Algorithm library matching system and difference calculation method for complex chronic common disease scene

Abstract

The invention belongs to the technical field of medical data processing, and particularly relates to an algorithm library matching system and a difference calculation method for a complex chronic common disease scene, and the system comprises a hierarchical algorithm library construction module which is used for constructing a common disease algorithm library and realizing the structured storage of an algorithm module; meanwhile, relying on a real-time updating interface, newest clinical evidence and diagnosis and treatment practice data are accessed, and algorithm rules of all levels are dynamically optimized; and the multi-dimensional matching module is used for completing algorithm matching by utilizing three-stage matching and ensuring that a matching result covers disease generality, interaction characteristics and individual differences. According to the method, the hierarchical algorithm library of the three-level architecture of the base layer, the interaction layer and the individual layer is constructed, and the multi-dimensional matching module is combined with the extraction of the disease interaction characteristics and the accurate matching of the interaction rule module, so that the interaction relationship among multiple diseases is effectively incorporated, and the diagnosis and treatment scheme conflict caused by the matching of a single disease algorithm is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data processing, in particular to an algorithm library matching system and a difference value calculation method for complex chronic comorbidity scenarios. BACKGROUND

[0002] With the aggravation of population aging, the proportion of complex chronic comorbidity patients is increasing year by year. Such patients have the characteristics of "disease interaction (such as kidney disease aggravating diabetes drug metabolism burden), significant individual differences (such as elderly patients with higher drug sensitivity), and multiple dimensions of diagnosis and treatment", which have very high requirements for the accuracy of clinical decision-making.

[0003] The current algorithm library matching and difference value calculation scheme for comorbidity diagnosis and treatment in the clinic has the following key problems:

[0004] 1. Algorithm library matching is single and ignores disease interaction: existing systems mostly match based on single disease algorithm library and do not construct multi-disease interaction rules. For example, for "type 2 diabetes (T2DM) + hypertension (HBP) + chronic kidney disease (CKD2 stage) " patients, the system only calls T2DM hypoglycemic algorithm and HBP antihypertensive algorithm, without considering the interaction between "CKD leading to metformin prohibition" and "antihypertensive drugs (such as nifedipine) may increase blood glucose", resulting in a risk of drug conflict in the matched diagnosis and treatment scheme.

[0005] 2. Difference value calculation dimension is incomplete and does not cover individual differences: existing difference value calculation (such as drug dosage adjustment difference and efficacy target difference) is only based on single disease benchmark value and does not include patient individual characteristics (such as age, liver and kidney function, medication history) and disease interaction weight. For example, when calculating the insulin dosage difference for T2DM patients with CKD, only the "normal T2DM patient benchmark dose" is calculated without considering the "30% decrease in insulin clearance rate for CKD patients" and "20% reduction in dose for patients over 65 years old", resulting in an excessively high final dose and causing the risk of hypoglycemia.

[0006] 3. Algorithm library update lags behind and cannot adapt to dynamic disease: the disease of comorbidity patients changes dynamically over time (such as CKD progressing from stage 2 to stage 3), but existing algorithm libraries are mostly static, unable to update the matched algorithm modules in real time according to the latest test results of the patient (such as eGFR decline), resulting in a disconnection between the difference value calculation result and the actual disease.

[0007] Based on the above, an algorithm library matching system and a difference value calculation method for complex chronic comorbidity scenarios are invented. SUMMARY

[0008] To solve the above technical problems, the present application provides the following technical solutions:

[0009] An algorithm library matching system for a complex chronic comorbidity scenario, comprising:

[0010] A hierarchical algorithm library construction module for constructing a comorbidity algorithm library to realize structured storage of algorithm modules; and relying on a real-time updating interface, accessing the latest clinical evidence and diagnosis and treatment practice data to dynamically optimize algorithm rules at each level;

[0011] A multi-dimensional matching module for completing algorithm matching by three-level matching to ensure that the matching results cover disease commonality, interaction characteristics and individual differences; and introducing a weighted matching degree scoring mechanism to quality check the matching results, and triggering expert intervention for matching results that do not meet the standards to ensure the safety and applicability of the output algorithm combination;

[0012] A comorbidity prognosis prediction module for predicting the short-term multi-type deterioration risk of comorbidity patients by a multi-task learning model based on the matched algorithm combination and multi-modal fusion features; and grading the risk results and generating corresponding intervention strategies.

[0013] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the hierarchical algorithm library construction module comprises:

[0014] A comorbidity algorithm library construction module for constructing a comorbidity algorithm library using a three-level architecture of a basic layer, an interaction layer and an individual layer;

[0015] An algorithm library association module for associating the algorithm library with a real-time updating interface to enable access to the latest clinical guidelines, multi-center comorbidity diagnosis and treatment data, and automatic updating of algorithm module parameters at each level every month.

[0016] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the basic layer is a standard algorithm module for storing a single chronic disease, including disease diagnosis algorithms, basic treatment algorithms and a benchmark parameter library.

[0017] The interaction layer is a multi-disease interaction rule algorithm module constructed based on evidence-based medical evidence and machine learning models.

[0018] The individual layer is a personalized adjustment algorithm module constructed based on patient individual characteristics.

[0019] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the multi-dimensional matching module comprises:

[0020] An information input module for accessing patient electronic medical records, laboratory examination data and medication history data through an interface, and supporting manual supplement of special information;

[0021] A multi-modal data fusion module is used to adapt multi-source data interfaces and collect full-dimensional data; and then through an algorithm, fragmented data is converted into an integrated fusion feature set of short-term fluctuations, long-term trends and individual differences;

[0022] A feature extraction module is used to automatically extract basic disease combination features, disease interaction features and individual features;

[0023] A hierarchical matching module is used to perform first-level matching, second-level matching and third-level matching;

[0024] A matching degree scoring module is used to calculate the total matching degree by using a weighted scoring method, and only when the total matching degree is greater than or equal to 85%, the matched algorithm combination is output; if less than 85%, an expert intervention interface is triggered to invite a comorbidity diagnosis and treatment expert to supplement the rules and then re-match.

[0025] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the multi-modal data fusion module comprises:

[0026] A multi-modal data access interface module is used to support dynamic physiological data, medical image data and patient subjective data access;

[0027] A data cleaning and fusion algorithm module is used to first match dynamic physiological data and laboratory data according to a time axis to generate a joint data set of short-term fluctuations and long-term trends; then, based on an isolation forest algorithm, false data of a wearable device is removed; and then, through an attention mechanism, different data is assigned a weight, and a multi-dimensional fusion feature set is output and directly transmitted to a subsequent feature extraction module.

[0028] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the dynamic physiological data is obtained by connecting a wearable device through Bluetooth / 5G, and patient data is obtained in real time;

[0029] The medical image data is accessed through a DICOM interface to access comorbidity-related images and extract image features;

[0030] The patient subjective data is collected through a WeChat applet / hospital APP to collect patient self-rating scales and quantify patient quality of life and symptom feedback.

[0031] As a preferred scheme of the algorithm library matching system for a complex chronic comorbidity scenario, the first-level matching is based on a disease combination algorithm group corresponding to a basic disease combination feature matching algorithm library basic layer;

[0032] The second-level matching is based on a multi-disease interaction rule algorithm module of an interaction layer based on disease interaction features;

[0033] The third level matching is based on individual characteristic matching individual layer personalized adjustment algorithm module.

[0034] As a preferred scheme of the algorithm library matching system of a complex chronic comorbidity scenario according to the application, wherein: the comorbidity prognosis prediction module comprises:

[0035] The prognosis prediction model construction module is used for training a multi-task learning model according to historical diagnosis and treatment data of the hierarchical algorithm library, and simultaneously predicting metabolic risk, organ damage risk and cardiovascular risk.

[0036] The risk grading and intervention suggestion generation module is used for dividing the prediction results into three levels of low risk, medium risk and high risk, and generating adjustment rules correspondingly.

[0037] A difference calculation method for a complex chronic comorbidity scenario, comprising the following specific steps:

[0038] S1: Based on the matched algorithm combination, a three-dimensional difference calculation model of benchmark value-interaction influence value-individual influence value is adopted, and the specific formula is as follows:

[0039] Final target difference / dose difference = basic layer benchmark value - interaction layer influence value x interaction weight - individual layer influence value x individual weight

[0040] Wherein, the basic layer benchmark value refers to a single disease standard parameter extracted from the basic layer; the interaction layer influence value refers to an adjustment amount of the benchmark value from the interaction of multiple diseases in the interaction layer; the interaction weight refers to a setting according to the disease interaction intensity; the individual layer influence value refers to an adjustment amount of the benchmark value from the individual characteristics in the individual layer; and the individual weight refers to a setting according to the individual characteristic influence intensity.

[0041] S2: Real-time monitoring of the change of the patient's examination results, automatic updating of the interaction influence value, individual influence value and corresponding weight, and dynamic adjustment of the difference calculation result.

[0042] Compared with the prior art:

[0043] 1. By constructing a hierarchical algorithm library with a three-level architecture of basic layer-interaction layer-individual layer, combining the extraction of disease interaction characteristics by the multi-dimensional matching module and the accurate matching of the interaction rule module, the effective inclusion of the interaction relationship between multiple diseases is realized, and the conflict of diagnosis and treatment scheme caused by single disease algorithm matching is avoided.

[0044] 2. By dynamic difference calculation, a three-dimensional calculation model of benchmark value-interaction influence value-individual influence value is adopted, the disease interaction weight and the patient individual characteristic weight are introduced, the comprehensive coverage of the difference calculation process to the patient individual characteristics and the disease interaction influence is realized, the limitations of single disease benchmark value calculation are made up, and the diagnosis and treatment parameter adaptability is improved.

[0045] 3. Access the latest clinical guidelines and multi-center diagnosis and treatment data through the real-time update interface associated with the hierarchical algorithm library, combine dynamic difference calculation, and automatically adjust the real-time monitoring and calculation parameters of the latest test results of the patient, so as to realize the dynamic adaptation of the algorithm library rules and difference calculation results with the changes of the patient's condition, and avoid the disconnection between the static algorithm system and the actual condition. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 It is a schematic diagram of the overall framework of the application;

[0047] Figure 2 It is a schematic diagram of the hierarchical algorithm library construction module framework of the application;

[0048] Figure 3 It is a schematic diagram of the multi-dimensional matching module framework of the application;

[0049] Figure 4 It is a schematic diagram of the comorbidity prognosis prediction module framework of the application;

[0050] Figure 5 It is a schematic diagram of the multi-modal data fusion module framework of the application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical scheme and advantages of the application clearer, the embodiments of the application will be further described in detail below with reference to the drawings.

[0052] The application provides an algorithm library matching system for complex chronic comorbidity scenarios, please refer to Figures 1-5 , which comprises:

[0053] The hierarchical algorithm library construction module is used to construct the comorbidity algorithm library and realize the structured storage of the algorithm module; at the same time, relying on the real-time update interface, the latest clinical evidence and diagnosis and treatment practice data are accessed, and the algorithm rules of each level are dynamically optimized;

[0054] The multi-dimensional matching module is used to complete algorithm matching by three-level matching, to ensure that the matching results cover disease commonness, interaction characteristics and individual differences; at the same time, the weighted matching degree scoring mechanism is introduced to quality check the matching results, and the matching results that do not meet the standard trigger expert intervention, to ensure the safety and applicability of the output algorithm combination;

[0055] The comorbidity prognosis prediction module is used to predict the short-term multi-type deterioration risk of comorbidity patients through a multi-task learning model according to the matched algorithm combination and multi-modal fusion features; then the risk results are classified, and the corresponding intervention strategies are generated.

[0056] The hierarchical algorithm library construction module comprises:

[0057] A comorbidity algorithm library construction module is configured to construct a comorbidity algorithm library in a three-level architecture of a basic layer, an interaction layer and an individual layer;

[0058] The basic layer is a standard algorithm module for storing a single chronic disease, including a disease diagnosis algorithm (such as an HbA1c diagnosis standard for T2DM), a basic treatment algorithm (such as an initial antihypertensive drug selection algorithm for HBP), and a benchmark parameter library (such as efficacy target values and initial drug doses for different diseases);

[0059] The interaction layer is an algorithm module for storing multi-disease interaction rules based on evidence-based medical evidence (such as the Comorbidity Drug Guide) and machine learning models (such as random forests), for example:

[0060] A “T2DM+CKD” interaction module stores rules such as “CKD patients in stage 2-3 are prohibited from using metformin” and “insulin doses need to be reduced by 15%-30% when eGFR is less than 60”;

[0061] A “HBP+T2DM” interaction module stores rules such as “monthly blood glucose monitoring is required when using calcium channel blockers (such as nifedipine)” and “ACEI drugs (such as enalapril) can improve kidney function in T2DM patients”;

[0062] The individual layer is an individualized adjustment algorithm module based on patient individual characteristics (age, gender, liver and kidney function, medication history, allergy history), such as rules such as “drug doses for patients over 65 years old need to be reduced by 10%-20%” and “statins are prohibited when liver enzymes (ALT) are greater than 3 times the normal value”;

[0063] An algorithm library association module is configured to associate the algorithm library with a real-time update interface to enable access to the latest clinical guidelines (such as the annual update of the China Type 2 Diabetes Prevention and Treatment Guidelines) and multi-center comorbidity diagnosis and treatment data (such as prognosis data for comorbid patients in top-tier hospitals), and to automatically update the parameters of the algorithm modules in each layer every month.

[0064] The multi-dimensional matching module includes:

[0065] An information input module is configured to access patient electronic medical records (EMR), laboratory test data (such as HbA1c, eGFR, blood pressure values), medication history data through an interface, and to support manual addition of special information (such as drug allergy history);

[0066] A multi-modal data fusion module is configured to adapt to multi-source data interfaces to collect full-dimensional data; and to convert fragmented data into an integrated fusion feature set of short-term fluctuations, long-term trends and individual differences through algorithms;

[0067] The feature extraction module is used for automatically extracting basic disease combination features (such as "T2DM+HBP+CKD"), disease interaction features (such as "CKD2 and use of nifedipine"), and individual features (such as "65-year-old male, eGFR=62 ml / min / 1.73 m2");

[0068] The hierarchical matching module is used for first-level matching, second-level matching, and third-level matching.

[0069] The first-level matching is based on a disease combination algorithm group corresponding to a basic layer of a basic disease combination feature matching algorithm library (such as matching three single disease basic algorithms corresponding to "T2DM+HBP+CKD");

[0070] The second-level matching is based on a multi-disease interaction rule algorithm module of an interaction layer matched based on disease interaction features (such as matching drug contraindications and efficacy influence rules of "T2DM+HBP+CKD");

[0071] The third-level matching is based on a personalized adjustment algorithm module of an individual layer matched based on individual features (such as matching a dose adjustment rule corresponding to "65 years old+eGFR=62");

[0072] The matching degree scoring module is used for calculating a total matching degree by using a weighted scoring method (basic layer weight 40%, interaction layer weight 35%, and individual layer weight 25%), and outputting a matched algorithm combination only when the total matching degree is greater than or equal to 85%; if less than 85%, triggering an expert intervention interface to invite a comorbidity diagnosis and treatment expert to supplement rules and then re-match.

[0073] The multi-modal data fusion module includes:

[0074] The multi-modal data access interface module is used for supporting access of dynamic physiological data, medical image data, and patient subjective data;

[0075] The dynamic physiological data is obtained by connecting a wearable device (such as a smart blood pressure meter, a dynamic blood glucose monitor (CGM), and a heart rate bracelet) through Bluetooth / 5G, and real-time acquisition of 24-hour blood pressure fluctuation, blood glucose curve, sleep quality, and other data of a patient;

[0076] The medical image data is accessed through a DICOM interface to access comorbidity-related images (such as kidney ultrasound of a chronic kidney disease patient and coronary CTA of a coronary heart disease patient), and image features such as "kidney cortex thickness" and "coronary stenosis degree" are extracted;

[0077] The patient subjective data is collected through a WeChat applet / hospital APP to collect patient self-rating scales (such as foot sensation scores of a diabetes patient and dizziness frequency scores of a hypertension patient), and quantifying patient quality of life and symptom feedback;

[0078] The data cleaning and fusion algorithm module is used to match the dynamic physiological data (such as blood glucose value every 5 minutes) with the laboratory data (such as HbA1c every week) according to the time axis, to generate a joint data set of short-term fluctuations and long-term trends; then, based on the isolation forest algorithm, the false data of the wearable device (such as the abnormally high value caused by the falling of the sphygmomanometer) is removed; then, through the attention mechanism (Attention), different data are assigned weights (such as the weight of CGM blood glucose fluctuation data for matching the insulin dose is 0.3, and the weight of kidney ultrasound data for kidney function protection algorithm is 0.4), and a multi-dimensional fusion feature set is output, which is directly transmitted to the subsequent feature extraction module.

[0079] The comorbidity prognosis prediction module comprises:

[0080] The prognosis prediction model construction module is used to train a multi-task learning model according to the historical diagnosis and treatment data (such as 5-year comorbidity patient prognosis data) of the hierarchical algorithm library, and simultaneously predict metabolic risk, organ damage risk and cardiovascular risk.

[0081] Metabolic risk: such as the hypoglycemia risk of T2DM+HBP patients (input features: insulin dose, CGM fluctuation, liver and kidney function);

[0082] Organ damage risk: such as the eGFR decline risk of CKD+T2DM patients (input features: kidney ultrasound data, urine protein quantification, blood pressure control situation);

[0083] Cardiovascular risk: such as the myocardial infarction risk of coronary heart disease+hyperlipidemia patients (input features: blood lipid target rate, coronary stenosis degree, heart rate variability);

[0084] The risk grading and intervention suggestion generation module is used to divide the prediction results into three levels of low risk (<10%), medium risk (10%-30%) and high risk (>30%), and generate adjustment rules correspondingly;

[0085] The high-risk patient: increase the "conservative coefficient" in dynamic difference calculation (such as reducing the insulin dose by 10% and relaxing the blood pressure control target by 10 / 5 mmHg);

[0086] The medium-risk patient: after the difference calculation result, the "monitoring frequency suggestion" is added (such as monitoring blood glucose 3 times a week and rechecking kidney function every 2 weeks);

[0087] The low-risk patient: execute according to the conventional difference calculation result, and only need monthly routine monitoring.

[0088] A difference calculation method for complex chronic comorbidity scenarios, comprising the following specific steps:

[0089] S1: Based on the matched algorithm combination, the three-dimensional difference calculation model of benchmark value-interaction value-individual value is adopted, and the specific formula is as follows:

[0090] Final target difference / dose difference = basic layer benchmark value-interaction layer influence value x interaction weight-individual layer influence value x individual weight

[0091] Wherein, the basic layer benchmark value refers to the single disease standard parameter (such as the HbA1c control target benchmark value 7.0% of T2DM patients and the insulin initial dose benchmark value 0.5U / kg / d) extracted from the basic layer; the interaction layer influence value refers to the adjustment amount of the benchmark value of multiple disease interactions (such as the HbA1c control target of “T2DM+CKD2” patients can be relaxed by 0.5%, and the insulin dose needs to be reduced by 0.1U / kg / d); the interaction weight refers to the setting according to the disease interaction intensity (such as “the influence weight of CKD on T2DM drug metabolism is 0.8” “the influence weight of HBP on T2DM blood glucose is 0.3”); the individual layer influence value refers to the adjustment amount of the benchmark value of individual characteristics (such as “the insulin dose of 65-year-old patients needs to be reduced by 0.05U / kg / d” “the statin dose of ALT>2 times normal value patients needs to be reduced by 50%”); the individual weight refers to the setting according to the individual characteristic influence intensity (such as “the individual weight of eGFR<60 is 0.9” “the individual weight of age>70 years old is 0.7”);

[0092] S2: Real-time monitoring of the change of the patient's examination results (such as eGFR from 62 to 58), automatic updating of interaction influence value, individual influence value and corresponding weight, dynamic adjustment of difference calculation result.

[0093] Although the present application has been described above with reference to the implementation, various improvements can be made and equivalent parts can be substituted without departing from the scope of the present application. In particular, each feature in the disclosed embodiments of the present application can be combined with any other feature in any manner, as long as there is no structural conflict. The combinations are not exhaustively described in the specification only for the purpose of omitting the length and saving resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. An algorithm library matching system for complex chronic comorbidity scenarios, characterized by, Comprising A hierarchical algorithm library construction module for constructing a comorbidity algorithm library and realizing structured storage of algorithm modules; meanwhile, relying on a real-time updating interface, the latest clinical evidence and diagnosis and treatment practice data are accessed, and algorithm rules at each level are dynamically optimized; A multi-dimensional matching module for completing algorithm matching by using three-level matching to ensure that the matching results cover disease commonality, interaction characteristics and individual differences; meanwhile, a weighted matching degree scoring mechanism is introduced to quality check the matching results, and expert intervention is triggered for matching results that do not meet the standard to guarantee the safety and applicability of the output algorithm combination; A comorbidity prognosis prediction module for predicting, by a multi-task learning model, short-term multi-type deterioration risks of a comorbidity patient simultaneously according to the matched algorithm combination and multi-modal fusion features; and grading the risk results and generating corresponding intervention strategies.

2. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 1, wherein, The hierarchical algorithm library construction module comprises: A comorbidity algorithm library construction module for constructing a comorbidity algorithm library by using a three-level architecture of a basic layer, an interaction layer and an individual layer; An algorithm library association module for associating the algorithm library with a real-time updating interface so as to access the latest clinical guidelines, multi-center comorbidity diagnosis and treatment data and automatically update parameters of algorithm modules at each level every month.

3. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 2, wherein, The basic layer is a standard algorithm module for storing a single chronic disease, including disease diagnosis algorithms, basic treatment algorithms and benchmark parameter libraries; The interaction layer is a multi-disease interaction rule algorithm module constructed based on evidence-based medical evidence and machine learning models; The individual layer is a personalized adjustment algorithm module constructed based on patient individual characteristics.

4. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 1, wherein, The multi-dimensional matching module comprises: An information input module for accessing patient electronic medical records, laboratory examination data and medication history data through an interface and supporting manual supplement of special information; A multi-modal data fusion module for adapting multi-source data interfaces to collect full-dimensional data; and then converting fragmented data into integrated fusion feature sets of short-term fluctuations, long-term trends and individual differences by using algorithms; A feature extraction module for automatically extracting basic disease combination features, disease interaction features and individual features; A hierarchical matching module for first-level matching, second-level matching and third-level matching; A matching degree scoring module for calculating a total matching degree by using a weighted scoring method, and outputting a matched algorithm combination only when the total matching degree is greater than or equal to 85%; if the total matching degree is less than 85%, an expert intervention interface is triggered to invite a comorbidity diagnosis and treatment expert to supplement rules and then re-match.

5. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 4, wherein, The multi-modal data fusion module comprises: A multi-modal data access interface module for supporting access of dynamic physiological data, medical image data and patient subjective data; A data cleaning and fusion algorithm module for first matching dynamic physiological data and laboratory data according to a time axis to generate a joint data set of short-term fluctuations and long-term trends; then eliminating false data of a wearable device based on an isolation forest algorithm; and then assigning weights to different data by using an attention mechanism to output a multi-dimensional fusion feature set and directly transmit the multi-dimensional fusion feature set to a subsequent feature extraction module.

6. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 5, wherein, The dynamic physiological data are obtained in real time by connecting a wearable device through Bluetooth / 5G; The medical image data are accessed through a DICOM interface to access comorbidity-related images and extract image features. The patient subjective data is collected by a WeChat applet / hospital APP, and the patient self-rating scale is used to quantify the patient's quality of life and symptom feedback.

7. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 4, wherein, The first level matching is based on the disease combination algorithm group corresponding to the basic layer of the basic disease combination feature matching algorithm library; The second level matching is based on the multi-disease interaction rule algorithm module of the disease interaction feature matching interaction layer; The third level matching is based on the individual characteristic matching individual layer personalized adjustment algorithm module.

8. The algorithmic library matching system for complex chronic comorbidity scenarios of claim 1, wherein, The comorbidity prognosis prediction module comprises: A prognosis prediction model construction module is used to train a multi-task learning model according to historical diagnosis and treatment data of the hierarchical algorithm library, and to simultaneously predict metabolic risk, organ damage risk and cardiovascular risk. A risk classification and intervention suggestion generation module is used to divide the prediction results into three levels of low risk, medium risk and high risk, and to generate adjustment rules accordingly.

9. A method for calculating difference values of complex chronic comorbidity scenarios, characterized in that The specific steps are as follows: S1: Based on the matched algorithm combination, a three-dimensional difference calculation model of benchmark value-interactive influence value-individual influence value is used, and the specific formula is as follows: Final target difference / dose difference=baseline value of basic layer-interaction influence value×interaction weight-individual influence value×individual weight Wherein, the baseline value of the basic layer refers to the single disease standard parameter extracted from the basic layer; the interaction influence value refers to the adjustment amount of the baseline value of multiple disease interactions extracted from the interaction layer; the interaction weight refers to the setting according to the disease interaction intensity; the individual influence value refers to the adjustment amount of the baseline value of the individual characteristics extracted from the individual layer; and the individual weight refers to the setting according to the individual characteristic influence intensity; S2: Real-time monitoring of patient test results changes, automatic updating of interaction influence value, individual influence value and corresponding weight, dynamic adjustment of difference calculation results.

Images