a suitable evaluation case model from multiple optimized case models; S7: Case model decision-making: Based on the specific opinions of the patient's attending physician and the selected evaluation case models, the clinical diagnosis and treatment case model that best matches the patient's characteristics is chosen.

2. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: The construction of the clinical diagnosis and treatment database in step S1 includes two parts: data model creation and data iteration and updating. Data model creation includes the collection part, which takes the patient's own characteristics as the center, the types and amounts of drugs used in the diagnosis and treatment process as clues, and prepares case models after the data is completely desensitized. It also includes the updating and iteration of the clinical diagnosis and treatment database, which is connected to the EMR / HIS system in various hospitals. After automatic desensitization, the database compares the feature information in the case models and classifies different case models into different departmental databases.

3. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: In step S2, the patient's past information is collected, and a patient diagnosis and treatment model is customized based on the patient's own characteristics and the characteristics during the diagnosis and treatment process.

4. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: Feature matching in step S3 refers to using the patient diagnosis and treatment model established in step S2 as a comparison file, executing a rule retrieval mechanism in the diagnosis and treatment database, and retrieving multiple similar case models. The rule-based retrieval mechanism refers to sequentially using the differences in specific features in the diagnostic and treatment model as input feature variables, comparing the input feature variables in the clinical diagnostic and treatment database, and filtering out past similar case models based on the matching degree of the input feature variables.

5. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: The optimization and reordering in step S4 refers to optimizing and reordering each of the selected past similar case models through the case model reordering optimization mechanism; The case model reordering optimization mechanism includes feature extraction, which extracts features from similar case models and uses the repeated features from multiple similar case models as input variables to feed them back into step S3. The rule enforcement retrieval mechanism therein uses these repeated features as input variables to re-select a batch of similar case models. All selected past case models are evaluated for multi-dimensional fit. After evaluation, they are initially sorted according to rules. The initial sorting is based on the scores. Local solutions are optimized and selected based on the selected case models. After optimization, the solutions are re-verified and re-sorted. The optimized case models are then selected based on the multi-dimensional fit scores.

6. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: In step S5, the new clinical characteristics of a patient refer to the characteristics that emerge after the patient is admitted to the hospital. The new clinical characteristics of the patient are used as the center and medication information is used as a clue to develop a patient clinical characteristic model.

7. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: In step S6, the clinical feature model established in step S5 is used as the input variable, and a data rule reasoning mechanism is executed to select a suitable evaluation case model from the optimized case model.

8. The decision-making intelligent reasoning method based on a large model according to claim 1, characterized in that: In step S7, external conditions are used as input variables, and the patient's attending physician comprehensively evaluates the case model to make the final appropriate clinical diagnosis and treatment decision.

9. A medical clinical diagnosis and treatment decision-making intelligent reasoning system based on a large model according to any one of claims 1-8, characterized in that: The system includes a clinical diagnosis and treatment database module, a feature matching module, an optimization and rearrangement module, a secondary feature matching module, and a case model decision module; The clinical diagnosis and treatment database module uses a unified representation to store various clinical diagnosis and treatment knowledge. The feature matching module takes the patient diagnosis and treatment model as input variable, executes the rule retrieval mechanism to output similar case models that match the features of new patients; The optimization and rearrangement module uses the common features extracted from the similar case model as input variables, executes the case model rearrangement optimization mechanism, and outputs an optimized case model with optimized features. The secondary feature matching module uses the patient's clinical feature model as input variable and executes a data rule reasoning mechanism to output a filtered and adapted evaluation case model. The case model decision module uses the specific external environment as input variables, executes the reasoning mechanism of the patient's attending physician, and makes the final appropriate clinical diagnosis and treatment decision by the patient's attending physician.

Images