Alzheimer's disease rehabilitation auxiliary system based on artificial intelligence

Abstract

The invention discloses an Alzheimer's disease rehabilitation assistance system based on artificial intelligence. The Alzheimer's disease rehabilitation assistance system comprises a data collection module, a cognitive evaluation module, a patient classification module, a rehabilitation planning module and a comprehensive assistance module. The invention belongs to the technical field of Alzheimer's disease rehabilitation management, and particularly relates to an Alzheimer's disease rehabilitation auxiliary system based on artificial intelligence, which realizes cognitive risk level judgment by acquiring multi-modal data of voice, writing, behavior and nursing evaluation, extracting structural features and constructing a cognitive evaluation method fusing rules and a lightweight model; performing multi-dimensional symptom grading in combination with patient behaviors and living abilities, screening personalized intervention tasks through an adaptive scoring mechanism, and generating a dynamic rehabilitation path and plan; closed-loop linkage among cognitive assessment, hierarchical classification and rehabilitation intervention is realized, and the accuracy and practicability of rehabilitation assistance are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of Alzheimer's disease rehabilitation management, and specifically relates to an Alzheimer's disease rehabilitation assistance system based on artificial intelligence. Background Art

[0002] An Alzheimer's disease rehabilitation assistance system based on artificial intelligence is an intelligent medical assistance system that uses intelligent technologies such as machine learning, natural language processing, computer vision, and the Internet of Things to provide personalized cognitive training, behavior monitoring, memory assistance, and rehabilitation support for Alzheimer's disease patients. This system represents a typical application of "artificial intelligence in medicine" in the field of neurodegenerative diseases and meets the technological development trend and the rigid needs of an aging society.

[0003] However, in existing Alzheimer's disease rehabilitation assistance systems, there are problems such as single data sources and information fragmentation between modules, which makes it difficult for traditional systems to dynamically track patients' rehabilitation status and provide intervention feedback. Since cognitive assessment results are often disconnected from subsequent care plans, real-time updates and precise connections cannot be achieved. At the same time, most systems focus on a single questionnaire or interview, ignoring the cognitive performance of patients in aspects such as language expression, motor coordination, and behavioral interaction; in existing patient cognitive assessment processes, there are traditional methods that generally rely on regular scales and manual question-and-answer. Such an assessment process lacks multi-dimensional input and is easily affected by subjective judgments. Especially in the early stage of the disease, mild cognitive decline is often masked. At the same time, the traditional assessment cycle is long and the tool standardization degree is not high, making it difficult to adapt to patient groups with different cultural and educational backgrounds; in the existing rehabilitation plan planning process, patient classification is usually based on clinical experience or a small number of index scores, lacking modeling analysis of the interaction between cognitive behaviors. In reality, patients with high cognitive scores but disordered daily routines and large mood swings may still belong to high-risk groups, and such complex situations are often difficult to identify in traditional methods; in existing rehabilitation plan planning, there is a problem with the production of rehabilitation plans based on a single rule, and the intervention content often still faces insufficient adaptability. The recommended training tasks may be too difficult, the execution frequency is unreasonable, or seriously inconsistent with the patient's life rhythm and compliance, affecting the actual implementation effect. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the existing technologies, the present invention provides an Alzheimer's disease rehabilitation assistance system based on artificial intelligence. In the existing Alzheimer's disease rehabilitation assistance systems, there are problems such as single data source and information fragmentation among various modules, which makes it difficult for traditional systems to dynamically track the rehabilitation status of patients and provide intervention feedback. Since the cognitive assessment results are often disjointed from the subsequent care plans, real-time updates and precise connection cannot be achieved. At the same time, most systems focus on a single questionnaire or interview, ignoring the cognitive performance of patients in aspects such as language expression, movement coordination, and behavioral interaction. This solution creatively adopts an automatic intelligent process design that combines cognitive assessment, patient classification, and rehabilitation support, realizing full-process dynamic linkage from data collection to intervention recommendation, and greatly improving the system's perception ability and response ability to the patient's status. In the existing patient cognitive assessment process, there are traditional methods that generally rely on regular scales and manual questions and answers. Such an assessment process lacks multi-dimensional input and is easily affected by subjective judgment. Especially in the early stage of the disease, mild cognitive decline is often masked. At the same time, the traditional assessment cycle is long and the tool standardization degree is not high, making it difficult to adapt to patient groups with different cultural and educational backgrounds. This solution creatively adopts a multi-modal lightweight interactive assessment method for patient cognitive assessment. By introducing a multi-modal lightweight assessment mechanism and constructing an assessment model with objective indicators such as speech rate, writing trajectory, and task completion degree, it not only enhances the ability to identify minor cognitive variations, but also reduces the dependence on professionals for assessment execution, and is more suitable for repeated assessments in daily follow-up and long-term management. In the existing rehabilitation plan planning process, patient classification is usually based on clinical experience or a small number of index scores, lacking modeling analysis of the interaction between cognitive behaviors. In reality, patients with high cognitive scores but disordered living patterns and large mood swings may still belong to high-risk groups, and such complex situations are often difficult to identify in traditional methods. This solution creatively adopts a cognitive-behavioral fusion hierarchical classification method for symptom severity classification, realizing the integration of comprehensive cognitive score trends, self-care ability changes, and emotional states, and establishing unified grading labels, making patient stratification more accurate and providing a more targeted basis for subsequent intervention tasks. In the existing rehabilitation plan planning, there is a problem that the production of rehabilitation plans based on a single rule still often faces insufficient adaptability. The recommended training tasks may be too difficult, the execution frequency is unreasonable, or seriously inconsistent with the patient's life rhythm and compliance, affecting the actual implementation effect. This solution creatively adopts a rehabilitation plan adaptive multi-dimensional generation method for rehabilitation plan planning, realizing comprehensive optimization of the intervention plan in terms of executability, matching degree, and continuity.

[0005] The technical solution adopted by the present invention is as follows: An Alzheimer's disease rehabilitation assistance system based on artificial intelligence provided by the present invention includes a data collection module, a cognitive assessment module, a patient classification module, a rehabilitation support module, and a comprehensive assistance module;

[0006] The data collection module is used for data collection and processing. Through data collection and processing, rehabilitation assistance optimization data is obtained, and the rehabilitation assistance optimization data is sent to the cognitive assessment module, the patient classification module, and the rehabilitation support module;

[0007] The cognitive assessment module is used for patient cognitive assessment. Through patient cognitive assessment, patient cognitive assessment reference data is obtained, and the patient cognitive assessment reference data is sent to the comprehensive assistance module;

[0008] The patient classification module is used for classifying the severity of symptoms. Through classifying the severity of symptoms, patient classification reference data is obtained, and the patient classification reference data is sent to the rehabilitation support module and the comprehensive assistance module;

[0009] The rehabilitation support module is used for rehabilitation plan planning. Through rehabilitation plan planning, rehabilitation plan reference planning data is obtained, and the rehabilitation plan reference planning data is sent to the comprehensive assistance module;

[0010] The comprehensive assistance module is used for Alzheimer's disease rehabilitation assistance. Through Alzheimer's disease rehabilitation assistance, comprehensive reference data for Alzheimer's disease rehabilitation assistance is obtained.

[0011] Further, the data collection and processing is used for collecting multi-dimensional health data and performing data optimization processing. Specifically, through multi-source and multi-modal data acquisition, original data of Alzheimer's disease rehabilitation records is obtained, and through data optimization processing, rehabilitation assistance optimization data is obtained;

[0012] The multi-source and multi-modal data acquisition specifically acquires voice interaction records, writing and drawing records, behavior interaction records, and nursing questionnaire records; the original data of Alzheimer's disease rehabilitation records includes voice interaction data, writing and drawing data, behavior interaction data, and nursing assessment data;

[0013] The data optimization processing specifically refers to sequentially performing artificial feature annotation extraction, data cleaning, data normalization, and structured encapsulation on the original data of Alzheimer's disease rehabilitation records to obtain rehabilitation assistance optimization data;

[0014] The rehabilitation assistance optimization data includes structured voice interaction feature data, structured writing and drawing feature data, structured behavior interaction feature data, and structured nursing assessment feature data.

[0015] Furthermore, the patient cognitive assessment is used to evaluate the cognitive state of Alzheimer's patients. Specifically, based on the rehabilitation assistance optimization data, a multi-modal lightweight interaction assessment method is adopted to conduct patient cognitive assessment and obtain patient cognitive assessment reference data, including the following steps: feature reconstruction and extraction, construction of a primary assessment layer, construction of a lightweight learning layer, construction of a comprehensive cognitive scoring layer, construction of an improved segmented nested cognitive assessment model, training of the patient cognitive assessment model, and patient cognitive assessment;

[0016] The feature reconstruction and extraction is specifically to extract speech features, writing trajectory features, and behavior interaction features from the rehabilitation assistance optimization data, and through feature normalization, construct combined cognitive feature data;

[0017] The construction of the primary assessment layer is specifically to construct a feature threshold rule based on the combined cognitive feature data, conduct rule matching assessment, and obtain risk primary assessment score vector data;

[0018] The construction of the lightweight learning layer is specifically to construct a standard LightGBM lightweight model and output machine assessment score vector data based on the combined cognitive feature data;

[0019] The construction of the comprehensive cognitive scoring layer is specifically to fuse the risk primary assessment score vector data and the machine assessment score vector data through a fusion coefficient to obtain comprehensive cognitive scoring data;

[0020] The construction of the improved segmented nested cognitive assessment model is specifically to construct a segmented nested structure through the primary assessment layer, the lightweight learning layer, and the comprehensive cognitive scoring layer to obtain an improved segmented nested cognitive assessment model;

[0021] The training of the patient cognitive assessment model is specifically to train the patient cognitive assessment model through the feature reconstruction and extraction and the construction of the improved segmented nested cognitive assessment model to obtain the patient cognitive assessment model Model MLIE ;

[0022] The patient cognitive assessment is specifically to use the patient cognitive assessment model Model MLIE , and based on the rehabilitation assistance optimization data, conduct patient cognitive assessment to obtain patient cognitive assessment reference data;

[0023] The patient cognitive assessment reference data includes a cognitive risk score and a risk level label.

[0024] Furthermore, the symptom severity classification is used to stage the patient's condition according to the patient's cognitive and behavioral data, specifically, based on the patient's cognitive assessment reference data and the rehabilitation assistance optimization data, a cognitive behavior fusion hierarchical classification method is used to classify the symptom severity to obtain patient classification reference data, including the following steps: preliminary risk judgment screening, symptom hierarchical classification, fusion comprehensive judgment and symptom severity classification;

[0025] The preliminary risk judgment and screening specifically includes constructing an initial classification rule, performing preliminary symptom risk screening based on the patient cognitive assessment reference data and the rehabilitation assistance optimization data, and obtaining preliminary classification and grading data; the preliminary classification and grading data specifically refers to the classification labels for the initial risk judgment, including mild risk, moderate risk and severe risk;

[0026] The initial classification rules specifically include severe initial judgment rules, moderate initial judgment rules, mild initial judgment rules and early warning initial judgment rules;

[0027] The symptom hierarchical classification is specifically to reconstruct features based on the patient cognitive assessment reference data and the rehabilitation auxiliary optimization data, use a lightweight multi-layer perceptron as a basic prediction model, and introduce a classifier improved by a regulating factor to perform symptom hierarchical classification output to obtain model classification data;

[0028] The calculation formula of the classifier improved by introducing the adjustment factor is:

[0029] ;

[0030] Where P i is the probability output of the i-th model classification data, i is the category index, e is the natural base, z i is the output probability of the basic prediction model corresponding to the i-th category, j is the summed category index, z j is the output probability of the basic prediction model corresponding to the i-th summation category. The value ranges of the category index i and the summation category index j are both {1, 2, 3}, which are used to represent mild risk, moderate risk and severe risk respectively. T is an adjustment factor, which is used to pull in the probability distribution of different categories and smoothly optimize the probability output;

[0031] The output structure calculation formula of the model classification data is:

[0032] ;

[0033] Where P is the model classification data, P moderate is the classification probability of the mild risk model, P mild is the classification probability of the moderate risk model, P severe is the classification probability of the severe risk model;

[0034] The fusion comprehensive judgment specifically refers to fusing the output results by setting the model confidence threshold to obtain the classification output data of the comprehensive symptom severity. The calculation formula is as follows:

[0035] The fusion comprehensive judgment specifically refers to fusing the output results by setting the model confidence threshold to obtain the classification output data of the comprehensive symptom severity. The calculation formula is as follows:

[0036] ;

[0037] In the formula, Level is the classification output data of the comprehensive symptom severity, Label rule is the initial risk judgment classification label, is the classification probability value of the model classification data, which is used as the confidence evaluation value of the model classification for the preliminary judgment classification. When the classification probability value of the model classification data is greater than 0.6, the initial risk judgment classification label is taken as the classification output data of the comprehensive symptom severity. argmax(·) is the function to find the maximum value, and P is the model classification data;

[0038] The symptom severity classification specifically refers to constructing a hierarchical classification system combining fixed rules and intelligent models through the preliminary risk judgment screening, the symptom stratification classification, and the fusion comprehensive judgment, and performing the symptom severity classification through the hierarchical classification system combining fixed rules and intelligent models to obtain the patient classification reference data;

[0039] The patient classification reference data includes the symptom severity category label and the classification evaluation trend data.

[0040] Furthermore, the rehabilitation plan planning is used to generate a personalized rehabilitation training and care plan according to the patient classification and cognitive evaluation results. Specifically, based on the patient classification reference data, the patient cognitive evaluation reference data, and the rehabilitation assistance optimization data, a rehabilitation plan self-adaptive multi-dimensional generation method is adopted to perform rehabilitation plan planning to obtain the rehabilitation plan reference planning data, including the following steps: constructing a rehabilitation intervention resource library, intervention preliminary selection, constructing a multi-objective comprehensive scoring function, rehabilitation path generation, and rehabilitation plan planning;

[0041] The construction of the rehabilitation intervention resource library specifically refers to setting the rehabilitation intervention resource library based on the patient classification reference data, the patient cognitive evaluation reference data, and the rehabilitation assistance optimization data by setting a basic rehabilitation plan to obtain the structured rehabilitation resource library data;

[0042] The basic rehabilitation plan includes rehabilitation intervention tasks, applicable populations, intervention intensities, execution time requirements, and patient rehabilitation tendency correlation scores;

[0043] The rehabilitation intervention tasks include cognitive training tasks, emotional support tasks, behavior intervention tasks, and mixed tasks; the applicable populations include low-risk populations, medium-risk populations, and high-risk populations; the intervention intensities include mild intervention, moderate intervention, and severe intervention; the patient rehabilitation tendency correlation score is specifically calculated using cosine similarity;

[0044] The initial intervention selection is specifically to construct a standard graph structure based on the patient's current cognitive state and symptom grading, and perform an initial screening of the rehabilitation intervention plan based on the data in the structured rehabilitation resource library to obtain initial screening rehabilitation intervention data;

[0045] The standard graph structure specifically uses symptom grading, intervention goals, intervention tasks, and behavioral characteristics as nodes, and uses the correlation degree between nodes as edges. The correlation degree between nodes specifically uses the patient rehabilitation tendency correlation score as the correlation degree value;

[0046] The construction of the multi-objective comprehensive scoring function is specifically to construct a multi-objective comprehensive scoring function and perform a comprehensive screening of tasks based on the data in the structured rehabilitation resource library to obtain optimized rehabilitation intervention list data;

[0047] The calculation formula for the comprehensive task screening is:

[0048] ;

[0049] In the formula, Score r is the output of the multi-objective comprehensive scoring function, used to represent the comprehensive score value of the rth rehabilitation intervention task, where r is the rehabilitation intervention task index, is the correlation degree weight, S r is the patient rehabilitation tendency correlation score, is the state fit weight, I r is the patient's cognitive state of the rth rehabilitation intervention task, I target is the target state of the cognitive state, is the patient cooperation weight, is the patient compliance estimate, specifically taking the weighted sum of the emotional state score and the daily routine score in the structured nursing assessment characteristic data, is the patient execution weight, is the patient physical condition estimate, specifically taking the weighted sum of the self-care ability score and the daily routine score in the structured nursing assessment characteristic data;

[0050] R finalIt is to optimize the rehabilitation intervention list data, which is used to represent the top K rehabilitation intervention tasks with the highest scoring values. Top-K(·) is a function to take the top K highest values, and K is the dimension of the optimized rehabilitation intervention list;

[0051] The generation of the rehabilitation path is specifically to group and construct a plan for rehabilitation planning based on the optimized rehabilitation intervention list data. The grouping is specifically to group the rehabilitation planning according to the types of rehabilitation intervention tasks. The plan construction is specifically to generate intervention details and an intervention schedule based on the intervention intensity and execution time requirements corresponding to the rehabilitation intervention tasks, so as to obtain rehabilitation path data;

[0052] The rehabilitation plan planning is specifically to conduct manual verification and integration based on the rehabilitation path data to obtain rehabilitation plan reference planning data;

[0053] The rehabilitation plan reference planning data includes a rehabilitation task list, a rehabilitation stage schedule, and rehabilitation intervention explanation information.

[0054] Furthermore, the Alzheimer's rehabilitation assistance is used to integrate the data and strategies of each module and provide comprehensive rehabilitation assistance suggestions. Specifically, it combines the rehabilitation plan reference planning data, the patient classification reference data, and the patient cognitive assessment reference data to conduct data integration and provide Alzheimer's rehabilitation assistance, so as to obtain comprehensive reference data for Alzheimer's rehabilitation assistance.

[0055] The beneficial effects obtained by the present invention using the above solution are as follows:

[0056] (1) In the existing Alzheimer's rehabilitation assistance system, there are problems such as single data source and information fragmentation between modules, which makes it difficult for traditional systems to dynamically track the patient's rehabilitation status and provide intervention feedback. Since the cognitive assessment results are often disconnected from the subsequent care plan, real-time update and precise connection cannot be achieved. At the same time, most systems focus on a single questionnaire or interview, ignoring the cognitive performance of patients in aspects such as language expression, movement coordination, and behavioral interaction. This solution creatively adopts an automatic intelligent process design that combines cognitive assessment, patient classification, and rehabilitation support, realizing the full-process dynamic linkage from data collection to intervention recommendation, and greatly improving the system's perception ability and response ability to the patient's status;

[0057] (2) In the existing patient cognitive assessment process, there are traditional methods that generally rely on periodic scales and manual questions and answers. Such an assessment process lacks multi-dimensional input and is easily affected by subjective judgment. Especially in the early stages of the disease, mild cognitive decline is often concealed. At the same time, the traditional assessment cycle is long, the tool standardization is not high, and it is difficult to adapt to the technical problems of patient groups with different cultural and educational backgrounds. This program creatively adopts a multimodal lightweight interactive assessment method to conduct patient cognitive assessment. By introducing a multimodal lightweight assessment mechanism, an assessment model is constructed with objective indicators such as speech speed, writing trajectory, and task completion. This not only enhances the ability to identify subtle cognitive variations, but also reduces the dependence of assessment execution on professionals. It is more suitable for repeated assessments in daily follow-up and long-term management.

[0058] (3) In the existing rehabilitation program planning process, patient classification is usually based on clinical experience or a small number of indicators, and there is a lack of modeling analysis of the interaction between cognition and behavior. In reality, patients with high cognitive scores but disordered lifestyles and large mood swings may still belong to the high-risk group. Such complex conditions are often difficult to identify in traditional methods. This program creatively adopts a cognitive-behavioral fusion hierarchical classification method to classify symptom severity, achieves a comprehensive cognitive score trend, changes in self-care ability and emotional state, and establishes a unified grading label, making patient stratification more accurate and providing a more targeted basis for subsequent intervention tasks.

[0059] (4) In the existing rehabilitation program planning, there are rehabilitation programs based on a single rule. The intervention content still often faces the problem of insufficient adaptability. The recommended training tasks may be difficult, the execution frequency is unreasonable, or they are seriously inconsistent with the patient's life rhythm and compliance, which affects the actual implementation effect. Technical problems, this program creatively adopts the adaptive multi-dimensional generation method of rehabilitation program to carry out rehabilitation program planning, and realizes the comprehensive optimization of the intervention plan in terms of feasibility, matching and continuity. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of the structure of an artificial intelligence-based Alzheimer's disease rehabilitation assistance system provided by the present invention;

[0061] Figure 2 A flowchart of the steps performed for the cognitive assessment module;

[0062] Figure 3 A flowchart of the steps performed by the patient classification module;

[0063] Figure 4 Flowchart of the steps performed for the rehabilitation support module.

[0064] The accompanying drawings are used to provide a further understanding of the present invention and form a part of the description. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. Detailed Embodiments

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0066] Example 1, refer to Figure 1 , the technical solutions adopted by the present invention are as follows: An Alzheimer's disease rehabilitation assistance system based on artificial intelligence provided by the present invention includes a data collection module, a cognitive assessment module, a patient classification module, a rehabilitation support module, and a comprehensive assistance module;

[0067] The data collection module is used for data collection and processing. Through data collection and processing, rehabilitation assistance optimization data is obtained, and the rehabilitation assistance optimization data is sent to the cognitive assessment module, the patient classification module, and the rehabilitation support module;

[0068] The cognitive assessment module is used for patient cognitive assessment. Through patient cognitive assessment, patient cognitive assessment reference data is obtained, and the patient cognitive assessment reference data is sent to the comprehensive assistance module;

[0069] The patient classification module is used for classifying the severity of symptoms. Through classifying the severity of symptoms, patient classification reference data is obtained, and the patient classification reference data is sent to the rehabilitation support module and the comprehensive assistance module;

[0070] The rehabilitation support module is used for rehabilitation plan planning. Through rehabilitation plan planning, rehabilitation plan reference planning data is obtained, and the rehabilitation plan reference planning data is sent to the comprehensive assistance module;

[0071] The comprehensive assistance module is used for Alzheimer's disease rehabilitation assistance. Through Alzheimer's disease rehabilitation assistance, Alzheimer's disease rehabilitation assistance comprehensive reference data is obtained.

[0072] Example 2, this example is based on the above example, refer to Figure 1 , the data collection and processing is used for collecting multi-dimensional health data and performing data optimization processing. Specifically, through multi-source and multi-modal data collection, the original data of Alzheimer's disease rehabilitation records is obtained, and through data optimization processing, the rehabilitation assistance optimization data is obtained;

[0073] The multi-source and multi-modal data collection specifically collects voice interaction records, writing and drawing records, behavior interaction records, and nursing questionnaire records; the original data of Alzheimer's rehabilitation records includes voice interaction data, writing and drawing data, behavior interaction data, and nursing assessment data;

[0074] The voice interaction records are collected through simple tasks or greeting conversations 1 to 3 times a day to obtain voice interaction data;

[0075] The writing and drawing records are collected through handwriting board drawing or writing 2 to 3 times a week to obtain writing and drawing data;

[0076] The behavior interaction records are collected through cognitive test tasks including graphic pairing, color selection, and clock filling to obtain behavior interaction data;

[0077] The nursing questionnaire records are collected through questionnaire filling tasks using questionnaires including activity ability scales and geriatric depression scales to obtain nursing assessment data;

[0078] The data optimization process specifically refers to sequentially performing manual feature annotation extraction, data cleaning, data normalization, and structured encapsulation on the original data of Alzheimer's rehabilitation records to obtain rehabilitation assistance optimization data;

[0079] The rehabilitation assistance optimization data includes structured voice interaction feature data, structured writing and drawing feature data, structured behavior interaction feature data, and structured nursing assessment feature data;

[0080] The structured voice interaction feature data includes original voice records, average speech rate features, average pause time features, intonation fluctuation features, and lexical richness features;

[0081] The structured writing and drawing feature data includes original writing and drawing records, number of writing strokes features, average trajectory speed features of writing and drawing, variance features of note turning angles, and trajectory complexity features;

[0082] The structured behavior interaction feature data includes original behavior interaction records, average reaction time features, total number of incorrect cognitions features, total number of correct cognitions features, and cognitive task completion rate features;

[0083] The structured nursing assessment feature data includes self-care ability scores, emotional state scores, daily routine scores, and number of times waking up at night features;

[0084] The structured voice interaction feature data, structured writing and drawing feature data, and structured behavior interaction feature data are used for patient cognitive assessment; the structured nursing assessment feature data is used for symptom severity classification and rehabilitation plan planning.

[0085] By performing the above operations, in the existing Alzheimer's disease rehabilitation assistance system, there are problems of single data source and information fragmentation between modules, which makes it difficult for traditional systems to dynamically track the rehabilitation status of patients and provide intervention feedback. Since the cognitive assessment results are often disconnected from the subsequent care plans and cannot be updated in real time and accurately connected, and most systems focus on a single questionnaire or interview, ignoring the technical problems of cognitive performance in aspects such as patients' language expression, motor coordination, and behavioral interaction. This solution creatively adopts an automatic intelligent process design that combines cognitive assessment, patient classification, and rehabilitation support, realizing the full-process dynamic linkage from data collection to intervention recommendation, and greatly improving the system's perception ability and response ability to patients' states.

[0086] Example 3, this example is based on the above example, referring to Figure 2 , the patient cognitive assessment is used to evaluate the cognitive status of Alzheimer's disease patients. Specifically, based on the optimized rehabilitation assistance data, a multi-modal lightweight interaction assessment method is adopted to conduct patient cognitive assessment to obtain patient cognitive assessment reference data, including the following steps: feature reconstruction extraction, construction of a primary assessment layer, construction of a lightweight learning layer, construction of a comprehensive cognitive scoring layer, construction of an improved segmented nested cognitive assessment model, patient cognitive assessment model training, and patient cognitive assessment;

[0087] The feature reconstruction extraction is specifically to extract speech features, writing trajectory features, and behavioral interaction features from the optimized rehabilitation assistance data, and through feature normalization, construct joint cognitive feature data. The calculation formula is:

[0088] ;

[0089] In the formula, X is the joint cognitive feature data, concat(·) is the feature connection operation function, norm(·) is the feature normalization operation function, x speech is the speech feature data, used to represent the feature data extracted from the structured writing and drawing feature data, x hand is the writing trajectory feature, used to represent the feature data extracted from the structured writing and drawing feature data, x click is the behavioral interaction feature, used to represent the feature data extracted from the structured behavioral interaction feature data;

[0090] The construction of the primary assessment layer is specifically to construct a feature threshold rule based on the joint cognitive feature data and conduct rule matching assessment to obtain risk primary assessment score vector data;

[0091] The construction of the lightweight learning layer specifically involves constructing a standard LightGBM lightweight model, and based on the combined cognitive feature data, outputting machine evaluation score vector data. The calculation formula is as follows:

[0092] ;

[0093] In the formula, P ML is the machine evaluation score vector data, f clf (·) is the prediction function of the standard LightGBM lightweight model, and X is the combined cognitive feature data;

[0094] The construction of the comprehensive cognitive scoring layer specifically involves fusing the risk primary evaluation score vector data and the machine evaluation score vector data through a fusion coefficient to obtain comprehensive cognitive scoring data. The calculation formula is as follows:

[0095] ;

[0096] In the formula, Score cog is the comprehensive cognitive scoring data, is the fusion coefficient with a value range of [0.3, 0.7], is the S-shaped activation function, R is the risk primary evaluation score vector data, and P ML is the machine evaluation score vector data;

[0097] The construction of the improved piecewise nested cognitive evaluation model specifically involves constructing a piecewise nested structure through the primary evaluation layer, the lightweight learning layer, and the comprehensive cognitive scoring layer to obtain the improved piecewise nested cognitive evaluation model;

[0098] The training of the patient cognitive evaluation model specifically involves training the patient cognitive evaluation model through the feature reconstruction extraction and the construction of the improved piecewise nested cognitive evaluation model to obtain the patient cognitive evaluation model Model MLIE ;

[0099] The patient cognitive evaluation specifically involves using the patient cognitive evaluation model Model MLIE , and based on the rehabilitation assistance optimization data, performing patient cognitive evaluation to obtain patient cognitive evaluation reference data;

[0100] The patient cognitive evaluation reference data includes a cognitive risk score and a risk level label;

[0101] The value range of the cognitive risk score is [0, 1];

[0102] The risk level label includes normal cognitive level, mild risk level, moderate risk level, and high risk level;

[0103] The cognitive risk score corresponding to the normal cognitive level has a value range of [0, 0.3);

[0104] The cognitive risk score corresponding to the mild risk level is in the range of [0.3, 0.6);

[0105] The cognitive risk score corresponding to the moderate risk level is in the range of [0.6, 0.8);

[0106] The cognitive risk score corresponding to the high risk level ranges from [0.8, 1].

[0107] By performing the above operations, in view of the fact that in the existing patient cognitive assessment process, there are traditional methods that generally rely on periodic scales and manual questions and answers. Such an assessment process lacks multi-dimensional input and is easily affected by subjective judgment, especially in the early stages of the disease, when mild cognitive decline is often concealed. At the same time, the traditional assessment cycle is long, the tool standardization is not high, and it is difficult to adapt to the technical problems of patient groups with different cultural and educational backgrounds. This solution creatively adopts a multimodal lightweight interactive assessment method to conduct patient cognitive assessment. By introducing a multimodal lightweight assessment mechanism, an assessment model is constructed with objective indicators such as speech speed, writing trajectory, and task completion. This not only enhances the ability to identify subtle cognitive variations, but also reduces the dependence of assessment execution on professionals, making it more suitable for repeated assessments in daily follow-up and long-term management.

[0108] Embodiment 4, this embodiment is based on the above embodiment, see Figure 3 The symptom severity classification is used to stage the patient's condition according to the patient's cognitive and behavioral data, specifically, based on the patient's cognitive assessment reference data and the rehabilitation assistance optimization data, a cognitive behavior fusion hierarchical classification method is used to classify the symptom severity to obtain patient classification reference data, including the following steps: preliminary risk judgment screening, symptom hierarchical classification, fusion comprehensive judgment and symptom severity classification;

[0109] The preliminary risk judgment and screening specifically includes constructing an initial classification rule, performing preliminary symptom risk screening based on the patient cognitive assessment reference data and the rehabilitation assistance optimization data, and obtaining preliminary classification and grading data; the preliminary classification and grading data specifically refers to the classification labels for the initial risk judgment, including mild risk, moderate risk and severe risk;

[0110] The initial classification rules specifically include severe initial judgment rules, moderate initial judgment rules, mild initial judgment rules and early warning initial judgment rules;

[0111] The calculation formula of the severe initial judgment rule is:

[0112] ;

[0113] Where Score cog is the comprehensive cognitive score data, used as the preliminary screening and evaluation value of symptom risk, and represents the cognitive risk score in the cognitive evaluation reference data of the patient. A is the self-care ability score in the structured nursing assessment feature data;

[0114] The calculation formula of the medium initial judgment rule is:

[0115] ;

[0116] The calculation formula of the mild initial judgment rule is:

[0117] ;

[0118] The calculation formula of the degree warning rule is:

[0119] ;

[0120] Where is the cognitive score change value, used to represent the difference between two consecutive cognitive scores;

[0121] The symptom stratification and classification is specifically to reconstruct the features based on the patient's cognitive evaluation reference data and the rehabilitation assistance optimization data, use a lightweight multi-layer perceptron as the basic prediction model, and introduce a classifier improved by a regulation factor to perform symptom stratification and classification output to obtain model classification data;

[0122] The calculation formula of the feature reconstruction is:

[0123] ;

[0124] Where X class is the reconstructed input feature data, Score cog is the comprehensive cognitive score data, s speech is the voice score sub-item data, s hand is the writing trajectory score sub-item data, s click is the behavior interaction score sharing data, A is the self-care ability score in the structured nursing assessment feature data, E is the emotional state score in the structured nursing assessment feature data, B is the living routine score in the structured nursing assessment feature data, is the cognitive score change value;

[0125] The calculation formula of the classifier improved by introducing a regulation factor is:

[0126] ;

[0127] Where P iis the probability output of the i-th model classification data, i is the category index, e is the natural base, z i is the output probability of the basic prediction model corresponding to the i-th category, j is the summed category index, z j is the output probability of the basic prediction model corresponding to the i-th summation category. The value ranges of the category index i and the summation category index j are both {1, 2, 3}, which are used to represent mild risk, moderate risk and severe risk respectively. T is an adjustment factor, which is used to pull in the probability distribution of different categories and smoothly optimize the probability output;

[0128] Preferably, the value of the adjustment factor T is set to 1.5;

[0129] The output structure calculation formula of the model classification data is:

[0130] ;

[0131] Where P is the model classification data, P moderate is the classification probability of the mild risk model, P mild is the classification probability of the moderate risk model, P severe is the classification probability of the severe risk model;

[0132] The fusion comprehensive judgment is specifically to set the model confidence threshold to fuse the output results to obtain the comprehensive symptom severity classification output data, and the calculation formula is:

[0133] ;

[0134] In the formula, Level is the comprehensive symptom severity classification output data, Label rule is the initial risk classification label, It is the classification probability value of the model classification data, which is used as the confidence evaluation value of the model classification for the preliminary judgment classification. When the value of is greater than 0.6, the risk initial judgment classification label is taken as the comprehensive symptom severity classification output data, argmax(·) is the maximum value function, and P is the model classification data;

[0135] The symptom severity classification is specifically to construct a hierarchical classification system combining fixed rules and intelligent models through the risk preliminary judgment screening, the symptom hierarchical classification and the fusion comprehensive judgment, and to classify the severity of symptoms through the hierarchical classification system combining fixed rules and intelligent models to obtain patient classification reference data;

[0136] The patient classification reference data includes symptom severity category labels and classification assessment trend data.

[0137] By performing the above operations, in the existing rehabilitation program planning process, there is a technical problem that patient classification is usually based on clinical experience or scoring with a small number of indicators, lacking the modeling and analysis of the interaction between cognitive behaviors. In reality, patients with high cognitive scores but disordered living patterns and large mood swings may still belong to the high-risk group, and such complex situations are often difficult to identify in traditional methods. In response to this, this solution creatively adopts a cognitive-behavior integration hierarchical classification method to classify the severity of symptoms, realizes the comprehensive cognitive score trend, self-care ability changes, and emotional state, and establishes a unified grading label, making patient stratification more accurate and providing a more targeted basis for subsequent intervention tasks.

[0138] Example 5, this example is based on the above example, refer to Figure 4 The rehabilitation program planning is used to generate a personalized rehabilitation training and care plan according to patient classification and cognitive assessment results. Specifically, based on the patient classification reference data, the patient cognitive assessment reference data, and the rehabilitation assistance optimization data, a rehabilitation program adaptive multi-dimensional generation method is adopted to perform rehabilitation program planning to obtain rehabilitation program reference planning data, including the following steps: constructing a rehabilitation intervention resource library, intervention preliminary selection, constructing a multi-objective comprehensive scoring function, rehabilitation path generation, and rehabilitation program planning;

[0139] The construction of the rehabilitation intervention resource library is specifically to set up the rehabilitation intervention resource library based on the patient classification reference data, the patient cognitive assessment reference data, and the rehabilitation assistance optimization data through setting a basic rehabilitation program to obtain structured rehabilitation resource library data;

[0140] The basic rehabilitation program includes rehabilitation intervention tasks, applicable populations, intervention intensities, execution time requirements, and patient rehabilitation tendency correlation scores;

[0141] The rehabilitation intervention tasks include cognitive training tasks, emotional support tasks, behavior intervention tasks, and mixed tasks; the applicable populations include low-risk populations, medium-risk populations, and high-risk populations; the intervention intensities include mild intervention, medium intervention, and high intervention; the patient rehabilitation tendency correlation score is specifically calculated using cosine similarity;

[0142] The intervention preliminary selection is specifically to construct a standard graph structure based on the patient's current cognitive state and symptom grading, and perform a preliminary screening of the rehabilitation intervention plan based on the structured rehabilitation resource library data to obtain preliminary screening rehabilitation intervention data;

[0143] For the standard graph structure, specifically, symptom grading, intervention goals, intervention tasks, and behavioral characteristics are used as nodes, and the degree of association between nodes is used as an edge. The degree of association between nodes specifically uses the correlation score of the patient's rehabilitation tendency as the value of the degree of association;

[0144] The construction of the multi-objective comprehensive scoring function is specifically to construct a multi-objective comprehensive scoring function, and based on the data in the structured rehabilitation resource library, perform task comprehensive screening to obtain optimized rehabilitation intervention list data;

[0145] The calculation formula for the task comprehensive screening is:

[0146] ;

[0147] In the formula, Score r is the output of the multi-objective comprehensive scoring function, used to represent the comprehensive score value of the r-th rehabilitation intervention task, r is the rehabilitation intervention task index, is the degree of association weight, S r is the correlation score of the patient's rehabilitation tendency, is the state fit weight, I r is the cognitive state of the patient for the r-th rehabilitation intervention task, I target is the target state of the cognitive state, is the patient cooperation weight, is the patient compliance estimate, and the specific value is the weighted sum of the emotional state score and the living pattern score in the structured nursing assessment feature data, is the patient executive weight, is the patient physical condition estimate, and the specific value is the weighted sum of the self-care ability score and the living pattern score in the structured nursing assessment feature data;

[0148] R final is the optimized rehabilitation intervention list data, used to represent the K rehabilitation intervention tasks with the highest score values, Top-K(·) is the function of taking the top K highest values, and K is the dimension of the optimized rehabilitation intervention list;

[0149] Preferably, the value of the dimension K of the optimized rehabilitation intervention list is set to 3; the degree of association weight , the state fit weight , the patient cooperation weight and the patient executive weight are initially set to 0.25;

[0150] The generation of the rehabilitation path is specifically to group the rehabilitation plan and construct a plan based on the data in the optimized rehabilitation intervention list. The grouping is specifically to group the rehabilitation plan according to the type of rehabilitation intervention tasks. The plan construction is specifically to generate intervention details and an intervention schedule based on the intervention intensity and execution time requirements corresponding to the rehabilitation intervention tasks, so as to obtain rehabilitation path data;

[0151] The rehabilitation plan planning is specifically to perform manual verification and integration based on the rehabilitation path data to obtain rehabilitation plan reference planning data;

[0152] The rehabilitation plan reference planning data includes a rehabilitation task list, a rehabilitation stage schedule, and rehabilitation intervention explanation information.

[0153] By performing the above operations, in the existing rehabilitation plan planning, there is a problem that the rehabilitation plan is produced based on a single rule, and its intervention content still often faces the problem of insufficient adaptability. The recommended training tasks may be too difficult, the execution frequency is unreasonable, or it seriously does not match the patient's life rhythm and compliance, affecting the actual implementation effect. The present solution creatively adopts a method for multi-dimensional adaptive generation of rehabilitation plans to perform rehabilitation plan planning, realizing a comprehensive optimization of the intervention plan in terms of executability, matching degree, and continuity.

[0154] Example 6, this example is based on the above example, refer to Figure 1 , the Alzheimer's disease rehabilitation assistance is used to integrate the data and strategies of each module and provide comprehensive rehabilitation assistance suggestions. Specifically, it combines the rehabilitation plan reference planning data, the patient classification reference data, and the patient cognitive assessment reference data to perform data integration and Alzheimer's disease rehabilitation assistance to obtain Alzheimer's disease rehabilitation assistance comprehensive reference data.

[0155] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprises", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion, so that a process and method comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process and method.

[0156] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0157] The above describes the present invention and its implementation manners. Such description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural manners and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. An Alzheimer's disease rehabilitation assistance system based on artificial intelligence, characterized in that: It includes a data collection module, a cognitive assessment module, a patient classification module, a rehabilitation support module, and a comprehensive assistance module; The data collection module is used for data collection and processing. Through data collection and processing, rehabilitation assistance optimization data is obtained, and the rehabilitation assistance optimization data is sent to the cognitive assessment module, the patient classification module, and the rehabilitation support module; The cognitive assessment module is used for patient cognitive assessment. Through patient cognitive assessment, patient cognitive assessment reference data is obtained, and the patient cognitive assessment reference data is sent to the comprehensive assistance module; The patient classification module is used for classifying the severity of symptoms. Through classifying the severity of symptoms, patient classification reference data is obtained, and the patient classification reference data is sent to the rehabilitation support module and the comprehensive assistance module; The rehabilitation support module is used for rehabilitation plan planning. Through rehabilitation plan planning, rehabilitation plan reference planning data is obtained, and the rehabilitation plan reference planning data is sent to the comprehensive assistance module; The comprehensive assistance module is used for Alzheimer's disease rehabilitation assistance. Through Alzheimer's disease rehabilitation assistance, Alzheimer's disease rehabilitation assistance comprehensive reference data is obtained.

2. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 1, characterized in that: The data collection and processing is used for collecting multi-dimensional health data and performing data optimization processing. Specifically, through multi-source and multi-modal data collection, original Alzheimer's disease rehabilitation record data is obtained, and through data optimization processing, rehabilitation assistance optimization data is obtained; The multi-source and multi-modal data collection specifically collects voice interaction records, writing and drawing records, behavior interaction records, and nursing questionnaire records; the original Alzheimer's disease rehabilitation record data includes voice interaction data, writing and drawing data, behavior interaction data, and nursing assessment data; The data optimization processing specifically refers to sequentially performing manual feature annotation extraction, data cleaning, data normalization, and structured encapsulation on the original Alzheimer's disease rehabilitation record data to obtain rehabilitation assistance optimization data; The rehabilitation assistance optimization data includes structured voice interaction feature data, structured writing and drawing feature data, structured behavior interaction feature data, and structured nursing assessment feature data.

3. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 2, characterized in that: The patient cognitive assessment is used for assessing the cognitive state of Alzheimer's disease patients. Specifically, based on the rehabilitation assistance optimization data, a multi-modal lightweight interaction assessment method is adopted to perform patient cognitive assessment to obtain patient cognitive assessment reference data, including the following steps: feature reconstruction extraction, constructing a primary assessment layer, constructing a lightweight learning layer, constructing a comprehensive cognitive scoring layer, constructing an improved segmented nested cognitive assessment model, patient cognitive assessment model training, and patient cognitive assessment.

4. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 3, characterized in that: The feature reconstruction extraction specifically extracts voice features, writing trajectory features, and behavior interaction features from the rehabilitation assistance optimization data, and through feature normalization, joint cognitive feature data is constructed; The constructing of the primary assessment layer specifically constructs a feature threshold rule based on the joint cognitive feature data, performs rule matching assessment, and obtains risk primary assessment score vector data; The construction of the lightweight learning layer specifically outputs the machine evaluation score vector data by constructing a standard LightGBM lightweight model based on the combined cognitive feature data. The construction of the comprehensive cognitive scoring layer specifically fuses the risk primary evaluation score vector data and the machine evaluation score vector data through a fusion coefficient to obtain the comprehensive cognitive scoring data. The construction of the improved segmented nested cognitive evaluation model specifically constructs a segmented nested structure through the primary evaluation layer, the lightweight learning layer, and the comprehensive cognitive scoring layer to obtain the improved segmented nested cognitive evaluation model. The training of the patient cognitive assessment model specifically involves training the patient cognitive assessment model Model through the feature reconstruction extraction and the construction of the improved segmented nested cognitive assessment model MLIE ; The patient cognitive assessment specifically uses the patient cognitive assessment model Model MLIE , and based on the rehabilitation assistance optimization data, conducts a patient cognitive assessment to obtain patient cognitive assessment reference data; The patient cognitive evaluation reference data includes cognitive risk scores and risk level labels.

5. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 4, characterized in that: The symptom severity classification is used to stage-classify the patient's condition based on the patient's cognitive and behavioral data. Specifically, based on the patient cognitive evaluation reference data and the rehabilitation assistance optimization data, a cognitive-behavioral fusion hierarchical classification method is used for symptom severity classification to obtain the patient classification reference data, including the following steps: risk preliminary judgment and screening, symptom hierarchical classification, fusion comprehensive judgment, and symptom severity classification.

6. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 5, characterized in that: The risk preliminary judgment and screening specifically constructs an initial classification rule, and based on the patient cognitive evaluation reference data and the rehabilitation assistance optimization data, conducts preliminary screening of symptom risks to obtain the preliminary classification and grading data; the preliminary classification and grading data specifically refers to the risk initial judgment classification labels, including mild risk, moderate risk, and severe risk. The initial classification rule specifically includes a severe initial judgment rule, a moderate initial judgment rule, a mild initial judgment rule, and a warning initial judgment rule. The symptom hierarchical classification specifically reconstructs features based on the patient cognitive evaluation reference data and the rehabilitation assistance optimization data, uses a lightweight multi-layer perceptron as the basic prediction model, and introduces a classifier improved by a regulation factor to output the symptom hierarchical classification to obtain the model classification data. The calculation formula for the classifier improved by introducing the regulation factor is: Wherein, P i is the probability output of the classification data of the i-th type of model, i is the class index, e is the natural base, z i is the output probability of the basic prediction model corresponding to the i-th class, j is the summation class index, z j is the output probability of the basic prediction model corresponding to the i-th summation class, and the value ranges of the class index i and the summation class index j are both {1, 2, 3}, which are used in sequence to represent low risk, medium risk, and high risk, and T is a regulation factor, which is used to pull in the probability distributions of different classes and smooth and optimize the probability output; The calculation formula for the output structure of the model classification data is: P = [P moderate , P mild , P severe ; Wherein, P is the model classification data, P moderate is the classification probability of the low-risk model, P mild is the classification probability of the medium-risk model, P severe is the classification probability of the high-risk model; The fusion comprehensive judgment specifically fuses the output results by setting a model confidence threshold to obtain the comprehensive symptom severity classification output data, and the calculation formula is: Wherein, Level is the output data of the comprehensive symptom severity classification, and Label rule is the initial risk judgment classification label, is the classification probability value of the model classification data, which is used as the confidence evaluation value of the model classification for the preliminary judgment classification. When the classification probability value of the model classification data is greater than 0.6, the initial risk judgment classification label is taken as the output data of the comprehensive symptom severity classification. argmax(·) is the function for obtaining the maximum value, and P is the model classification data; The symptom severity classification specifically constructs a hierarchical classification system combining fixed rules and intelligent models through the risk preliminary judgment and screening, the symptom hierarchical classification, and the fusion comprehensive judgment, and conducts symptom severity classification through the hierarchical classification system combining fixed rules and intelligent models to obtain the patient classification reference data. The patient classification reference data includes symptom severity category labels and classification evaluation trend data.

7. An Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 6, characterized in that: The rehabilitation plan is used to generate a personalized rehabilitation training and care plan according to the patient classification and cognitive assessment results. Specifically, based on the patient classification reference data, the patient cognitive assessment reference data, and the rehabilitation assistance optimization data, a self-adaptive multi-dimensional generation method for the rehabilitation plan is used to perform rehabilitation plan planning to obtain rehabilitation plan reference planning data, including the following steps: constructing a rehabilitation intervention resource library, initial intervention selection, constructing a multi-objective comprehensive scoring function, generating a rehabilitation path, and planning a rehabilitation plan.

8. The Alzheimer's disease rehabilitation assistance system based on artificial intelligence according to claim 7, characterized in that: The construction of the rehabilitation intervention resource library specifically involves setting up the rehabilitation intervention resource library by setting a basic rehabilitation plan based on the patient classification reference data, the patient cognitive assessment reference data, and the rehabilitation assistance optimization data to obtain structured rehabilitation resource library data. The basic rehabilitation plan includes rehabilitation intervention tasks, applicable populations, intervention intensities, execution time requirements, and a correlation score for the patient's rehabilitation tendency. The rehabilitation intervention tasks include cognitive training tasks, emotional support tasks, behavior intervention tasks, and mixed tasks; the applicable populations include low-risk populations, medium-risk populations, and high-risk populations; the intervention intensities include mild intervention, moderate intervention, and severe intervention; the correlation score for the patient's rehabilitation tendency is specifically calculated using cosine similarity. The initial intervention selection specifically involves constructing a standard graph structure based on the patient's current cognitive state and symptom grading, and performing an initial screening of the rehabilitation intervention plan based on the structured rehabilitation resource library data to obtain initial screening rehabilitation intervention data. The standard graph structure specifically uses symptom grading, intervention goals, intervention tasks, and behavior characteristics as nodes, and the degree of association between the nodes as edges. The degree of association between the nodes specifically uses the correlation score for the patient's rehabilitation tendency as the value of the degree of association. The construction of the multi-objective comprehensive scoring function specifically involves constructing a multi-objective comprehensive scoring function and performing a comprehensive task screening based on the structured rehabilitation resource library data to obtain optimized rehabilitation intervention list data. The calculation formula for the comprehensive task screening is: Where Score r is the output of the multi-objective comprehensive scoring function, used to represent the comprehensive score value of the r-th rehabilitation intervention task, r is the rehabilitation intervention task index, β1 is the correlation weight, S r is the correlation score of the patient's rehabilitation tendency, β2 is the state fitness weight, I r is the cognitive state of the patient for the r-th rehabilitation intervention task, I target is the target state of the cognitive state, β3 is the patient cooperation weight, C adhere is the patient compliance estimate, specifically the weighted sum of the emotional state score and the daily routine score in the structured nursing assessment feature data, β4 is the patient executive weight, F mobility is the patient physical condition estimate, specifically the weighted sum of the self-care ability score and the daily routine score in the structured nursing assessment feature data; R final is the optimized rehabilitation intervention list data, used to represent the K rehabilitation intervention tasks with the highest score values. Top-K(·) is the function of taking the top K highest values, and K is the dimension of the optimized rehabilitation intervention list; The generation of the rehabilitation path specifically involves grouping and plan construction for rehabilitation planning based on the optimized rehabilitation intervention list data. The grouping specifically groups the rehabilitation planning according to the type of rehabilitation intervention task, and the plan construction specifically generates intervention details and an intervention schedule based on the intervention intensity and execution time requirements corresponding to the rehabilitation intervention task to obtain rehabilitation path data. The rehabilitation plan planning specifically involves performing manual verification and integration based on the rehabilitation path data to obtain rehabilitation plan reference planning data. The rehabilitation plan reference planning data includes a rehabilitation task list, a rehabilitation stage schedule, and rehabilitation intervention explanation information. The Alzheimer's disease rehabilitation assistance is used to integrate the data and strategies of each module and provide comprehensive rehabilitation assistance suggestions. Specifically, by combining the rehabilitation plan reference planning data, the patient classification reference data, and the patient cognitive assessment reference data, data integration is performed and Alzheimer's disease rehabilitation assistance is carried out to obtain Alzheimer's disease rehabilitation assistance comprehensive reference data.

Images