A method and system for managing intelligent scientific research data on geriatric syndrome

Abstract

The present invention discloses a method and system for intelligent scientific research data management of geriatric syndrome, relating to the technical field of data processing. The present invention integrates structured data of hospital information systems, parses unstructured texts, extracts key entities and time series relationships, and obtains multi-source patient structured data and time series feature data. The present invention cleans, standardizes and terminologically normalizes the data based on a geriatric syndrome-specific domain ontology. A disease-specific data matrix is ​​generated by dragging and dropping variables, and scientific research cohorts are quickly screened in combination with Boolean logic grouping rules. The wild horse optimization algorithm is used to dynamically adjust the parameters of a random forest model, and a high-risk prediction model is trained based on historical data to output real-time high-risk cases. A scientific research report is generated by combining SHAP value quantification of variable contribution and statistical tests. The present invention solves the problems of medical research data silos, inconsistent terminology, and poor model interpretability, and achieves efficient integration, accurate prediction, and interpretable analysis of geriatric syndrome data.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and in particular, relates to a method and system for managing intelligent scientific research data on geriatric syndrome. Background Art

[0002] In the current environment, the prevention and treatment of geriatric syndromes (such as sarcopenia and cognitive impairment) has become a core challenge in the healthcare sector. Although hospital information systems (HIS, EMR, etc.) accumulate vast amounts of patient data, their heterogeneity (a mix of structured and unstructured data), inconsistent terminology, and lack of temporal features hinder their direct application in research and clinical decision-making. Existing technologies rely on manual data cleaning and static statistical models, resulting in long research cycles and insufficient predictive accuracy. Intelligent tools are needed to achieve data fusion, standardization, dynamic risk prediction, and interpretable analysis to improve the efficiency of geriatric research and the precision of clinical interventions.

[0003] Existing technologies have problems such as isolated medical research data, insufficient processing of unstructured data, inefficient research processes, and poor domain adaptability. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In response to the problems in the related art, the present invention provides an intelligent scientific research data management method for geriatric syndrome to overcome the above-mentioned technical problems existing in the existing related art.

[0006] (2) Technical solution

[0007] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0008] S1. Integrate the hospital's structured and unstructured data to obtain multi-source patient structured data and time series feature data;

[0009] S2. Construct a domain-specific ontology for geriatric syndromes; clean and normalize multi-source patient structured data based on the domain-specific ontology for geriatric syndromes to obtain standardized patient terminology data;

[0010] S3. Construct a disease-specific data matrix based on the patient's standardized terminology data, time series feature data, and research variable requirements;

[0011] Based on the case enrollment rules and the disease-specific data matrix, the target research medical record cohort data was obtained;

[0012] S4. Use historical medical record data combined with optimization algorithms to train and optimize the initial random forest model to obtain a high-risk medical record prediction model;

[0013] Input the target research medical record cohort data into the high-risk medical record prediction model to obtain real-time high-risk cases;

[0014] S5. Perform significance analysis on the disease-specific data matrix to obtain significance analysis results; perform explainable AI analysis on real-time high-risk cases to obtain variable contribution data;

[0015] Generate a scientific research analysis report based on the significance analysis results and variable contribution data;

[0016] By combining data processing and analysis with deep learning technology, this invention achieves efficient integration and standardization of multi-source medical data, rapid construction of scientific research cohorts and training of high-precision prediction models, as well as interpretable quantitative analysis of risk factors. This solves the problem of data silos, improves the efficiency of geriatric syndrome research, and provides a scientific basis for clinical decision-making that combines statistical significance and practical contribution.

[0017] Preferably, the S1 comprises the following steps:

[0018] S11. Setting a hospital information system set, wherein the hospital information system set includes various information systems of the hospital;

[0019] Connect to each information system in the hospital information system through a standardized interface to extract the patient's structured data and obtain the patient's structured data;

[0020] Perform format conversion on patient structured data to obtain patient structured data in a unified format;

[0021] S12. Use natural language processing technology to parse text data in unstructured data and extract key entities and temporal relationships to obtain structured entity data;

[0022] Use OCR technology to scan images and paper data in unstructured data to obtain the original text; call the AI ​​semantic model to perform secondary cleaning and structural processing on the original text to obtain structured field data;

[0023] The patient structured data, structured entity data, and structured field data in a unified format together constitute multi-source patient structured data;

[0024] S13. Integrate the device data in the unstructured data, store the dynamic monitoring indicators in the time series database, and associate them with the patient ID to obtain time series feature data;

[0025] The present invention integrates the structured data of various hospital systems through standardized interfaces and converts them into JSON / XML formats. It uses NLP to extract key entities and temporal relationships in unstructured texts, and combines OCR and AI models to process image / paper data to generate structured fields. Finally, it associates equipment monitoring indicators with patient IDs through a time series database to form multi-source structured data and time series feature data, thus solving the problem of data heterogeneity.

[0026] Preferably, said S2 comprises the following steps:

[0027] S21. Construct a domain-specific ontology for geriatric syndromes;

[0028] S22. Combined with the geriatric syndrome domain-specific ontology, data cleaning and normalization are performed on multi-source patient structured data to obtain standardized patient terminology data;

[0029] The present invention constructs a geriatric syndrome ontology and combines logical rules with the edit distance algorithm to clean multi-source data, thereby achieving terminology and unit standardization.

[0030] Preferably, the S21 includes the following steps:

[0031] S211. Define the core data element of geriatric syndrome in the geriatric syndrome domain ontology;

[0032] S212. Define the inter-entity relationships of the geriatric syndrome domain ontology through the knowledge graph and obtain logical constraint rules;

[0033] The logical constraint rules include time logic rules, numerical logic rules and semantic logic rules;

[0034] The present invention defines the core data elements of geriatric syndrome and establishes logical constraint rules between entities based on the knowledge graph to achieve data standardization and logical verification.

[0035] Preferably, the S22 includes the following steps:

[0036] S221. Use the temporal logic rules of the geriatric syndrome domain ontology to detect data that does not conform to temporal logic in multi-source patient structured data and trigger automatic correction or manual review;

[0037] S222. Use the numerical logic rules of the geriatric syndrome domain ontology to detect data that does not conform to numerical logic in multi-source patient structured data and trigger automatic correction or manual review;

[0038] S223. Use the semantic logic rules of the geriatric syndrome domain ontology and combine it with the edit distance algorithm to calculate semantic similarity, and convert non-standard terms and units in multi-source patient structured data into standard semantics;

[0039] S224. Obtain standardized terminology data through steps S221, S223, and S224.

[0040] The present invention detects anomalies in multi-source data through time, numerical value, and semantic logic rules, converts non-standard terms in combination with the edit distance algorithm, and finally generates standardized terminology data.

[0041] Preferably, the step S3 includes the following steps:

[0042] S31. Generate an electronic case report form by dragging and dropping variables based on the patient's standardized terminology data and time series feature data, combined with research requirement variables;

[0043] S32. Upload the electronic case report form to the Excel template and automatically map it to the system fields to obtain the disease-specific data matrix;

[0044] S33. Set the case enrollment rules; based on the case enrollment rules and the disease-specific data matrix, use Boolean logic to screen cases and target the medical record cohort;

[0045] The present invention dynamically generates electronic case report forms by dragging variables, automatically maps fields to generate disease-specific data matrices, and combines Boolean logic to quickly screen scientific research case cohorts that meet the inclusion rules, thereby improving research efficiency.

[0046] Preferably, the S4 comprises the following steps:

[0047] S41. Construct an initial random forest model, set the number of trees in the initial random forest model; set the training accuracy of the initial random forest model and the training accuracy threshold;

[0048] S42. Collect historical medical record data, where the historical case data includes each patient's physical data and high-risk labels;

[0049] S43. Using historical medical record data to train an initial random forest model. During the training process, an optimization algorithm is used to find the number of trees in the initial random forest model. The convergence of the initial random forest model is determined based on the training accuracy and the training accuracy threshold to obtain the optimal solution.

[0050] The optimal solution is used as the number of trees in the initial random forest model to obtain a high-risk medical record prediction model;

[0051] S44, inputting the target research medical record cohort data into the high-risk medical record prediction model to obtain real-time high-risk cases;

[0052] The present invention constructs an initial random forest model, trains it based on historical medical record data, uses an optimization algorithm to dynamically adjust the number of tree parameters, and combines convergence judgment to generate a high-precision prediction model. Ultimately, it outputs high-risk cases in the target cohort in real time, thereby improving prediction accuracy and clinical decision-making efficiency.

[0053] Preferably, the step of finding the number of trees of the initial random forest model by using an optimization algorithm in S43, and judging the convergence of the initial random forest model according to the training accuracy and the training accuracy threshold, to obtain the optimal solution comprises the following steps:

[0054] S431. Construct a wild horse population and set the size of the wild horse population; set the maximum number of optimization iterations;

[0055] S432. Randomly set the initial position of the wild horse population according to the number of trees in the initial random forest to obtain an initial set of wild horse populations;

[0056] S433. Defining a fitness function of an initial position of wild horses in a wild horse population according to the training accuracy and the training accuracy threshold of the initial random forest;

[0057] S434, performing an iterative operation on the initial position set of the wild horse population, calculating the fitness value of each position in the initial position set of the wild horse population according to the fitness function in each round of iteration, updating the position concentration of each wild horse in the initial position set of the wild horse population in descending order of fitness value, and obtaining the best individual wild horse position in the wild horse population and the global best wild horse position in each round of iteration;

[0058] S435, repeat S434, when the maximum number of optimization iterations is reached, stop the iteration, and take the global best wild horse position as the optimal solution;

[0059] The present invention simulates the behavior of wild horse populations, iteratively updates the individual position concentration based on the fitness function, dynamically adjusts the tree number parameter, and finally selects the global optimal solution as the number of trees in the random forest model to achieve automatic optimization of model hyperparameters and improve prediction accuracy and convergence efficiency.

[0060] Preferably, the S5 comprises the following steps:

[0061] S51. Divide the data in the disease-specific data matrix into continuous data and categorical data according to the data type;

[0062] S52. Use the t-test algorithm to perform significance analysis on the continuous data to obtain the continuous data analysis results;

[0063] S53. Use the chi-square test to perform significance analysis on the categorical data to obtain the categorical data analysis results;

[0064] S54. The continuous data analysis result and the categorical data analysis result together constitute a significance analysis result;

[0065] S55. Calculate the variable contribution of the real-time high-risk case through the SHAP value to obtain variable contribution data;

[0066] S56. Generate a scientific research analysis report based on the significance analysis results and the variable contribution data;

[0067] The present invention generates significance analysis results by analyzing continuous data through t-test and categorical data through chi-square test, combines SHAP value to quantify variable contribution, and finally integrates statistical significance and clinical impact weight to generate a scientific research report, providing a decision-making support basis with both statistical significance and interpretability for geriatric syndrome research.

[0068] An intelligent scientific research data management system for geriatric syndrome, used to implement the above-mentioned intelligent scientific research data management method for geriatric syndrome, comprising a multi-source medical data integration and feature extraction module, a domain ontology construction and data standardization module, a disease-specific data matrix construction and case screening module, a high-risk prediction model training and optimization module, and a scientific research analysis and report generation module;

[0069] The multi-source medical data integration and feature extraction module connects to the hospital information system through a standardized interface, integrates the patient's structured data, and uses natural language processing, OCR technology and time series database to parse unstructured data, extract key entities and time series features, and ultimately generate multi-source patient structured data and time series feature data in a unified format, providing standardized input for subsequent analysis;

[0070] The domain ontology construction and data standardization module builds a proprietary domain ontology based on the core data elements and knowledge graph of geriatric syndromes and defines logical constraint rules between entities. Through time logic detection, numerical logic correction, and semantic logic mapping, combined with the edit distance algorithm, multi-source data is converted into standardized terminology data to ensure data consistency and scientific research usability.

[0071] The disease-specific data matrix construction and case screening module dynamically generates electronic case report forms based on research requirement variables, constructs a disease-specific data matrix through Excel template mapping, and uses Boolean logic to screen target research cohorts in combination with case enrollment rules; converts standardized data into a research-oriented structured data set to support subsequent statistical analysis and model training;

[0072] The high-risk prediction model training and optimization module uses the random forest model as the basic model, combined with the Mustang optimization algorithm to dynamically adjust the number of tree parameters, trains the model through historical medical record data and evaluates convergence, and ultimately generates a high-risk medical record prediction model; the high-risk medical record prediction model achieves accurate prediction of real-time high-risk cases in the target cohort through feature selection and hyperparameter optimization;

[0073] The scientific research analysis and report generation module performs significance analysis and interpretability analysis on the disease-specific data matrix, identifies key variables, and generates a scientific research analysis report based on statistical results and variable contributions; it reveals the inherent correlation of the data through quantitative analysis, and provides statistical evidence and clinical decision-making support for geriatric syndrome research.

[0074] (3) Beneficial effects

[0075] The present invention has the following beneficial effects:

[0076] By combining data processing and analysis with deep learning technology, this invention achieves efficient integration and standardization of multi-source medical data, rapid construction of scientific research cohorts and training of high-precision prediction models, as well as interpretable quantitative analysis of risk factors. This solves the problem of data silos, improves the efficiency of geriatric syndrome research, and provides a scientific basis for clinical decision-making that combines statistical significance and practical contribution.

[0077] This invention integrates data from HIS, EMR and other systems through standardized interfaces, combines NLP and OCR technologies to parse unstructured text, and uses time-series databases to store dynamic monitoring indicators to solve the problem of medical data silos; the logical rules and edit distance algorithm of the geriatric syndrome domain-specific ontology realize the standardization of domain-specific terminology, improve data quality and consistency, and provide a reliable foundation for data analysis.

[0078] The present invention is based on the disease-specific data matrix generated by dragging variables and the Boolean logic inclusion rules to quickly screen target cases and shorten the scientific research data preparation cycle; combined with the Mustang optimization algorithm to dynamically adjust the number of trees in the random forest, and optimize the random forest model parameters through the fitness function, significantly improving the accuracy and convergence efficiency of high-risk predictions.

[0079] The present invention quantifies the contribution of variables through SHAP values ​​and generates a multidimensional scientific research report in combination with the significance results of t-test and chi-square test. The scientific research report not only reveals the statistical significance of the data, but also clarifies the influence weights of clinical variables, providing a decision-making basis with both statistical and clinical interpretability for mechanism research and risk intervention of geriatric syndromes.

[0080] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying any creative work.

[0082] Figure 1 This is a flow chart of an intelligent scientific research data management method for geriatric syndrome according to the present invention;

[0083] Figure 2 This is a flow chart of obtaining a high-risk medical record prediction model in an intelligent scientific research data management method for geriatric syndrome according to the present invention;

[0084] Figure 3 This is a schematic diagram of a process for obtaining significance analysis results in a method for managing intelligent scientific research data on geriatric syndrome according to the present invention;

[0085] Figure 4 This is a module diagram of an intelligent scientific research data management system for geriatric syndrome according to the present invention. DETAILED DESCRIPTION

[0086] The following will clearly and completely describe the technical solutions in the embodiments of the invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0087] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0088] Example 1

[0089] See also Figure 1 、 Figure 2 、 Figure 3 The present invention discloses a method for managing geriatric syndrome intelligent scientific research data, comprising the following steps:

[0090] S1. Integrate the hospital's structured and unstructured data to obtain multi-source patient structured data and time series feature data;

[0091] Said S1 comprises the following steps:

[0092] S11. Set up hospital information system ,in, a i Indicates the hospital's i information systems, b Indicates the total number of hospital information systems; the hospital information systems include HIS, EMR, LIS, and PACS;

[0093] Connecting to each information system in the hospital information system through a standardized interface, extracting the patient's structured data, and obtaining the patient's structured data; the patient's structured data includes basic information, diagnosis records, test reports, etc.;

[0094] Convert the patient structured data to a unified format; unify it into a preset JSON or XML format to ensure that the data fields are consistent with the target database mapping;

[0095] S12. Use natural language processing technology to parse text data in unstructured data and extract key entities and temporal relationships to obtain structured entity data; the text data includes medical records, medical history records, discharge summaries, etc.; the key entities include symptoms, disease names, and drug dosages; and the temporal relationships include "blood pressure dropped three days after taking the medication";

[0096] Scan images and paper data in unstructured data using OCR technology to obtain the original text; call the AI ​​semantic model to perform secondary cleaning and structural processing on the original text to obtain structured field data; the OCR technology mentioned above includes Tesseract and Alibaba Cloud OCR;

[0097] The patient structured data, structured entity data, and structured field data in a unified format together constitute multi-source patient structured data;

[0098] S13. Integrate device data in unstructured data, store dynamic monitoring indicators in a time series database, and associate them with patient IDs to obtain time series feature data; the sensor data may include muscle strength, gait speed, etc.;

[0099] S2. Construct a domain-specific ontology for geriatric syndromes; clean and normalize multi-source patient structured data based on the domain-specific ontology for geriatric syndromes to obtain standardized patient terminology data;

[0100] The S2 comprises the following steps:

[0101] S21. Construct a domain-specific ontology for geriatric syndromes;

[0102] The S21 includes the following steps:

[0103] S21. Define the core data elements of geriatric syndrome in the geriatric syndrome domain ontology; the geriatric syndromes include sarcopenia and cognitive impairment; the core data elements include Chinese name, English name, value range, and data level;

[0104] S22. Define the inter-entity relationships of the geriatric syndrome domain ontology through the knowledge graph to obtain logical constraint rules; for example, the inter-entity relationships include "sarcopenia diagnosis must include three indicators: muscle mass, grip strength, and gait speed";

[0105] The logical constraint rules include time logic rules, numerical logic rules and semantic logic rules;

[0106] S22. Combined with the geriatric syndrome domain-specific ontology, data cleaning and normalization are performed on multi-source patient structured data to obtain standardized patient terminology data;

[0107] The S22 includes the following steps:

[0108] S221. Use the temporal logic rules of the geriatric syndrome domain ontology to detect data that does not conform to temporal logic in multi-source patient structured data and trigger automatic correction or manual review; for example, detect "admission time is later than discharge time";

[0109] S222. Use the numerical logic rules of the geriatric syndrome domain ontology to detect data that does not conform to numerical logic in multi-source patient structured data and trigger automatic correction or manual review; for example, a height of 205 cm exceeds the numerical rules;

[0110] S223. Use the semantic logic rules of the geriatric syndrome domain ontology and combine it with the edit distance algorithm to calculate semantic similarity, and convert non-standard terms and units in multi-source patient structured data into standard semantics; for example, "senile dementia" is mapped to "Alzheimer's disease", and "pound" is converted to "kg". The edit distance algorithm formula is as follows:

[0111] ;

[0112] in, S i Indicates semantic similarity, s 1 represents non-standard terminology and unit data in multi-source patient structured data, s 2 represents the standard semantics in the geriatric syndrome domain ontology; ED ( s 1, s 2) Indicates s 1 and s The edit distance of 2 (about s 1 is converted to s 2 minimum number of editing operations required); express s 1 and s The maximum value of the data length is 2;

[0113] S224. Obtain standardized terminology data through steps S221, S223, and S224.

[0114] S3. Construct a disease-specific data matrix based on the patient's standardized terminology data, time series feature data, and research variable requirements;

[0115] Based on the case enrollment rules and the disease-specific data matrix, the target research medical record cohort data was obtained;

[0116] The S3 includes the following steps:

[0117] S31. Generate an electronic case report form by dragging and dropping variables based on the patient's standardized terminology data and time series feature data, combined with research requirement variables such as age, gait speed, and number of falls;

[0118] S32. Upload the electronic case report form to the Excel template and automatically map it to the system field to obtain the disease-specific data matrix C. The disease-specific data matrix is ​​as follows:

[0119] ;

[0120] in, C in Indicates the i The first disease n The data of each patient, m Indicates the total number of types of special diseases;

[0121] S33. Setting the case inclusion rules, such as age ≥ 65 years old;

[0122] Based on the case inclusion rules and the disease-specific data matrix, Boolean logic was used to select cases and target the medical record cohort;

[0123] S4. Use historical medical record data combined with optimization algorithms to train and optimize the initial random forest model to obtain a high-risk medical record prediction model;

[0124] Input the target research medical record cohort data into the high-risk medical record prediction model to obtain real-time high-risk cases;

[0125] The S4 comprises the following steps:

[0126] S41. Construct an initial random forest model, set the number of trees in the initial random forest model; set the training accuracy of the initial random forest model to α , the training accuracy threshold is β;

[0127] The single tree depth of the initial random forest model is 15 layers, the minimum number of samples for node splitting is 3, the minimum number of samples for leaf nodes is 2, and each tree randomly selects √n or log2(n) features;

[0128] S42. Collect historical medical record data, where the historical case data includes each patient's physical data and high-risk labels;

[0129] S43. Using historical medical record data to train an initial random forest model. During the training process, an optimization algorithm is used to find the number of trees in the initial random forest model. The convergence of the initial random forest model is determined based on the training accuracy and the training accuracy threshold to obtain the optimal solution.

[0130] The optimal solution is used as the number of trees in the initial random forest model to obtain a high-risk medical record prediction model;

[0131] In S43, the number of trees of the initial random forest model is found by the optimization algorithm, and the convergence of the initial random forest model is judged according to the training accuracy and the training accuracy threshold. Obtaining the optimal solution includes the following steps:

[0132] S431, build a wild horse population, set the size of the wild horse population to z , then the wild horse population is expressed as ,in, p i Indicates the first i A wild horse; set the maximum number of optimization iterations;

[0133] S432, randomly set the initial position of the wild horse population according to the number of trees in the initial random forest, and obtain the initial set of wild horse population as ,in q i Indicates the first i The initial position of the wild horse; the position of the wild horse can reflect the distance between the wild horse and the river;

[0134] S433, according to the training accuracy of the initial random forest α , training accuracy threshold β , define the fitness function of the initial position of the wild horse in the wild horse population. The fitness function formula is as follows,

[0135] ;

[0136] in ，g represents the fitness function, η Represents the bias (bias is used to help the model better fit the data);

[0137] S434, performing an iterative operation on the initial position set of the wild horse population, wherein the higher the fitness value, the closer the position of the wild horse is to the river; in each round of iteration, calculating the fitness value of each position in the initial position set of the wild horse population according to the fitness function, updating the position concentration of each wild horse in the initial position set of the wild horse population in descending order of fitness value, and obtaining the best individual wild horse position in the wild horse population and the global best wild horse position in each round of iteration;

[0138] S435, repeat S434, when the maximum number of optimization iterations is reached, stop the iteration, and take the global best wild horse position as the optimal solution;

[0139] S44, inputting the target research medical record cohort data into the high-risk medical record prediction model to obtain real-time high-risk cases;

[0140] S5. Perform significance analysis on the disease-specific data matrix to obtain significance analysis results; perform explainable AI analysis on real-time high-risk cases to obtain variable contribution data;

[0141] Generate a scientific research analysis report based on the significance analysis results and variable contribution data;

[0142] The S5 comprises the following steps:

[0143] S51. Divide the data in the disease-specific data matrix into continuous data and categorical data according to the data type;

[0144] S52. Use the t-test algorithm to perform significance analysis on the continuous data to obtain the continuous data analysis results; the t-test algorithm formula is as follows:

[0145] ;

[0146] in, t Indicates the results of continuous data analysis. X i 、 X j Represents the continuous data i The mean of the group data (e.g. the average walking speed of sarcopenia patients is 0.75 m / s) is the same as the j The mean of the group data (the average walking speed of non-sarcopenic patients was 0.95 m / s), k 2 i 、 k 2 j Represents the continuous data i The variance of the group data (e.g. the variance of the walking speed of patients with sarcopenia is 0.02 m² / s²) is the same as the variance of the firstj The variance of the group data (e.g. the variance of walking speed of non-sarcopenic patients is 0.01 m² / s²), h i 、 h j Represents the continuous data i The total amount of group data (e.g. the number of patients with sarcopenia is 100) is the same as the j The total amount of group data (e.g., the number of patients with sarcopenia is 100);

[0147] S53. Use the chi-square test to perform significance analysis on the categorical data to obtain the categorical data analysis results; the chi-square test formula is as follows:

[0148] ;

[0149] in, λ 2 Indicates the results of categorical data analysis; O de Indicates the observation frequency, for example, the actual number of patients with “sarcopenia and falls” observed was 45; E de Indicates the expected frequency, for example, the expected number of patients with “sarcopenia and falls” is 36;

[0150] S54. The continuous data analysis result and the categorical data analysis result together constitute a significance analysis result;

[0151] S55. Calculate the variable contribution of real-time high-risk cases using the SHAP value to obtain variable contribution data; for example, the contribution of walking speed to fall risk accounts for 30%;

[0152] S56. Generate a scientific research analysis report based on the significance analysis results and the variable contribution data.

[0153] Example 2

[0154] See also Figure 4 , an intelligent scientific research data management system for geriatric syndrome, used to implement the above-mentioned intelligent scientific research data management method for geriatric syndrome, including a multi-source medical data integration and feature extraction module, a domain ontology construction and data standardization module, a disease-specific data matrix construction and case screening module, a high-risk prediction model training and optimization module, and a scientific research analysis and report generation module;

[0155] The multi-source medical data integration and feature extraction module connects to the hospital information system through a standardized interface, integrates the patient's structured data, and uses natural language processing, OCR technology and time series database to parse unstructured data, extract key entities and time series features, and ultimately generate multi-source patient structured data and time series feature data in a unified format, providing standardized input for subsequent analysis;

[0156] The domain ontology construction and data standardization module builds a proprietary domain ontology based on the core data elements and knowledge graph of geriatric syndromes and defines logical constraint rules between entities. Through time logic detection, numerical logic correction, and semantic logic mapping, combined with the edit distance algorithm, multi-source data is converted into standardized terminology data to ensure data consistency and scientific research usability.

[0157] The disease-specific data matrix construction and case screening module dynamically generates electronic case report forms based on research requirement variables, constructs a disease-specific data matrix through Excel template mapping, and uses Boolean logic to screen target research cohorts in combination with case enrollment rules; converts standardized data into a research-oriented structured data set to support subsequent statistical analysis and model training;

[0158] The high-risk prediction model training and optimization module uses the random forest model as the basic model, combined with the Mustang optimization algorithm to dynamically adjust the number of tree parameters, trains the model through historical medical record data and evaluates convergence, and ultimately generates a high-risk medical record prediction model; the high-risk medical record prediction model achieves accurate prediction of real-time high-risk cases in the target cohort through feature selection and hyperparameter optimization;

[0159] The scientific research analysis and report generation module performs significance analysis and interpretability analysis on the disease-specific data matrix, identifies key variables, and generates a scientific research analysis report based on statistical results and variable contributions; it reveals the inherent correlation of the data through quantitative analysis, and provides statistical evidence and clinical decision-making support for geriatric syndrome research.

[0160] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0161] The preferred embodiments of the invention disclosed above are intended only to help illustrate the invention. These preferred embodiments do not exhaust all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for managing intelligent scientific research data on geriatric syndrome, characterized in that: The following steps are involved: S1. Integrate the hospital's structured and unstructured data to obtain multi-source patient structured data and time series feature data; Said S1 comprises the following steps: S11. Setting a hospital information system set, wherein the hospital information system set includes various information systems of the hospital; Connect to each information system in the hospital information system through a standardized interface to extract the patient's structured data and obtain the patient's structured data; Perform format conversion on patient structured data to obtain patient structured data in a unified format; S12. Use natural language processing technology to parse text data in unstructured data and extract key entities and temporal relationships to obtain structured entity data; Use OCR technology to scan images and paper data in unstructured data to obtain the original text; call the AI ​​semantic model to perform secondary cleaning and structural processing on the original text to obtain structured field data; The patient structured data, structured entity data, and structured field data in a unified format together constitute multi-source patient structured data; S13. Integrate the device data in the unstructured data, store the dynamic monitoring indicators in the time series database, and associate them with the patient ID to obtain time series feature data; S2. Construct a domain-specific ontology for geriatric syndromes; clean and normalize multi-source patient structured data based on the domain-specific ontology for geriatric syndromes to obtain standardized patient terminology data; S3. Construct a disease-specific data matrix based on the patient's standardized terminology data, time series feature data, and research variable requirements; Based on the case enrollment rules and disease-specific data matrix, the target research medical record cohort data was obtained; S4. Use historical medical record data combined with optimization algorithms to train and optimize the initial random forest model to obtain a high-risk medical record prediction model; Input the target research medical record cohort data into the high-risk medical record prediction model to obtain real-time high-risk cases; The S4 comprises the following steps: S43. Using historical medical record data to train the initial random forest model. During the training process, the number of trees in the initial random forest model is found through an optimization algorithm. The convergence of the initial random forest model is judged based on the training accuracy and the training accuracy threshold to obtain the optimal solution. The optimal solution is used as the number of trees in the initial random forest model to obtain a high-risk medical record prediction model; The S43 includes the following steps: S431. Construct a wild horse population and set the size of the wild horse population; set the maximum number of optimization iterations; S432. Randomly set the initial position of the wild horse population according to the number of trees in the initial random forest to obtain an initial set of wild horse populations; S433. Defining a fitness function of an initial position of wild horses in a wild horse population according to the training accuracy and the training accuracy threshold of the initial random forest; S434, performing an iterative operation on the initial position set of the wild horse population, calculating the fitness value of each position in the initial position set of the wild horse population according to the fitness function in each round of iteration, updating the position concentration of each wild horse in the initial position set of the wild horse population in descending order of fitness value, and obtaining the best individual wild horse position in the wild horse population and the global best wild horse position in each round of iteration; S435, repeat 434, when the maximum number of optimization iterations is reached, stop the iteration and take the global best wild horse position as the optimal solution S5. Perform a significance analysis on the disease-specific data matrix to obtain significance analysis results; perform an explainable AI analysis on real-time high-risk cases to obtain variable contribution data; Generate a scientific research analysis report based on the significance analysis results combined with the variable contribution data.

2. The method for managing geriatric syndrome intelligent scientific research data according to claim 1, characterized in that: The S2 comprises the following steps: S21. Construct a domain-specific ontology for geriatric syndromes; S22. Combined with the geriatric syndrome-specific domain ontology, multi-source patient structured data is cleaned and normalized to obtain patient standardized terminology data.

3. The method for managing geriatric syndrome intelligent scientific research data according to claim 2, characterized in that: The S21 includes the following steps: S211. Define the core data element of geriatric syndrome in the geriatric syndrome domain ontology; S212. Define the relationship between entities of the geriatric syndrome domain ontology through the knowledge graph and obtain logical constraint rules; The logic constraint rules include time logic rules, value logic rules and semantic logic rules.

4. The method for managing geriatric syndrome intelligent scientific research data according to claim 2, characterized in that: The S22 includes the following steps: S221. Use the temporal logic rules of the geriatric syndrome domain ontology to detect data that does not conform to temporal logic in multi-source patient structured data and trigger automatic correction or manual review; S222. Use the numerical logic rules of the geriatric syndrome domain ontology to detect data that does not conform to numerical logic in multi-source patient structured data and trigger automatic correction or manual review; S223. Use the semantic logic rules of the geriatric syndrome domain ontology and combine it with the edit distance algorithm to calculate semantic similarity, and convert non-standard terms and units in multi-source patient structured data into standard semantics; S224. Through steps S221, S223 and S223, obtain standardized terminology data.

5. The method for managing geriatric syndrome intelligent scientific research data according to claim 1, characterized in that: The S3 includes the following steps: S31. Generate an electronic case report form by dragging and dropping variables based on the patient's standardized terminology data and time series feature data, combined with research requirement variables; S32. Upload the electronic case report form to the Excel template and automatically map it to the system fields to obtain the disease-specific data matrix; S33. Set case inclusion rules; based on the case inclusion rules and the disease-specific data matrix, use Boolean logic to screen cases and target the medical record cohort.

6. The method for managing geriatric syndrome intelligent scientific research data according to claim 1, characterized in that: The S4 further comprises the following steps: S41. Construct an initial random forest model, set the number of trees in the initial random forest model; set the training accuracy of the initial random forest model and the training accuracy threshold; S42. Collect historical medical record data, which includes physical data and high-risk labels of each patient.

7. The method for managing geriatric syndrome intelligent scientific research data according to claim 1, characterized in that: The S5 comprises the following steps: S51. Divide the data in the disease-specific data matrix into continuous data and categorical data according to the data type; S52. Use the t-test algorithm to perform significance analysis on the continuous data to obtain the continuous data analysis results; S53. Use the chi-square test to perform significance analysis on the categorical data to obtain the categorical data analysis results; S54. The continuous data analysis result and the categorical data analysis result together constitute a significance analysis result; S55. Calculate the variable contribution of the real-time high-risk case through the SHAP value to obtain variable contribution data; S56. Generate a scientific research analysis report based on the significance analysis results and the variable contribution data.

8. An intelligent scientific research data management system for geriatric syndrome, characterized in that: A method for intelligent scientific research data management of geriatric syndrome as described in any one of claims 1 to 7 is implemented, wherein the system includes a multi-source medical data integration and feature extraction module, a domain ontology construction and data standardization module, a disease-specific data matrix construction and case screening module, a high-risk prediction model training and optimization module, and a scientific research analysis and report generation module.

Images