Self-monitoring system for stress-related obstacle typing

Abstract

The invention relates to a self-monitoring system for stress-related disorder typing, and belongs to the technical field of medical data analysis. The system comprises a data collection module which collects medical data of a patient, wherein the medical data comprises fMRI data, questionnaires and scales. The data is divided into a stress disorder group and a non-stress disorder group. And the preprocessing module is used for preprocessing various data and improving the data quality. And the model training module is used for carrying out deep learning training on the data and generating a stress-related obstacle typing model. And the evaluation module is used for carrying out effect evaluation on the generated stress-related obstacle typing model so as to ensure the accuracy and the reliability of the model. The evaluation result is used for selecting or optimizing an optimal model. And the integration module is used for integrating the evaluated and optimized optimal models into a complete system, so that the system can comprehensively analyze fMRI data and questionnaire scale results of the patients and can divide the patients into different stress-related disorder types.

Description

Technical Field

[0001] The invention relates to a self-monitoring system for stress-related disorder classification, belonging to the technical field of medical data analysis. Background Art

[0002] Stress-related disorders include acute stress reaction, post-traumatic stress disorder, and adjustment disorder. However, doctors can usually determine whether a patient has a stress-related disorder, but it may be difficult to identify the specific type of disorder. Therefore, it is crucial to develop a classification system.

[0003] This classification system will extract key features from a large amount of clinical data and use deep learning and medical data analysis to quantify the patient's symptoms and performance. Through in-depth analysis and pattern recognition of these data, the system can classify patients into different types of stress-related disorders. This classification system helps doctors better develop personalized help plans, thereby providing more comprehensive support for patients. Summary of the invention

[0004] The technical problem to be solved by the present invention is: the present invention provides a self-monitoring system for stress-related disorder classification, which is used to extract key features from a large amount of clinical data, and use deep learning and medical data analysis methods to quantitatively analyze the patient's symptoms and manifestations. Through in-depth analysis and pattern recognition of these data, the system can classify patients into different types of stress-related disorders.

[0005] The technical solution of the present invention is: a self-monitoring system for stress-related disorder classification, the system comprising:

[0006] The data collection module, preprocessing module, model training module, model evaluation module and system integration module;

[0007] The data collection module includes a data screening unit and a data grouping unit;

[0008] The preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit;

[0009] The fMRI data preprocessing unit includes basic preprocessing, time series extraction and functional connection matrix construction;

[0010] The questionnaire data preprocessing unit includes data cleaning, coding conversion, and normalization processing;

[0011] The model training module includes an fMRI feature extraction unit, a questionnaire scale feature extraction unit and a feature fusion unit;

[0012] The fMRI feature extraction unit includes a graph convolutional neural network (GCN) algorithm to perform deep feature extraction on the functional connection matrix and output an fMRI feature vector;

[0013] The questionnaire scale feature extraction unit includes natural language processing technology and GloVe model to process questionnaire and scale data and output questionnaire scale feature vector;

[0014] The feature fusion unit inputs the fMRI feature vector and the questionnaire scale feature vector, and outputs the subtype classification result by applying a multi-layer perceptron MLP;

[0015] The model evaluation module includes a testing unit, a result acquisition unit and an evaluation unit;

[0016] The system integration module includes a system integration unit, an operation interface unit, a result presentation unit, a maintenance and update unit, and a performance monitoring and log recording unit.

[0017] Furthermore, the data collection module is used to obtain fMRI data and questionnaire scale data of stress-related disorders from the stress-related disorder clinical database, screen out data that meet quality requirements, and divide the two types of data into a post-traumatic stress disorder group, an acute stress reaction group, and an adjustment disorder group according to data attributes.

[0018] Furthermore, the preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit:

[0019] an fMRI data preprocessing unit, for performing a series of preprocessing steps on the collected fMRI data, the preprocessing including removing noise in the time series, spatial standardization, normalization of the time series, and any necessary data conversion, after the preprocessing, dividing the fMRI data according to predefined brain regions, and extracting the time series data of each region, and calculating the functional connectivity between brain regions using the extracted time series data;

[0020] The questionnaire and scale data preprocessing unit is used to preprocess the questionnaire and scale data, including data cleaning, encoding conversion, and normalization, to provide standardized input for model training.

[0021] Furthermore, the model training module includes:

[0022] An fMRI feature extraction unit is used to perform deep feature extraction on the preprocessed functional connection matrix using a graph convolutional neural network (GCN), identify brain activity patterns specific to various stress-related disorder subtypes, and output fMRI feature vectors;

[0023] A questionnaire scale feature extraction unit, used to process questionnaire and scale data using natural language processing technology and GloVe model to capture psychological and behavioral characteristics related to stress-related disorder subtypes and output questionnaire scale feature vectors;

[0024] The feature fusion unit is a strategy for integrating feature vectors from different data sources. By applying a multi-layer perceptron (MLP), various features and their interactions are considered to improve the accuracy of identifying subtypes of stress-related disorders.

[0025] Furthermore, the model evaluation module includes:

[0026] The test unit is used to run the model trained by the model training module on an independent test data set to simulate the model performance in actual application scenarios and ensure the accuracy and practicality of the evaluation;

[0027] The result acquisition unit is used to collect the running results of the test unit, including the prediction accuracy, recall rate, and F1 score key performance indicators of the model, so as to fully understand the model performance;

[0028] The evaluation unit is used to comprehensively analyze the performance data collected by the result acquisition unit, compare different models or different configurations of the same model, and determine the optimal model configuration and parameters; in addition, this unit is also responsible for identifying potential problems in model performance, such as overfitting or underfitting, and making corresponding optimization suggestions.

[0029] Furthermore, the system integration module includes:

[0030] System integration unit, which is used to integrate various modules into a unified operating system to ensure the automation and seamless operation of the process from data input to diagnosis output;

[0031] Operation interface unit, used to provide an intuitive and friendly operation interface, including data upload, result query and system setting functions;

[0032] A result presentation unit that presents diagnostic results in an easy-to-understand format, ensuring that users can quickly access key information, including disease classification, severity, and possible treatment recommendations;

[0033] Maintenance and update unit, used for regular inspection and update of the system, including model update, system function enhancement and user feedback processing;

[0034] The performance monitoring and logging unit is used to monitor and record the system's operating status and user activities to support system security, stable operation and subsequent audit and improvement activities.

[0035] Furthermore, in the fMRI data preprocessing unit, the Pearson correlation coefficient PCC is calculated by the following formula to construct a functional connectivity matrix, where r nm represents the Pearson correlation coefficient between the signal time series of the nth brain region and the mth brain region;

[0036]

[0037] in, and are the mean signals of brain regions n and m, respectively. N represents the total number of time points, and T tn represents the signal value of brain area n at time point t, T tm represents the signal value of brain area m at time point t.

[0038] Furthermore, the model evaluation module calculates the accuracy, recall R and F1score indicators to optimize the model parameters, where TP is the number of samples that are actually positive and predicted as positive by the model, FN is the number of samples that are actually positive but predicted as negative by the model, FP is the number of samples that are actually negative but predicted as positive by the model, and TN is the number of samples that are actually negative and predicted as negative by the model.

[0039]

[0040] The beneficial effects of the present invention are as follows: the present invention first collects the medical data of the patient, including fMRI data, questionnaires and scales, and these data are divided into a stress disorder group and a non-stress disorder group; pre-processes various types of data through a pre-processing module to improve data quality; performs deep learning training on the data through a model training module to generate a stress-related disorder classification model; performs effect evaluation on the stress-related disorder classification model generated above through an evaluation module to ensure the accuracy and reliability of the model, and the evaluation result will be used to select or optimize the optimal model; and integrates the evaluated and optimized optimal model into a complete system through an integration module, so that it can comprehensively analyze the patient's fMRI data and questionnaire scale results, and can classify the patient into different types of stress-related disorders. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the present invention;

[0042] Figure 2 Obtaining process maps for fMRI time series;

[0043] Figure 3 It is the sparse process map of fMRI functional connectivity matrix;

[0044] Figure 4Figure 2 is a feature fusion model diagram for feature fusion. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0046] Example 1: Figure 1-Figure 4 As shown, a self-monitoring system for classification of stress-related disorders, the system comprising:

[0047] The data collection module, preprocessing module, model training module, model evaluation module and system integration module;

[0048] The data collection module includes a data screening unit and a data grouping unit;

[0049] The preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit;

[0050] The fMRI data preprocessing unit includes basic preprocessing, time series extraction and functional connection matrix construction;

[0051] The questionnaire data preprocessing unit includes data cleaning, coding conversion, and normalization processing;

[0052] The model training module includes an fMRI feature extraction unit, a questionnaire scale feature extraction unit and a feature fusion unit;

[0053] The fMRI feature extraction unit includes a graph convolutional neural network (GCN) algorithm to perform deep feature extraction on the functional connection matrix and output an fMRI feature vector;

[0054] The questionnaire scale feature extraction unit includes natural language processing technology and GloVe model to process questionnaire and scale data and output questionnaire scale feature vector;

[0055] The feature fusion unit inputs the fMRI feature vector and the questionnaire scale feature vector, and outputs the subtype classification result by applying a multi-layer perceptron MLP;

[0056] The model evaluation module includes a testing unit, a result acquisition unit and an evaluation unit;

[0057] The system integration module includes a system integration unit, an operation interface unit, a result presentation unit, a maintenance and update unit, and a performance monitoring and log recording unit.

[0058] Furthermore, the data collection module is used to obtain fMRI data and questionnaire scale data of stress-related disorders from the stress-related disorder clinical database, screen out data that meet quality requirements, and divide the two types of data into a post-traumatic stress disorder group, an acute stress reaction group, and an adjustment disorder group according to data attributes.

[0059] Furthermore, the preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit:

[0060] an fMRI data preprocessing unit, for performing a series of preprocessing steps on the collected fMRI data, the preprocessing including removing noise in the time series, spatial standardization, normalization of the time series, and any necessary data conversion, after the preprocessing, dividing the fMRI data according to predefined brain regions, and extracting the time series data of each region, and calculating the functional connectivity between brain regions using the extracted time series data;

[0061] The questionnaire and scale data preprocessing unit is used to preprocess the questionnaire and scale data, including data cleaning, encoding conversion, and normalization, to provide standardized input for model training.

[0062] Furthermore, the model training module includes:

[0063] An fMRI feature extraction unit is used to perform deep feature extraction on the preprocessed functional connection matrix using a graph convolutional neural network (GCN), identify brain activity patterns specific to various stress-related disorder subtypes, and output fMRI feature vectors;

[0064] A questionnaire scale feature extraction unit, used to process questionnaire and scale data using natural language processing technology and GloVe model to capture psychological and behavioral characteristics related to stress-related disorder subtypes and output questionnaire scale feature vectors;

[0065] The feature fusion unit is a strategy for integrating feature vectors from different data sources. By applying a multi-layer perceptron (MLP), various features and their interactions are considered to improve the accuracy of identifying subtypes of stress-related disorders.

[0066] Furthermore, the model evaluation module includes:

[0067] The test unit is used to run the model trained by the model training module on an independent test data set to simulate the model performance in actual application scenarios and ensure the accuracy and practicality of the evaluation;

[0068] The result acquisition unit is used to collect the running results of the test unit, including the prediction accuracy, recall rate, and F1 score key performance indicators of the model, so as to fully understand the model performance;

[0069] The evaluation unit is used to comprehensively analyze the performance data collected by the result acquisition unit, compare different models or different configurations of the same model, and determine the optimal model configuration and parameters; in addition, this unit is also responsible for identifying potential problems in model performance, such as overfitting or underfitting, and making corresponding optimization suggestions.

[0070] Furthermore, the system integration module includes:

[0071] System integration unit, which is used to integrate various modules into a unified operating system to ensure the automation and seamless operation of the process from data input to diagnosis output;

[0072] Operation interface unit, used to provide an intuitive and friendly operation interface, including data upload, result query and system setting functions;

[0073] A result presentation unit that presents diagnostic results in an easy-to-understand format, ensuring that users can quickly access key information, including disease classification, severity, and possible treatment recommendations;

[0074] Maintenance and update unit, used for regular inspection and update of the system, including model update, system function enhancement and user feedback processing;

[0075] The performance monitoring and logging unit is used to monitor and record the system's operating status and user activities to support system security, stable operation and subsequent audit and improvement activities.

[0076] Furthermore, in the fMRI data preprocessing unit, the Pearson correlation coefficient PCC is calculated by the following formula to construct a functional connectivity matrix, where r nm represents the Pearson correlation coefficient between the signal time series of the nth brain region and the mth brain region;

[0077]

[0078] in, and are the mean signals of brain regions n and m, respectively. N represents the total number of time points, and T tn represents the signal value of brain area n at time point t, T tm represents the signal value of brain area m at time point t.

[0079] Furthermore, the model evaluation module calculates the accuracy, recall R and F1score indicators to optimize the model parameters, where TP is the number of samples that are actually positive and predicted as positive by the model, FN is the number of samples that are actually positive but predicted as negative by the model, FP is the number of samples that are actually negative but predicted as positive by the model, and TN is the number of samples that are actually negative and predicted as negative by the model.

[0080]

[0081] The modules included in the self-monitoring system for the classification of stress-related disorders of the present invention realize the complete process from data collection to disease classification. The process of a self-checking method for the classification of stress-related disorders is as follows:

[0082] Step 1: Data Collection,The system first obtains fMRI data and questionnaire scale data of patients with stress-related disorders from a clinical database through the data collection module.,The data screening unit ensures that the selected data meets the quality standards,,while the data grouping unit divides the data into post-traumatic stress disorder,,acute stress reaction, and adjustment disorder groups according to their different attributes.

[0083] Step 2: Data preprocessing,In the preprocessing module, the fMRI data undergoes basic processing steps such as denoising, spatial normalization, and time series normalization. After time series extraction, the functional connectivity matrix between brain regions is generated using the functional connectivity matrix construction unit, and the sparse connectivity matrix is ​​obtained using the graph autoencoder. At the same time, the questionnaire scale data is prepared for subsequent processing through steps such as cleaning, encoding conversion, and normalization.

[0084] Step 3: Model training. In the model training module, the sparse connection matrix is ​​extracted through deep features of the graph convolutional neural network (GCN). The questionnaire data is processed through natural language processing technology and the GloVe model. Subsequently, the feature fusion unit integrates the feature vectors of the two data sources and uses the multi-layer perceptron (MLP) to output the final subtype classification results.

[0085] Step 4: Model evaluation, the model evaluation module optimizes the parameters by evaluating the model on an independent test set and calculates key indicators such as the model's accuracy, recall, and F1 score to ensure the effectiveness and reliability of the model.

[0086] Step 5: System Integration. The system integration module ensures that all sub-modules work together, provides a user-friendly operation interface, displays easy-to-understand diagnostic results, and implements system maintenance updates and performance monitoring.

[0087] Through the above steps, the system in this embodiment can effectively identify and classify different subtypes of stress-related disorders.

[0088] The specific embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A self-monitoring system for the classification of stress-related disorders, characterized in that: The system comprises: The data collection module, preprocessing module, model training module, model evaluation module and system integration module; The data collection module includes a data screening unit and a data grouping unit; The preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit; The fMRI data preprocessing unit includes basic preprocessing, time series extraction and functional connection matrix construction; The questionnaire data preprocessing unit includes data cleaning, coding conversion, and normalization processing; The model training module includes an fMRI feature extraction unit, a questionnaire scale feature extraction unit and a feature fusion unit; The fMRI feature extraction unit includes a graph convolutional neural network (GCN) algorithm to perform deep feature extraction on the functional connection matrix and output an fMRI feature vector; The questionnaire scale feature extraction unit includes natural language processing technology and GloVe model to process questionnaire and scale data and output questionnaire scale feature vector; The feature fusion unit inputs the fMRI feature vector and the questionnaire scale feature vector, and outputs the subtype classification result by applying a multi-layer perceptron MLP; The model evaluation module includes a testing unit, a result acquisition unit and an evaluation unit; The system integration module includes a system integration unit, an operation interface unit, a result presentation unit, a maintenance and update unit, and a performance monitoring and log recording unit.

2. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The data collection module is used to obtain fMRI data and questionnaire scale data of stress-related disorders from the stress-related disorder clinical database, screen out data that meet quality requirements, and divide the two types of data into a post-traumatic stress disorder group, an acute stress reaction group, and an adjustment disorder group according to data attributes.

3. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The preprocessing module includes an fMRI data preprocessing unit and a questionnaire scale data preprocessing unit: an fMRI data preprocessing unit, for performing a series of preprocessing steps on the collected fMRI data, the preprocessing including removing noise in the time series, spatial standardization, normalization of the time series, and any necessary data conversion, after the preprocessing, dividing the fMRI data according to predefined brain regions, and extracting the time series data of each region, and calculating the functional connectivity between brain regions using the extracted time series data; The questionnaire and scale data preprocessing unit is used to preprocess the questionnaire and scale data, including data cleaning, encoding conversion, and normalization, to provide standardized input for model training.

4. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The model training module includes: An fMRI feature extraction unit is used to perform deep feature extraction on the preprocessed functional connection matrix using a graph convolutional neural network (GCN), identify brain activity patterns specific to various stress-related disorder subtypes, and output fMRI feature vectors; A questionnaire scale feature extraction unit, used to process questionnaire and scale data using natural language processing technology and GloVe model to capture psychological and behavioral characteristics related to stress-related disorder subtypes and output questionnaire scale feature vectors; The feature fusion unit is a strategy for integrating feature vectors from different data sources. By applying a multi-layer perceptron (MLP), various features and their interactions are considered to improve the accuracy of identifying subtypes of stress-related disorders.

5. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The model evaluation module comprises: The test unit is used to run the model trained by the model training module on an independent test data set to simulate the model performance in actual application scenarios and ensure the accuracy and practicality of the evaluation; The result acquisition unit is used to collect the running results of the test unit, including the prediction accuracy, recall rate, and F1 score key performance indicators of the model, so as to fully understand the model performance; The evaluation unit is used to comprehensively analyze the performance data collected by the result acquisition unit, compare different models or different configurations of the same model, and determine the optimal model configuration and parameters; in addition, this unit is also responsible for identifying potential problems in model performance, such as overfitting or underfitting, and making corresponding optimization suggestions.

6. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The system integration module comprises: System integration unit, which is used to integrate various modules into a unified operating system to ensure the automation and seamless operation of the process from data input to diagnosis output; Operation interface unit, used to provide an intuitive and friendly operation interface, including data upload, result query and system setting functions; A result presentation unit that presents diagnostic results in an easy-to-understand format, ensuring that users can quickly access key information, including disease classification, severity, and possible treatment recommendations; Maintenance and update unit, used for regular inspection and update of the system, including model update, system function enhancement and user feedback processing; The performance monitoring and logging unit is used to monitor and record the system's operating status and user activities to support system security, stable operation and subsequent audit and improvement activities.

7. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: In the fMRI data preprocessing unit, the Pearson correlation coefficient PCC is calculated by the following formula to construct the functional connectivity matrix. nm represents the Pearson correlation coefficient between the signal time series of the nth brain region and the mth brain region; in, and are the mean signals of brain regions n and m, respectively. N represents the total number of time points, and T tn represents the signal value of brain area n at time point t, T tm represents the signal value of brain area m at time point t.

8. The self-monitoring system for classification of stress-related disorders according to claim 1, characterized in that: The model evaluation module calculates the accuracy, recall R and F1score indicators to optimize the model parameters, where TP true positive examples: the number of samples that are actually positive examples and predicted as positive examples by the model, FN false negative examples: the number of samples that are actually positive examples but predicted as negative examples by the model, FP false positive examples: the number of samples that are actually negative examples but predicted as positive examples by the model, TN true negative examples: the number of samples that are actually negative examples and predicted as negative examples by the model;

Images