Standardized fusion and comprehensive information model construction system for clinical data

Abstract

The invention discloses a clinical data standardized fusion and comprehensive information model construction system, which comprises an infectious lung disease clinical queue data acquisition system used for establishing a unified data acquisition system and standard based on a prospective infectious lung disease special disease research queue of a cooperative unit, clinical indexes, laboratory examination data, iconography data and treatment response multi-dimensional data are collected and integrated, and the data are collected, processed and stored. According to the clinical data standardized fusion and comprehensive information model construction system, by integrating multi-modal data and constructing an advanced model, the death rate of infectious lung disease patients, the ICU admission rate, the breathing machine requirement and the like can be more accurately predicted, and accurate AI auxiliary decision support is provided for critical disease treatment.

Description

Technical Field

[0001] The present invention belongs to the technical field of clinical data, and particularly relates to a system for standardizing the integration of clinical data and constructing a comprehensive information model. Background Art

[0002] Infectious pulmonary diseases are a major public health challenge threatening human health. Each year, a large number of people die from this disease globally, and the incidence rate in China is relatively high and shows an upward trend. Currently, the diagnosis and treatment of infectious pulmonary diseases face challenges such as time urgency, disease complexity, and individual differences. Traditional assessment tools have problems such as the lag of the scoring system and the limitations of single markers.

[0003] With the development of artificial intelligence technology, AI4S provides a path to solve these problems. However, when applying it to clinical practice, key issues such as data integration, model construction, and clinical verification still need to be solved. Although there have been some progress in related research, there is still room for improvement in data integration, model optimization, etc.

[0004] Therefore, the present invention provides a system for standardizing the integration of clinical data and constructing a comprehensive information model. Summary of the Invention

[0005] The object of the present invention is to provide a system for standardizing the integration of clinical data and constructing a comprehensive information model to solve the technical problems raised in the background art.

[0006] To achieve the above object, the specific technical solution of the present invention is as follows: A system for standardizing the integration of clinical data and constructing a comprehensive information model, including an infectious pulmonary disease clinical cohort data collection system, which is used to establish a unified data collection system and standards based on the prospective infectious pulmonary disease specialty research cohort of cooperative units, collect and integrate multi-dimensional data such as clinical indicators, laboratory examination data, imaging data, and treatment responses, collect, process, and store the data, including converting laboratory examination data into international standard units, storing imaging data in DICOM format, cleaning and standardizing the data, storing it using a hybrid mode architecture database, and setting user permission management and data encryption;

[0007] A multi-modal deep algorithm model construction module, including:

[0008] A multi-modal fusion model that normalizes, augments data, extracts features, and reduces the dimensionality of imaging data; preprocesses text data, performs deep semantic encoding, and unifies dimensions; standardizes and extracts features from structured data; uses a GBM model to perform regression and classification predictions on multi-timepoint data; uses AFD to extract features from time series data and inputs them into an LSTM model; uses a large language model to extract features and establish associations between text information and laboratory results; and fuses multi-modal features through a self-attention mechanism.

[0009] A multi-task output layer that constructs a multi-task learning framework to concurrently execute the risk prediction of infectious pulmonary diseases and the decision support task based on Task 1. The risk prediction task determines the high-risk state of severe pneumonia in patients and evaluates the necessity of mechanical ventilation. The decision support task combines a large language model and an expert knowledge graph to assist doctors in making decisions regarding ICU requirements.

[0010] A clinical performance verification module, including:

[0011] Clinical enrollment: Relying on partner hospitals, patients with infectious pulmonary diseases and non-infectious controls are collected according to specific time ranges and inclusion and exclusion criteria. Humoral samples are longitudinally collected from "the entire course of the disease, multiple time points, and multiple sites" and standardized management is carried out.

[0012] Prospective cohort data integration and management: Relying on the hospital's advantages, a unified data processing process is established to ensure data quality, remove redundant information, protect privacy, and standardize data processing.

[0013] Preferably, in the clinical cohort data collection system for infectious pulmonary diseases, the data collection scope covers the patient's baseline information, intervention measures, and clinical outcomes, and additional metadata such as data source, collection time, and disease type are added.

[0014] Preferably, in the multi-modal fusion model, for the extraction of imaging data features, a pre-trained EfficientNet model combined with principal component analysis is used; for the extraction of text data features, a BioBERT model combined with average pooling operation is used; for the extraction of structured data features, a random forest model and a support vector machine method are used.

[0015] Preferably, in the multi-task output layer, for the risk prediction task, the fused multi-modal features are combined with imaging lesions, clinical symptoms, and laboratory data, and the classifier outputs the high-risk state of severe pneumonia in patients and the evaluation results of the necessity of mechanical ventilation; for the decision support task, the large language model semantically analyzes the risk prediction results and provides decision-making suggestions in combination with the knowledge graph.

[0016] Preferably, in the clinical performance verification module, the inclusion criteria for infectious pulmonary disease patients in clinical enrollment include meeting the diagnostic criteria for community-acquired pneumonia, signing an informed consent form, and being less than 80 years old; the inclusion criteria for patients without lung infections include being less than 80 years old, having no serious underlying diseases, and not participating in relevant clinical trials.

[0017] The system for standardizing the integration of clinical data and constructing a comprehensive information model of the present invention has the following advantages:

[0018] By integrating multi-modal data and constructing an advanced model, the system for standardizing the integration of clinical data and constructing a comprehensive information model can more accurately predict the mortality rate, ICU admission rate, and ventilator requirements of infectious pulmonary disease patients, providing precise AI-assisted decision-making support for the treatment of critically ill patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is the overall system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In the following text, only some exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and the description are considered to be exemplary in nature rather than restrictive.

[0022] In the description of the embodiments of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "vertical", "horizontal", "top", "bottom", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the embodiments of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present invention.

[0023] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, "a plurality" means two or more unless otherwise specifically defined.

[0024] In the embodiments of the present invention, unless otherwise clearly defined and limited, terms such as "installation", "connection", "connection", "fixation" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection, an electrical connection, or a communication; it can be directly connected, or indirectly connected through an intermediate medium, and can be the communication inside two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific situations.

[0025] The following disclosure provides many different embodiments or examples for implementing different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, the components and settings of specific examples are described below. Of course, they are only examples and are not intended to limit the embodiments of the present invention. In addition, the embodiments of the present invention may repeat reference numerals and / or reference letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0026] In order to better understand the purpose, structure and function of the present invention, the following further describes in detail a system for standardizing the integration of clinical data and constructing a comprehensive information model of the present invention with reference to the accompanying drawings.

[0027] As Figure 1 shown, a system for standardizing the integration of clinical data and constructing a comprehensive information model of the present invention includes an infectious pulmonary disease clinical cohort data acquisition system, which is used to establish a unified data acquisition system and standard based on the prospective infectious pulmonary disease specialized research cohort of cooperative units, collect and integrate multi-dimensional data such as clinical indicators, laboratory test data, imaging data, and treatment responses, collect, process and store the data, including converting laboratory test data into international standard units, storing imaging data in DICOM format, cleaning and standardizing the data, storing it in a hybrid mode architecture database, and setting user permission management and data encryption;

[0028] In the infectious pulmonary disease clinical cohort data acquisition system, the data acquisition scope covers patient baseline information, intervention measures and clinical outcomes, and additional metadata such as data source, acquisition time, and disease type.

[0029] The specific embodiments of the infectious pulmonary disease clinical cohort data acquisition system are as follows:

[0030] Clinical indicators: Obtain the clinical information of patients from the electronic medical record system (EMR) of the collaborating unit, including demographic information (such as age A, gender G), symptom information (such as cough duration t, cough volume V), and symptoms can be collected through questionnaires or doctor records. For symptom scoring, relevant scales can be used. For example, for the cough severity score S, different score ranges can be set according to different symptom degrees;

[0031] Laboratory test data: Collect the results of various laboratory tests, such as white blood cell count WBG. If the unit is 10 9 / L, but some units are 10 3 / uL. When converting to the international standard unit, use the formula: When the original unit is 10 3 / uL, WBG std =WBG original / 1000; For biochemical indicators, such as serum creatinine C T , unify the results of different detection methods into the internationally recommended unit. According to the conversion formula provided by the reagent manufacturer, such as C Tstd =a*Cr original +b, where a and b are the conversion coefficients provided by the manufacturer.

[0032] Imaging data: Use the medical imaging device interface to store the acquired imaging data in DICOM format. For the image resolution, ensure that it meets the research set standards during acquisition, such as Res = X*Y*Z, where X and Y are the planar resolutions and Z is the slice thickness; Label the images. For example, the lesion volume Vlesion can be calculated through an image segmentation algorithm, and the formula can be used

[0033] Treatment response data: Record the treatment drug dose D1 and the medication time t1, as well as the changes in clinical indicators after treatment. For example, the difference in body temperature △T = T2 - T3 between the body temperature T2 after treatment and the body temperature T3 before treatment.

[0034] The construction module of the multi-modal deep algorithm model includes:

[0035] Multi-modal fusion model, which normalizes, enhances data, extracts features and reduces dimensions for imaging data; preprocesses, performs deep semantic encoding and unifies dimensions for text data; standardizes and extracts features for structured data; uses the GBM model to perform regression and classification predictions on multi-time point data; uses AFD to extract features from time series data and input them into the LSTM model; uses the large language model to extract features and associate text information and laboratory results; fuses multi-modal features through the self-attention mechanism;

[0036] The multi-task output layer constructs a multi-task learning framework to parallelly execute the risk prediction of infectious pulmonary diseases and the decision support task based on Task 1. The risk prediction task determines the high-risk state of severe pneumonia in patients and evaluates the necessity of mechanical ventilation. The decision support task combines a large language model and an expert knowledge graph to assist doctors in making decisions on ICU requirements. In the multi-task output layer, the risk prediction task uses the fused multi-modal features combined with imaging lesions, clinical symptoms, and laboratory data, and outputs the high-risk state of severe pneumonia in patients and the evaluation results of the necessity of mechanical ventilation through a classifier. The decision support task semantically analyzes the risk prediction results through a large language model and provides decision-making suggestions in combination with the knowledge graph.

[0037] In the multi-modal fusion model, for the extraction of imaging data features, a pre-trained EfficientNet model is combined with principal component analysis. For the extraction of text data features, a BioBERT model is combined with average pooling operation. For the extraction of structured data features, a random forest model and a support vector machine method are used.

[0038] Specific implementation manners of the technical solutions of the multi-modal fusion model and the multi-task output layer:

[0039] Processing of imaging data:

[0040] Normalization: For the imaging data X irng , the min-max normalization formula can be used to normalize the data to the range (a, b);

[0041]

[0042] where min(X img ) and max(X img ) are the minimum and maximum values of the imaging data, and a and b are the lower and upper limits of the normalization range.

[0043] Data augmentation: Methods such as geometric transformation (such as rotation, translation, scaling) and intensity transformation (such as contrast adjustment, noise addition) can be used to expand the data. For example, for the image rotation operation, assuming the rotation angle is Q, the rotated image X img -rot can be transformed from the original image through the corresponding rotation matrix R(Q): X img -rot = R(Q)X img .

[0044] Feature extraction and dimensionality reduction: Use a pre-trained EfficientNet model to extract features. Assuming the features output by EfficientNet are F EfficientNet , it can be expressed as:

[0045] F EfficientNet = EfficientNet(Ximg -norn-aug), where X img -norn-aug is the imaging data after normalization and data augmentation;

[0046] Then, principal component analysis (PCA) is used for dimensionality reduction. Let the original feature dimension be n, the dimension to be reduced to be k, and the PCA transformation matrix be W PCA , and the feature F after dimensionality reduction img -pca can be expressed as:

[0047] F img -pca = F EfficientNet W PCA ;

[0048] Text data processing: Operations such as cleaning (removing noise and stop words), word segmentation, stemming, or lemmatization are performed on the text data.

[0049] Deep semantic encoding: The BioBERT model is used for semantic encoding, and the processed text data is input into the BioBERT model to obtain the encoded feature F BioBERT :

[0050] F BioBERT = BioBERT(X text-preprocessed ), where X text-preprocessed is the preprocessed text data.

[0051] Multi-modal feature fusion is to fuse features of different modalities using the self-attention mechanism.

[0052] Multi-task output layer: Risk prediction task: Using the fused multi-modal feature F fusion , combined with the imaging lesion feature F lesion , the clinical symptom feature F symptom and the laboratory data feature F lab , input into a classifier (such as a logistic regression or neural network classifier) to evaluate the high-risk state of severe pneumonia and the necessity of mechanical ventilation.

[0053] For the logistic regression classifier, assuming the input feature is Xrisk = (F fusion , F lesion , F symptom , F lab ), the probability prediction formula is:

[0054]

[0055] where y = 1 indicates being in a high-risk state or requiring mechanical ventilation, w is the weight vector, and b is the bias term.

[0056] Decision support task: Semantically parse the risk prediction results, input the risk prediction result Rrisk into a large language model (such as the GPT series), and obtain the semantic interpretation Srisk:

[0057] Srisk = GPT(Rrisk);

[0058] Combine expert knowledge graphs to provide decision-making suggestions, which can be expressed as:

[0059] Ddecision = Combine(Srisk, K).

[0060] Clinical performance verification module, including:

[0061] Clinical enrollment: Relying on partner hospitals, collect infectious pulmonary disease patients and non-infectious controls according to specific time ranges and inclusion and exclusion criteria, longitudinally collect "whole-course, multi-timepoint, multi-site" body fluid samples and conduct standardized management; in the clinical performance verification module, the inclusion criteria for infectious pulmonary disease patients in clinical enrollment include meeting the diagnostic criteria for community-acquired pneumonia, signing an informed consent form, and being less than 80 years old; the inclusion criteria for patients without lung infections include being less than 80 years old, having no serious underlying diseases, and not participating in relevant clinical trials.

[0062] Prospective cohort data integration and management: Rely on the hospital's advantages to establish a unified data processing process to ensure data quality, remove redundant information, protect privacy, and standardize data processing.

[0063] Clinical performance verification indicators: Use accuracy, precision, recall, and F1-score indicators to evaluate the effectiveness and reliability of the model in the diagnosis of infectious pulmonary diseases.

[0064] It can be understood that the present invention is described through some embodiments. Those skilled in the art know that without departing from the spirit and scope of the present invention, various changes or equivalent substitutions can be made to these features and embodiments. Additionally, under the teaching of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.

Claims

1. A system for constructing a standardized integration and comprehensive information model of clinical data, characterized in that: It includes an infectious pulmonary disease clinical cohort data collection system, which is used to establish a unified data collection system and standards based on the prospective infectious pulmonary disease specialized disease research cohort of partner units, collect and integrate multi-dimensional data such as clinical indicators, laboratory test data, imaging data, and treatment responses, and collect, process, and store the data, including converting laboratory test data into international standard units, storing imaging data in DICOM format, cleaning and standardizing the data, storing it using a hybrid-mode architecture database, and setting user permission management and data encryption; A deep algorithm model construction module based on multi-modalities, including: A multi-modal fusion model that normalizes, data-augments, extracts features, and reduces the dimension of imaging data; preprocesses, deeply semantically encodes, and unifies the dimension of text data; standardizes and extracts features from structured data; uses a GBM model to perform regression and classification predictions on multi-time point data; uses AFD to extract features from time series data and inputs them into an LSTM model; uses a large language model to extract features and make associations between text information and laboratory results; and fuses multi-modal features through a self-attention mechanism; A multi-task output layer that constructs a multi-task learning framework to concurrently execute the risk prediction of infectious pulmonary diseases and the decision support task based on Task 1. The risk prediction task determines the high-risk status of severe pneumonia in patients and evaluates the necessity of mechanical ventilation, and the decision support task combines a large language model and an expert knowledge graph to assist doctors in making decisions on ICU requirements A clinical performance verification module, including: Clinical enrollment, relying on partner hospitals, collecting infectious pulmonary disease patients and non-infectious controls according to specific time ranges and inclusion and exclusion criteria, longitudinally collecting "whole-course, multi-time point, multi-site" body fluid samples and conducting standardized management; Prospective cohort data integration and management, relying on the hospital's advantages to establish a unified data processing process to ensure data quality, remove redundant information, protect privacy, and standardize the data. Clinical performance verification indicators, using accuracy, precision, recall, and F1-score indicators to evaluate the effectiveness and reliability of the model in the diagnosis of infectious pulmonary diseases.

2. The standardized fusion of clinical data and the system for constructing an integrated information model according to claim 1, characterized in that: In the infectious pulmonary disease clinical cohort data collection system, the data collection scope covers the patient's baseline information, intervention measures, and clinical outcomes, and additional metadata such as data source, collection time, and disease type are added.

3. The standardized fusion of clinical data and the system for constructing an integrated information model according to claim 1, characterized in that: The specific embodiments of the infectious pulmonary disease clinical cohort data collection system are as follows: Clinical indicators: Obtain the patient's clinical information from the electronic medical record system of the partner unit, including demographic information and symptom information. Symptoms can be collected through questionnaires or doctor records. For symptom scoring, relevant scales can be used. For example, for the cough severity score S, different score ranges can be set according to different symptom degrees; Laboratory test data: Collect the results of various laboratory tests, such as the white blood cell count WBG. If the unit is 10 9 / L, but some units are 10 3 / uL. When converting to the international standard unit, use the formula: When the original unit is 10 3 / uL, WBG std = WBG original / 1000; For biochemical indicators, such as serum creatinine C T , the results of different detection methods are uniformly converted into internationally recommended units. The conversion formula provided by the reagent manufacturer can be used, such as C Tstd = a * Cr original + b, where a and b are the conversion coefficients provided by the manufacturer; Imaging data: Using the interface of medical imaging equipment, the acquired imaging data is stored in DICOM format. For the imaging resolution, ensure that it meets the research set standards during acquisition, such as Res = X * Y * Z, where X and Y are the planar resolutions and Z is the slice thickness; label the images, such as the lesion volume V lesion It can be calculated through image segmentation algorithms and the formula can be used Treatment response data: Record the treatment drug dose D1 and medication time t1, as well as the changes in clinical indicators after treatment, such as the difference △T = T2 - T3 between the body temperature T2 after treatment and the body temperature T3 before treatment.

4. A system for standardizing the integration of clinical data and constructing a comprehensive information model according to claim 1, characterized in that: In the multi-modal fusion model, for the extraction of imaging data features, a pre-trained EfficientNet model is combined with principal component analysis; for the extraction of text data features, a BioBERT model is combined with average pooling operation; and for the extraction of structured data features, a random forest model and a support vector machine method are used.

5. A system for standardizing the integration of clinical data and constructing a comprehensive information model according to claim 1, characterized in that: Specific implementation manners of the multi-modal fusion model and the multi-task output layer technical solution: Imaging data processing: Normalization: For imaging data X irng , the data can be normalized to the range (a, b) using the min-max normalization formula; where min(X img ) and max(X img ) are the minimum and maximum values of the imaging data, and a and b are the lower and upper limits of the normalization range; Data augmentation: Methods such as geometric transformations (e.g., rotation, translation, scaling) and intensity transformations (e.g., contrast adjustment, noise addition) can be used to augment data. For example, for the image rotation operation, let the rotation angle be Q, and the rotated image X img -rot can be transformed from the original image through the corresponding rotation matrix R(Q): X img -rot = R(Q)X img .

6. The standardized fusion of clinical data and the construction system of the comprehensive information model according to claim 5, characterized in that: Feature extraction and dimensionality reduction: Use a pre-trained EfficientNet model to extract features. Assume that the output features of EfficientNet are F EfficientNet , which can be expressed as: F EfficientNet = EfficientNet(X img -norn-aug), where X img -norn-aug is the imaging data after normalization and data augmentation; Then, principal component analysis (PCA) is used for dimensionality reduction. Let the original feature dimension be n, the dimension to be reduced to be k, and the PCA transformation matrix be W PCA , and the feature F after dimensionality reduction img -pca can be expressed as: F img -pca = F EfficientNet W PCA ; Text data processing: cleaning, tokenization, stemming or lemmatization operations are performed on the text data; Deep semantic encoding: Use the BioBERT model for semantic encoding, and input the processed text data into the BioBERT model to obtain the encoded feature F BioBERT : F BioBERT = BioBERT(X text-preprocessed ), X text-preprocessed is the preprocessed text data.

7. A system for standardizing the integration of clinical data and constructing a comprehensive information model according to claim 1, characterized in that: In the multi-task output layer, for the risk prediction task, the fused multi-modal features are combined with imaging lesions, clinical symptoms and laboratory data, and the classifier outputs the assessment results of the high-risk state of severe pneumonia and the necessity of mechanical ventilation for the patient; for the decision support task, a large language model semantically analyzes the risk prediction results and provides decision-making suggestions in combination with the knowledge graph.

8. A system for standardizing the integration of clinical data and constructing a comprehensive information model according to claim 1, characterized in that: In the clinical performance verification module, the inclusion criteria for patients with infectious pulmonary diseases enrolled clinically include meeting the diagnostic criteria for community-acquired pneumonia, signing the informed consent form and being less than 80 years old; the inclusion criteria for patients without pulmonary infections include being less than 80 years old, having no serious underlying diseases and not participating in relevant clinical trials.

Images