Community elderly cognitive decline prediction and intervention system based on machine learning

Abstract

The invention relates to the technical field of medical big data information processing, and particularly discloses a community elderly cognitive decline prediction and intervention system based on machine learning. The system comprises a data acquisition module, a data processing module, a data analysis module and an intervention scheme construction and follow-up feedback module, according to the method, multi-modal data of a patient is acquired and processed by utilizing a big data technology, two cognitive weakness prediction models are constructed by utilizing a machine learning model in combination with different characteristics of the multi-modal data, and prediction of the cognitive weakness degree of the patient is realized by fusing results of the two models; corresponding standard intervention schemes are given according to different prediction results, medical staff can adjust the standard intervention schemes according to the bad health event atlas of the patient to form personalized intervention schemes corresponding to the patient, and follow-up feedback is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical big data information processing, and specifically relates to a community elderly cognitive decline and intervention system based on machine learning. Background Art

[0002] In the current social environment, the elderly usually suffer from two or more chronic diseases. In this case, with the decline of physical function, malnutrition and cognitive impairment, the risk of dementia in the elderly increases. Cognitive decline is a state of impaired cognitive function, accompanied by physical weakness and cognitive dysfunction. It is regarded as the early stage of dementia and a key intervention point for dementia prevention. It has a certain degree of reversibility. Timely identification and follow-up of elderly people with cognitive decline is not only in line with the country's policy of promoting health protection throughout the life cycle, but also an important part of the "preventive treatment" strategy.

[0003] Traditional methods of identifying cognitive decline and follow-up are often limited by low efficiency and data quality issues. The elderly with cognitive decline are mainly distributed in the community, and different elderly people have their own personalized needs, which cannot be generalized. In addition, with the development of big data and artificial intelligence technologies, as well as the gradual popularization of smart medicine, medical data analysis and identification based on big data and artificial intelligence have become possible; sample data from more sources are collected and processed, target prediction models are established, and personalized intervention plans are set according to different prediction situations. Therefore, the present invention aims to provide a community elderly cognitive decline prediction and intervention system based on machine learning. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned problems existing in the prior art, and to provide a community elderly cognitive decline prediction and intervention system based on machine learning, which comprehensively realizes the multi-source sample data collection of the community elderly population, the classification processing and feature extraction of the multi-source sample data, the cognitive decline prediction model constructed based on the multi-source sample data features, the cognitive decline prediction results of the current sample data based on the cognitive decline prediction model, and the personalized intervention plan formulated based on the cognitive decline prediction results. The system uses big data and artificial intelligence technology to make a more efficient prediction of cognitive decline in the elderly in the community, and the prediction results provide auxiliary results for medical personnel to judge the specific condition; at the same time, according to the different categories of the prediction results, a targeted intervention plan is provided, which can improve the effect and accuracy of the intervention and improve the efficiency of medical work.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions: A community elderly cognitive decline prediction and intervention system based on machine learning, including the following functional modules: Data collection module: including basic data collection unit, neuroimaging data collection unit, cognitive function data collection unit, motor ability data collection unit, psychological state data collection unit and adverse health event collection unit; Basic data collection unit, used to collect basic patient data, including age, height, weight, sleep data, personal chronic disease history and family chronic disease history; A neuroimaging data acquisition unit, used to acquire neuroimaging data of patients, including brain structure magnetic resonance imaging data sMRI; Cognitive function data collection unit, used to collect patients' cognitive function data, including cognitive assessment data of memory, attention, executive function, language ability and visual-spatial ability; The sports ability data acquisition unit is used to collect sports ability data of the patient, including sports video data, climbing test data, pace data and grip strength data, wherein the sports video data includes sit-up test video and balance test video; the sports video data and climbing test data are used to determine physical fitness data; A psychological state data collection unit is used to collect the patient's psychological state data, including depression assessment data, anxiety assessment data and quality of life assessment data; Adverse health event collection unit, used to collect patients' adverse health event data, including falls, delirium and hospitalization events; Data processing module: including basic data processing unit, neuroimaging data processing unit, cognitive function data processing unit, motor ability data processing unit, psychological state data processing unit and adverse health event processing unit; A basic data processing unit, used for processing the basic data to obtain a basic data feature vector; A neuroimaging data processing unit, used to process the neuroimaging data to obtain brain network feature data; A cognitive function data processing unit, used to process the cognitive function data to obtain a cognitive function data feature vector; A sports ability data processing unit, used for processing the sports ability data to obtain a sports ability data feature vector; A psychological state data processing unit, used for processing the psychological state data to obtain a psychological state data feature vector; An adverse health event processing unit, used to process the adverse health event to obtain a characteristic graph of the adverse health event; Data analysis module: including cognitive decline prediction model building unit and cognitive decline prediction unit; A cognitive decline prediction model construction unit, used to establish cognitive decline prediction models A and B, wherein the prediction model A is constructed by graph convolutional neural network training; and the prediction model B is constructed by multi-layer perceptron model training; A cognitive decline prediction unit: inputting the brain network feature data into a cognitive decline prediction model A to obtain a cognitive decline prediction result A; performing feature selection processing and splicing operations on the basic data feature vector, cognitive function data feature vector, motor ability data feature vector, and psychological state data feature vector to obtain a patient comprehensive data feature vector, inputting the patient comprehensive data feature vector into a cognitive decline prediction model B to obtain a cognitive decline prediction result B; and using the cognitive decline prediction results A and B to obtain a final cognitive decline prediction result; Intervention plan construction and follow-up feedback module: including standard intervention plan generation unit, personalized intervention plan generation unit and follow-up feedback unit; A standard intervention program generating unit, used for determining a standard intervention program corresponding to the patient according to the final cognitive decline prediction result; A personalized intervention plan generating unit is used to adjust the standard intervention plan corresponding to the patient according to the patient's adverse health event characteristic map to obtain a personalized intervention plan for the patient; The follow-up feedback unit is used to follow up and intervene in patients according to their personalized intervention plans, and collect patients' basic data, neuroimaging data, cognitive function data, motor ability data, psychological state data and adverse health events according to preset cycles; and compare the changes in the degree of cognitive decline of patients after intervention.

[0006] Furthermore, the basic data processing unit specifically includes: constructing a basic data-score value mapping table, including score values ​​corresponding to different numerical intervals of age, height, weight, sleep data, personal chronic disease history and family chronic disease history, quantizing the basic data into component values ​​according to the basic data-score value mapping table, and then normalizing all score values ​​to the [0,1] interval and splicing them to obtain a basic data feature vector.

[0007] Furthermore, the neuroimaging data processing unit specifically includes: performing T1-weighted imaging processing on the brain structure magnetic resonance imaging data sMRI to obtain a T1-weighted image, using freesurfer software to perform brain region segmentation on the T1-weighted image to obtain a brain region segmentation result; using the brain region included in the brain region segmentation result as an sMRI brain network node, and constructing an sMRI connection matrix based on the connection relationship between brain regions, and the sMRI brain network nodes and the sMRI connection matrix together constitute the brain network feature data.

[0008] Furthermore, the cognitive function data processing unit specifically includes: processing each type of data in the cognitive assessment data separately to obtain a memory assessment score, an attention assessment score, an executive function assessment score, a speech ability assessment score, and a visual-spatial ability assessment score; normalizing and mapping the above scores to the range of [0,1] and splicing them to obtain a cognitive function data feature vector.

[0009] Furthermore, the motor ability data processing unit specifically includes: performing motion posture detection and tracking on the sit-up test video data, calculating the number of times num that the patient completes the prescribed action in a test cycle T and the average time time to complete a prescribed action; performing motion posture detection and tracking on the balance test video, calculating the longest time t-max that the patient can maintain a balanced posture in a test cycle T; processing the climbing test data, extracting the patient's climbing height high and forward distance distance in a test cycle T; constructing a patient fitness data set fitness_data=(num, time, t-max, high, distance), evaluating the patient's physical fitness based on the patient fitness data set fitness_data, and obtaining a physical fitness evaluation result including excellent, good and poor; constructing a motor ability-score value mapping table, including score values ​​corresponding to different numerical intervals of pace and grip strength, and score values ​​corresponding to different evaluation results of physical fitness, quantizing the motor ability data into component values ​​based on the motor ability-score value mapping table, and then normalizing all the score values ​​to the [0,1] interval range and splicing them to obtain a motor ability data feature vector.

[0010] Furthermore, the psychological state data processing unit specifically includes: obtaining the patient's depression level, anxiety level and quality of life level according to the depression assessment data, anxiety assessment data and quality of life assessment data, constructing a psychological state-score value mapping table, including score values ​​corresponding to different levels of depression, anxiety and quality of life, quantizing the psychological state data into component values ​​according to the psychological state-score value mapping table, and then normalizing all the score values ​​to the [0,1] interval and splicing them to obtain a psychological state data feature vector.

[0011] Furthermore, the adverse health event processing unit specifically includes: establishing a visualization map of adverse health events, taking three types of events, namely, falls, delirium, and hospitalization, as the first-level nodes of the map, and setting second-level nodes under the first-level nodes. The second-level nodes are specific events of falls, delirium, and hospitalization each time, and the specific time and cause of the events are marked on the second-level nodes. The co-occurrence probability between the first-level nodes is calculated. If the co-occurrence probability is greater than a preset threshold, the first-level nodes are connected with lines; the co-occurrence probability is calculated as follows: , where p ijrepresents the co-occurrence probability between first-level nodes i and j, num_total(i,j) represents the sum of the number of second-level nodes corresponding to first-level nodes i and j, num_co(i,j) represents the number of co-occurrences between second-level nodes corresponding to first-level nodes i and j. If the difference in the occurrence time of the second-level nodes is within the preset time range, it is considered that the second-level nodes co-occur.

[0012] Furthermore, the cognitive decline prediction unit specifically includes: determining the importance of basic data, cognitive function data, motor ability data and psychological state data according to the expert scoring method, determining the percentage of retained data feature values ​​of the basic data, cognitive function data, motor ability data and psychological state data according to the importance of each of the above data, the higher the importance, the higher the percentage of retained feature values; splicing the feature vectors after feature selection processing to form the patient's comprehensive data feature vector.

[0013] Furthermore, the cognitive decline prediction unit specifically includes: the cognitive decline prediction results A and B are probability value vectors of the patient's cognitive decline levels being high, medium and low, respectively, output by models A and B; weighted summation of the cognitive decline prediction results A and B is performed to obtain a final probability value vector V_final of the patient's cognitive decline levels being high, medium and low, respectively, and the decline level corresponding to the maximum probability value in V_final is used as the patient's final cognitive decline prediction result.

[0014] Furthermore, the personalized intervention plan generation unit specifically includes: showing the patient's adverse health event characteristic map and the patient's corresponding standard intervention plan to medical personnel, and the medical personnel adjust the patient's corresponding standard intervention plan to generate a personalized intervention plan for the patient.

[0015] By adopting the above technology, compared with the prior art, the beneficial effects of the present invention are: This application collects multi-source data and multi-modal data of patients, uses big data technology to implement data processing based on the inherent characteristics of different data, and forms a patient data feature vector.

[0016] This application combines the different characteristics of multimodal data and uses machine learning models to construct two cognitive decline prediction models. The prediction results of the two models are integrated to predict the degree of cognitive decline in patients.

[0017] This application provides corresponding standard intervention plans based on different prediction results. On this basis, medical personnel can adjust the standard intervention plans according to the patient's adverse health event map to form a personalized intervention plan corresponding to the patient, and achieve follow-up feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] 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 therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A functional module diagram of a community elderly cognitive decline prediction and intervention system based on machine learning for this application; Figure 2 This is the functional flow chart of the data collection module of this application; Figure 3 This is the functional flow chart of the data processing module of this application; Figure 4 This is the functional flow chart of the data analysis module of this application; Figure 5 Functional flow chart of the intervention program construction and follow-up feedback module for this application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Example 1 like Figure 1 As shown, this embodiment provides a community elderly cognitive decline prediction and intervention system based on machine learning, including the following functional modules: Data collection module: including basic data collection unit, neuroimaging data collection unit, cognitive function data collection unit, motor ability data collection unit, psychological state data collection unit and adverse health event collection unit; Data processing module: including basic data processing unit, neuroimaging data processing unit, cognitive function data processing unit, motor ability data processing unit, psychological state data processing unit and adverse health event processing unit; A basic data processing unit, used for processing the basic data to obtain a basic data feature vector; A neuroimaging data processing unit, used to process the neuroimaging data to obtain brain network feature data; A cognitive function data processing unit, used to process the cognitive function data to obtain a cognitive function data feature vector; A sports ability data processing unit, used for processing the sports ability data to obtain a sports ability data feature vector; A psychological state data processing unit, used for processing the psychological state data to obtain a psychological state data feature vector; An adverse health event processing unit, used to process the adverse health event to obtain a characteristic graph of the adverse health event; Data analysis module: including cognitive decline prediction model building unit and cognitive decline prediction unit; A cognitive decline prediction model construction unit, used to establish cognitive decline prediction models A and B, wherein the prediction model A is constructed by graph convolutional neural network training; and the prediction model B is constructed by multi-layer perceptron model training; A cognitive decline prediction unit: inputting the brain network feature data into a cognitive decline prediction model A to obtain a cognitive decline prediction result A; performing feature selection processing and splicing operations on the basic data feature vector, cognitive function data feature vector, motor ability data feature vector, and psychological state data feature vector to obtain a patient comprehensive data feature vector, inputting the patient comprehensive data feature vector into a cognitive decline prediction model B to obtain a cognitive decline prediction result B; and using the cognitive decline prediction results A and B to obtain a final cognitive decline prediction result; Intervention plan construction and follow-up feedback module: includes standard intervention plan generation unit, personalized intervention plan generation unit and follow-up feedback unit.

[0022] Example 2 like Figure 2 As shown, this embodiment provides a community elderly cognitive decline prediction and intervention system based on machine learning, wherein the data acquisition module specifically includes: Basic data collection unit, used to collect basic patient data, including age, height, weight, sleep data, personal chronic disease history and family chronic disease history; Sleep data is collected through wearable devices, which record sleep data within a statistical period, including the following indicators: total sleep time, deep sleep time, light sleep time and rapid eye movement time; the average indicator within a statistical period is used as the patient's sleep data.

[0023] Chronic disease history refers to the number and types of chronic diseases that the patient has, including hypertension and diabetes.

[0024] Family chronic disease history refers to the number and type of chronic diseases suffered by the patient's family members, including hypertension and diabetes.

[0025] A neuroimaging data acquisition unit, used to acquire neuroimaging data of patients, including brain structure magnetic resonance imaging data sMRI; Cognitive function data collection unit, used to collect patients' cognitive function data, including cognitive assessment data of memory, attention, executive function, language ability and visual-spatial ability; A sports ability data acquisition unit, used to acquire sports ability data of the patient, including sports video data, climbing test data, pace data and grip strength data, wherein the sports video data includes sit-up test video and balance test video; the sports video data and climbing test data are used to determine physical fitness data; A psychological state data collection unit is used to collect the patient's psychological state data, including depression assessment data, anxiety assessment data and quality of life assessment data; The adverse health event collection unit is used to collect patients' adverse health event data, including falls, delirium and hospitalization events.

[0026] This embodiment determines the types of multi-source and multi-modal data to be collected. This is based on the many influencing factors of cognitive decline identification determined in the early stage through reviewing research literature and expert discussion and scoring, and is collected and obtained through questionnaires, wearable devices, medical imaging, and sports video shooting.

[0027] Example 3 like Figure 3 As shown, this embodiment provides a community elderly cognitive decline prediction and intervention system based on machine learning, wherein the data processing module specifically includes: The basic data processing unit specifically includes: constructing a basic data-score value mapping table, including score values ​​corresponding to different numerical intervals of age, height, weight, sleep data, personal chronic disease history and family chronic disease history, quantizing the basic data into component values ​​according to the basic data-score value mapping table, and then normalizing all score values ​​to the [0,1] interval and splicing them to obtain the basic data feature vector.

[0028] The neuroimaging data processing unit specifically includes: performing T1-weighted imaging processing on the brain structure magnetic resonance imaging data sMRI to obtain a T1-weighted image, using freesurfer software to perform brain region segmentation on the T1-weighted image to obtain a brain region segmentation result; using the brain region included in the brain region segmentation result as an sMRI brain network node, and constructing an sMRI connection matrix based on the connection relationship between brain regions, and the sMRI brain network nodes and the sMRI connection matrix together constitute the brain network feature data.

[0029] The cognitive function data processing unit specifically includes: processing each type of data in the cognitive assessment data separately to obtain a memory assessment score, an attention assessment score, an executive function assessment score, a speech ability assessment score, and a visual-spatial ability assessment score; normalizing and mapping the above scores to the range of [0,1] and splicing them to obtain a cognitive function data feature vector.

[0030] The motor ability data processing unit specifically includes: performing motion posture detection and tracking on the sit-up test video data, calculating the number of times num that the patient completes the prescribed action in a test cycle T and the average time time to complete a prescribed action; performing motion posture detection and tracking on the balance test video, calculating the longest time t-max that the patient can maintain a balanced posture in a test cycle T; processing the climbing test data, extracting the patient's climbing height high and forward distance distance in a test cycle T; constructing a patient fitness data set fitness_data=(num, time, t-max, high, distance), evaluating the patient's physical fitness based on the patient fitness data set fitness_data, and obtaining a physical fitness evaluation result including excellent, good and poor; constructing a motor ability-score value mapping table, including score values ​​corresponding to different numerical intervals of pace and grip strength, and score values ​​corresponding to different evaluation results of physical fitness, quantizing the motor ability data into component values ​​based on the motor ability-score value mapping table, and then normalizing all the score values ​​to the [0,1] interval range and splicing them to obtain a motor ability data feature vector.

[0031] The psychological state data processing unit specifically includes: obtaining the patient's depression level, anxiety level and quality of life level according to the depression assessment data, anxiety assessment data and quality of life assessment data, constructing a psychological state-score value mapping table, including score values ​​corresponding to different levels of depression, anxiety and quality of life, quantizing the psychological state data into component values ​​according to the psychological state-score value mapping table, and then normalizing all the score values ​​to the range of [0,1] and splicing them to obtain a psychological state data feature vector.

[0032] The adverse health event processing unit specifically includes: establishing a visualization map of adverse health events, taking three types of events, namely, falls, delirium and hospitalization, as the first-level nodes of the map, and setting second-level nodes under the first-level nodes. The second-level nodes are specific events of falls, delirium and hospitalization each time, marking the specific time and cause of the events on the second-level nodes, calculating the co-occurrence probability between the first-level nodes, and if the co-occurrence probability is greater than a preset threshold, connecting the first-level nodes with lines; the co-occurrence probability is calculated as follows: , where p ijrepresents the co-occurrence probability between first-level nodes i and j, num_total(i,j) represents the sum of the number of second-level nodes corresponding to first-level nodes i and j, num_co(i,j) represents the number of co-occurrences between second-level nodes corresponding to first-level nodes i and j. If the difference in the occurrence time of the second-level nodes is within the preset time range, it is considered that the second-level nodes co-occur.

[0033] This embodiment determines the processing and feature extraction methods of different categories of data, converts the original unstructured data into structured data through a score value mapping table, and performs normalization mapping to obtain the characteristic values ​​corresponding to different categories of data. The brain structure magnetic resonance imaging data sMRI is processed in order to adapt to the subsequent cognitive decline prediction model A, which is built based on a graph convolutional neural network model, and its input is graph structure data. Establish a visualization map of adverse health events in order to more intuitively show the relationship map corresponding to the patient's adverse health events to medical staff in the future, so as to assist medical staff in judging the degree of cognitive decline of patients, make adjustments based on the standard intervention plan, and form a personalized intervention plan for patients.

[0034] Example 4 like Figure 4 As shown, this embodiment provides a community elderly cognitive decline prediction and intervention system based on machine learning, wherein the data analysis module specifically includes: The cognitive decline prediction unit specifically includes: determining the importance of basic data, cognitive function data, motor ability data and psychological state data according to the expert scoring method, and determining the percentage of retained data feature values ​​of basic data, cognitive function data, motor ability data and psychological state data according to the importance of each of the above data, the higher the importance, the higher the percentage of retained feature values; splicing the feature vectors after feature selection processing to form the patient's comprehensive data feature vector.

[0035] For example, if the importance of data is determined from high to low according to the expert scoring method: cognitive function data, motor ability data, basic data and psychological state data. Then the percentages of the number of retained eigenvalues ​​are determined to be 100%, 80%, 60% and 40% respectively. According to the above percentages, the original eigenvalues ​​of cognitive function data, motor ability data, basic data and psychological state data are screened to reduce the dimension of data processing. There are three original eigenvalues ​​of motor ability data, namely physical fitness eigenvalue, pace eigenvalue and grip strength eigenvalue. The corresponding eigenvalue retention percentage is 80%. After calculation and rounding down, the number of eigenvalues ​​obtained is 2. Therefore, the first two eigenvalues, namely physical fitness eigenvalue and pace eigenvalue, are retained, and the third eigenvalue, namely grip strength eigenvalue, is discarded.

[0036] The specific percentage can be set according to actual conditions and is not limited in this application.

[0037] The cognitive decline prediction unit specifically includes: the cognitive decline prediction results A and B are probability value vectors of the patient's cognitive decline levels being high, medium and low, respectively, output by models A and B; a final probability value vector V_final of the patient's cognitive decline levels being high, medium and low, respectively, is obtained by weighted summing the cognitive decline prediction results A and B, and the decline level corresponding to the maximum probability value in V_final is used as the patient's final cognitive decline prediction result.

[0038] This embodiment elaborates on the data analysis module. The cognitive decline prediction model construction unit realizes the construction of cognitive decline prediction models A and B. The prediction model A is constructed by graph convolutional neural network (GCN) training: features are extracted from sMRI data, including hippocampal volume, cortical thickness, shape, texture, etc. Studies have shown that the three-dimensional texture features of the hippocampus can reflect the pathological changes in the hippocampal structure of patients with AD and MCI, and the analysis combined with multiple features can better reflect the substantial differences in cognitive impairment. Different regions of the brain are regarded as nodes of the graph, and the connections between regions (such as white matter fiber connections) are regarded as edges. The features of each node can include various biomarkers extracted from sMRI data. The GCN model is designed to process graph structure data. GCN can capture complex relationships between nodes and is suitable for processing brain network structures. Advanced features can be extracted by stacking multiple graph convolution layers; the GCN model is trained using the collected sMRI data and the corresponding cognitive decline labels, and the performance of the model can be evaluated by cross-validation.

[0039] Prediction model B is constructed using multi-layer perceptron (MLP) training: Design the MLP structure, including input layer, hidden layer and output layer. There can be multiple hidden layers, each containing multiple neurons. Select a suitable activation function, such as ReLU or sigmoid function. The input feature vector is multi-source fusion data, including basic data, cognitive function data, motor ability data and psychological state data. The MLP model is trained using the extracted features and cognitive decline degree labels (such as high, medium and low). The model weights are optimized using the back propagation algorithm and gradient descent method.

[0040] Example 5 like Figure 5 As shown, this embodiment provides a community elderly cognitive decline prediction and intervention system based on machine learning, wherein the intervention program construction and follow-up feedback module specifically includes: A standard intervention program generating unit, used for determining a standard intervention program corresponding to the patient according to the final cognitive decline prediction result; A personalized intervention plan generating unit is used to adjust the standard intervention plan corresponding to the patient according to the patient's adverse health event characteristic map to obtain a personalized intervention plan for the patient; Specifically, the patient's adverse health event characteristic map and the patient's corresponding standard intervention plan are shown to medical staff, who then adjust the patient's corresponding standard intervention plan to generate a personalized intervention plan for the patient.

[0041] The follow-up feedback unit is used to follow up and intervene in patients according to their personalized intervention plans, and collect patients' basic data, neuroimaging data, cognitive function data, motor ability data, psychological state data and adverse health events according to preset cycles; and compare the changes in the degree of cognitive decline of patients after intervention.

[0042] In this embodiment, the intervention plan specifically includes the staff side and the patient side, wherein the staff side needs to collect and save patient information, conduct real-time health monitoring, behavioral guidance, and health education; the patient side needs to set goals, self-monitor, and self-feedback.

[0043] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, computer devices, or computer-readable storage media. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0044] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0045] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. A community elderly cognitive decline prediction and intervention system based on machine learning, characterized by: Includes the following functional modules: Data collection module: including basic data collection unit, neuroimaging data collection unit, cognitive function data collection unit, motor ability data collection unit, psychological state data collection unit and adverse health event collection unit; The basic data collection unit is used to collect basic data of the patient, including age, height, weight, sleep data, chronic disease history and family chronic disease history; The neuroimaging data acquisition unit is used to acquire the patient's neuroimaging data, including brain structure magnetic resonance imaging data sMRI; The cognitive function data acquisition unit is used to collect cognitive function data of the patient, including cognitive assessment data of memory, attention, executive function, language ability and visual-spatial ability; The sports ability data acquisition unit is used to collect sports ability data of the patient, including sports video data, climbing test data, pace data and grip strength data, wherein the sports video data includes sit-up test video and balance test video; the sports video data and climbing test data are used to determine physical fitness data; The psychological state data collection unit is used to collect the patient's psychological state data, including depression assessment data, anxiety assessment data and quality of life assessment data; The adverse health event collection unit is used to collect adverse health event data of patients, including fall events, delirium events and hospitalization events; Data processing module: including basic data processing unit, neuroimaging data processing unit, cognitive function data processing unit, motor ability data processing unit, psychological state data processing unit and adverse health event processing unit; The basic data processing unit is used to process the basic data to obtain a basic data feature vector; The neural image data processing unit is used to process the neural image data to obtain brain network feature data; A cognitive function data processing unit, used to process the cognitive function data to obtain a cognitive function data feature vector; The athletic ability data processing unit is used to process the athletic ability data to obtain an athletic ability data feature vector; The mental state data processing unit is used to process the mental state data to obtain a mental state data feature vector; The adverse health event processing unit is used to process the adverse health event to obtain a characteristic graph of the adverse health event; Data analysis module: including cognitive decline prediction model building unit and cognitive decline prediction unit; The cognitive decline prediction model construction unit is used to establish cognitive decline prediction models A and B, wherein the prediction model A is constructed by graph convolutional neural network training; and the prediction model B is constructed by multi-layer perceptron model training; The cognitive decline prediction unit: inputs the brain network feature data into the cognitive decline prediction model A to obtain cognitive decline prediction result A; performs feature selection processing and splicing operations on the basic data feature vector, cognitive function data feature vector, motor ability data feature vector and psychological state data feature vector to obtain a patient comprehensive data feature vector, inputs the patient comprehensive data feature vector into the cognitive decline prediction model B to obtain cognitive decline prediction result B; uses the cognitive decline prediction results A and B to obtain a final cognitive decline prediction result; Intervention plan construction and follow-up feedback module: including standard intervention plan generation unit, personalized intervention plan generation unit and follow-up feedback unit; The standard intervention program generating unit is used to determine a standard intervention program corresponding to the patient according to the final cognitive decline prediction result; The personalized intervention plan generating unit is used to adjust the standard intervention plan corresponding to the patient according to the patient's adverse health event characteristic map to obtain the patient's personalized intervention plan; The follow-up feedback unit is used to follow up and intervene in the patient according to the patient's personalized intervention plan, and collect the patient's basic data, neuroimaging data, cognitive function data, motor ability data, psychological state data and adverse health events according to a preset cycle; and compare the changes in the patient's cognitive decline after the intervention.

2. A community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The basic data processing unit specifically includes: constructing a basic data-score value mapping table, including score values ​​corresponding to different numerical intervals of age, height, weight, sleep data, personal chronic disease history and family chronic disease history, quantizing the basic data into component values ​​according to the basic data-score value mapping table, and then normalizing all score values ​​to the range of [0,1] and splicing them to obtain a basic data feature vector.

3. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The neuroimaging data processing unit specifically includes: performing T1-weighted imaging processing on the brain structure magnetic resonance imaging data sMRI to obtain a T1-weighted image, using freesurfer software to perform brain region segmentation on the T1-weighted image to obtain a brain region segmentation result; using the brain region included in the brain region segmentation result as an sMRI brain network node, and constructing an sMRI connection matrix based on the connection relationship between brain regions, and the sMRI brain network nodes and the sMRI connection matrix together constitute the brain network feature data.

4. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The cognitive function data processing unit specifically includes: processing each type of data in the cognitive assessment data separately to obtain a memory assessment score, an attention assessment score, an executive function assessment score, a speech ability assessment score, and a visual-spatial ability assessment score; normalizing and mapping the above scores to the range of [0,1] and splicing them to obtain a cognitive function data feature vector.

5. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The sports ability data processing unit specifically includes: performing motion posture detection and tracking on the sit-up test video data, calculating the number of times num that the patient completes the prescribed action in a test cycle T and the average time time to complete a prescribed action; performing motion posture detection and tracking on the balance test video, calculating the longest time t-max that the patient can maintain a balanced posture in a test cycle T; processing the climbing test data, extracting the patient's climbing height high and forward distance distance in a test cycle T; constructing a patient fitness data set fitness_data=(num, time, t-max, high, distance), evaluating the patient's fitness according to the patient fitness data set fitness_data, and obtaining fitness evaluation results including excellent, good and poor; constructing a sports ability-score value mapping table, including score values ​​corresponding to different numerical intervals of pace and grip strength, and score values ​​corresponding to different evaluation results of physical fitness, quantizing the sports ability data into component values ​​according to the sports ability-score value mapping table, and then normalizing all the score values ​​to the range of [0,1] and splicing them to obtain a sports ability data feature vector.

6. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The psychological state data processing unit specifically includes: obtaining the patient's depression level, anxiety level and quality of life level according to the depression assessment data, anxiety assessment data and quality of life assessment data, constructing a psychological state-score value mapping table, including score values ​​corresponding to different levels of depression, anxiety and quality of life, quantizing the psychological state data into component values ​​according to the psychological state-score value mapping table, and then normalizing all the score values ​​to the range of [0,1] and splicing them to obtain a psychological state data feature vector.

7. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The adverse health event processing unit specifically includes: establishing a visualization map of adverse health events, taking three types of events, namely, falls, delirium and hospitalization, as the first-level nodes of the map, and setting second-level nodes under the first-level nodes. The second-level nodes are specific events of falls, delirium and hospitalization each time, marking the specific time and cause of the events on the second-level nodes, calculating the co-occurrence probability between the first-level nodes, and if the co-occurrence probability is greater than a preset threshold, connecting the first-level nodes with lines; the co-occurrence probability is calculated as follows: , where p ij represents the co-occurrence probability between first-level nodes i and j, num_total(i,j) represents the sum of the number of second-level nodes corresponding to first-level nodes i and j, num_co(i,j) represents the number of co-occurrences between second-level nodes corresponding to first-level nodes i and j. If the difference in the occurrence time of the second-level nodes is within the preset time range, it is considered that the second-level nodes co-occur.

8. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The cognitive decline prediction unit specifically includes: determining the importance of basic data, cognitive function data, motor ability data and psychological state data according to the expert scoring method, and determining the percentage of retained data feature values ​​of basic data, cognitive function data, motor ability data and psychological state data according to the importance of each of the above data, the higher the importance, the higher the percentage of retained feature values; splicing the feature vectors after feature selection processing to form the patient's comprehensive data feature vector.

9. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The cognitive decline prediction unit specifically includes: the cognitive decline prediction results A and B are probability value vectors of the patient's cognitive decline levels being high, medium and low, respectively, output by models A and B; a final probability value vector V_final of the patient's cognitive decline levels being high, medium and low, respectively, is obtained by weighted summing the cognitive decline prediction results A and B, and the decline level corresponding to the maximum probability value in V_final is used as the patient's final cognitive decline prediction result.

10. The community elderly cognitive decline prediction and intervention system based on machine learning as claimed in claim 1, characterized in that: The personalized intervention plan generating unit specifically includes: showing the patient's adverse health event characteristic map and the patient's corresponding standard intervention plan to medical personnel, and the medical personnel adjusts the patient's corresponding standard intervention plan to generate a personalized intervention plan for the patient.

Images