Evaluation system and method applied to cognitive disorder intervention

Abstract

The invention belongs to the technical field of cognitive impairment intervention evaluation, and particularly relates to an evaluation system and method applied to cognitive impairment intervention. According to the method, the cognitive state is evaluated and the training tasks are allocated based on the real data of the user in the natural life scene, so that the one-sidedness of only depending on a laboratory or one-time evaluation is avoided, the training and intervention are closer to the real demand and life situation of the user, and the training efficiency is improved through multi-source data fusion. The influence of single signal noise or accidental fluctuation on judgment is reduced, the robustness and objectivity of evaluation are improved, indexes such as cognitive performance, intervention indexes and deviation degree are numeralized, long-term tracking and transverse comparison of the effects of different intervention measures are facilitated, interpretable quantitative evidences are provided for clinical personnel, and the accuracy of evaluation is improved. Intervention indexes are calculated in real time, intervention measures are automatically adjusted or reselected according to feedback results, and a quick-response closed-loop system is formed, so that the time from problem discovery to effective intervention is shortened, and the intervention conversion rate is increased.

Description

Technical Field

[0001] This invention belongs to the field of cognitive impairment intervention and evaluation technology, specifically relating to an evaluation system and method for cognitive impairment intervention. Background Technology

[0002] With the accelerating aging of the global population, cognitive impairment has become a major public health challenge affecting the health and quality of life of the elderly. Cognitive impairment not only places a heavy burden of care and economic pressure on patients and their families, but also poses a severe test to social medical resources. Early identification and effective intervention for cognitive impairment are key to slowing disease progression and improving patient prognosis. The incidence of cognitive impairment (including mild cognitive impairment (MCI), Alzheimer's disease (AD), and vascular dementia) continues to rise with the increasing global aging population. Currently, there is no cure, and early intervention to slow disease progression has become a core clinical strategy. Numerous non-pharmacological interventions (such as cognitive training, exercise therapy, and neuromodulation) and pharmacological intervention programs have emerged, but how to scientifically quantify the effectiveness of interventions remains a challenge for the industry.

[0003] Existing technologies in the field of cognitive impairment intervention cannot continuously and comprehensively collect users' multimodal data in natural life scenarios, lack the ability to continuously perceive the real state, and fail to personalize and adapt assessments and task selection to the context, resulting in insufficient intervention accuracy. Furthermore, there is a lack of quantitative effect verification and closed-loop optimization after intervention, making it impossible to achieve dynamic strategy iteration based on results. The safety and effectiveness of intervention measures are difficult to judge in real time, increasing the risk of use. Summary of the Invention

[0004] The purpose of this invention is to provide an evaluation system and method for cognitive impairment intervention, which can improve the effectiveness of cognitive impairment intervention by collecting multimodal data from natural scenes, conducting real-time assessment and dynamic intervention, and performing closed-loop verification.

[0005] The specific technical solution adopted by this invention is as follows: An evaluation method applied to cognitive impairment interventions includes: Acquire multimodal data of target users in natural life scenarios, and acquire cognitive training tasks based on the multimodal data; Cognitive training data of target users is obtained based on cognitive training tasks, and cognitive performance information is obtained based on cognitive training data; Intervention indices are obtained based on multimodal data and cognitive performance information; Intervention measures are obtained based on the intervention index, the intervention measures are implemented, and cognitive performance information after the intervention is re-obtained; Obtain the cognitive shift before and after the intervention and generate a multidimensional evaluation report.

[0006] In a preferred embodiment, the steps of acquiring multimodal data of the target user in natural life scenarios and obtaining cognitive training tasks based on the multimodal data include: Acquire multimodal data of target users in natural life scenarios, including behavioral data, physiological data, and environmental data; Behavioral, physiological, and environmental values ​​are obtained based on multimodal data; Obtain multimodal weights, which include behavioral weights, physiological weights, and environmental weights; Cognitive state values ​​are obtained based on behavioral values, physiological values, environmental values, and multimodal weights. Obtain the task table, which includes multiple cognitive state value intervals and the cognitive training task corresponding to each cognitive state value interval; The corresponding cognitive training task is obtained from the task table based on the cognitive state value range corresponding to the cognitive state value.

[0007] In a preferred embodiment, the steps of acquiring cognitive training data of the target user based on a cognitive training task, and acquiring cognitive performance information based on the cognitive training data, include: Perform cognitive training tasks and acquire cognitive training data, which includes operational feature information, memory matching information, and attention duration information. Based on cognitive training data, obtain corresponding operational feature values, memory matching values, and attention duration values; Obtain the training weights, which include operational feature weights, memory matching weights, and attention persistence weights. Cognitive performance values ​​are obtained based on operational feature values, memory matching values, attention duration values, and cognitive training weights, and are then labeled as cognitive performance information.

[0008] In a preferred embodiment, the step of obtaining the intervention index based on multimodal data and cognitive performance information includes: Obtain the corresponding cognitive state value based on multimodal data; Obtain the corresponding cognitive performance value based on cognitive performance information; Cognitive weights are obtained based on cognitive training tasks, where cognitive weights include cognitive state weights and cognitive performance weights. An intervention index is obtained based on cognitive state value, cognitive performance value, and cognitive weight.

[0009] In a preferred embodiment, the steps of obtaining intervention measures based on intervention indices, implementing intervention measures, and re-obtaining post-intervention cognitive performance information include: Obtain the intervention table, which includes multiple intervention index ranges and the corresponding intervention measures for each intervention index range; Based on the intervention index range corresponding to the intervention index, obtain the corresponding intervention measures from the intervention table; Implement intervention measures for target users; Feedback period based on intervention measures; Obtain cognitive performance information after intervention during the feedback period.

[0010] In a preferred embodiment, the step of obtaining feedback time periods based on intervention measures includes: Obtain the time points for implementing intervention measures and mark them as the start time of the feedback period; Obtain the duration compensation table, which includes multiple intervention measures and the duration coefficient corresponding to each intervention measure; Obtain the corresponding duration coefficient from the duration compensation table based on the intervention measures; Obtain the benchmark intervention index and benchmark duration; The feedback duration is obtained based on the intervention index, the baseline intervention index, the duration coefficient, and the baseline duration. The end time of the feedback period is obtained based on the feedback duration and the start time of the feedback period, and the feedback period is constructed.

[0011] In a preferred embodiment, the steps of obtaining the cognitive intervention shift before and after the intervention and generating a multidimensional evaluation report include: Obtain the corresponding pre-intervention cognitive performance value based on pre-intervention cognitive performance information; Obtain the corresponding cognitive performance value after intervention based on the cognitive performance information after intervention; Cognitive intervention offset is obtained based on cognitive performance values ​​before and after intervention. Obtain the scoring table, which includes multiple cognitive intervention offset intervals and the intervention score corresponding to each cognitive intervention offset interval; Based on the cognitive intervention offset interval corresponding to the cognitive intervention offset, the corresponding intervention score is obtained from the scoring table, and a multidimensional evaluation report is generated.

[0012] In a preferred embodiment, after obtaining the cognitive intervention shift before and after the intervention and generating a multidimensional evaluation report, the method further includes: Obtain the cognitive intervention offset threshold; Determine whether the cognitive intervention offset is lower than the cognitive intervention offset; If the cognitive intervention deviation is not less than the cognitive intervention deviation, the intervention is deemed effective. If the cognitive intervention deviation is lower than the cognitive intervention deviation, the intervention is deemed ineffective, an alarm is issued, the intervention is reacquired and implemented, until the intervention is effective.

[0013] The present invention also provides an evaluation system for cognitive impairment intervention, used in the aforementioned evaluation method for cognitive impairment intervention, comprising: The cognitive training module is used to acquire multimodal data of target users in natural life scenarios, and to acquire cognitive training tasks based on the multimodal data; The cognitive performance module acquires the cognitive training data of the target user based on the cognitive training task, and obtains cognitive performance information based on the cognitive training data; The intervention numerical module obtains intervention indices based on multimodal data and cognitive performance information; The cognitive intervention module obtains intervention measures based on the intervention index, implements the intervention measures, and re-acquires cognitive performance information after the intervention. The intervention feedback module is used to obtain the cognitive intervention deviation before and after the intervention and generate a multidimensional evaluation report.

[0014] And, an assessment terminal for cognitive impairment intervention, comprising: One or more processors; A storage device on which one or more programs are stored; When one or more programs are executed by one or more processors, the one or more processors implement evaluation methods applicable to cognitive impairment interventions.

[0015] The technical effects achieved by this invention are as follows: This invention assesses cognitive states and assigns training tasks based on real-world data from users in natural life scenarios, avoiding the bias of relying solely on laboratory tests or one-off assessments. This makes training and interventions more closely aligned with users' real needs and life situations. By fusing multi-source data on behavior, physiology, environment, and training performance, it reduces the impact of single-signal noise or occasional fluctuations on judgment, improving the robustness and objectivity of the assessment. It quantifies indicators such as cognitive performance, intervention index, and deviation, facilitating long-term tracking, horizontal comparison of the effects of different interventions, and providing interpretable quantitative evidence to clinicians. It calculates the intervention index in real time and automatically adjusts or reselects interventions based on feedback results, forming a rapid response closed-loop system. This shortens the time from problem discovery to effective intervention and improves the intervention conversion rate. Through modular configuration of task tables, intervention tables, duration compensation tables, and scoring tables, it facilitates rapid adjustment of strategies according to different populations or scenarios, possessing good engineering implementation and large-scale promotion potential. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method provided by the present invention; Figure 2 This is a system module diagram provided by the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0020] Furthermore, the present invention will be described in detail with reference to the schematic diagrams. When describing the embodiments of the present invention in detail, the schematic diagrams are merely examples for ease of explanation and should not limit the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 As shown, an evaluation method for cognitive impairment intervention is provided, including: S1. Obtain multimodal data of target users in natural life scenarios, and obtain cognitive training tasks based on multimodal data; S2. Obtain cognitive training data of the target user based on the cognitive training task, and obtain cognitive performance information based on the cognitive training data; S3. Obtain intervention indices based on multimodal data and cognitive performance information; S4. Obtain intervention measures based on the intervention index, implement the intervention measures, and re-obtain cognitive performance information after the intervention; S5. Obtain the cognitive intervention deviation before and after the intervention, and generate a multidimensional evaluation report.

[0022] As described in steps S1 to S5 above, behavioral (e.g., activity patterns), physiological (e.g., heart rate), and environmental (e.g., noise) data are continuously or periodically collected in the user's daily life environment. Feature extraction is performed on the raw data to obtain behavioral, physiological, and environmental values. These values ​​are then weighted and fused according to preset or learned multimodal weights (behavioral weights, physiological weights, and environmental weights) to calculate a comprehensive cognitive state value. Based on this cognitive state value, a corresponding interval is matched in the task table, thereby selecting or generating a cognitive training task that is suitable for the current state. The user executes the issued cognitive training task, and simultaneously, operational characteristic information (e.g., reaction time), memory matching information (e.g., recall accuracy), and attention duration information are recorded during the training process. Information such as fixation duration is used to convert training data into operational feature values, memory matching values, and attention duration values. These are then fused according to training weights (operational feature weights, memory matching weights, and attention duration weights) to obtain cognitive performance values, which serve as a quantitative representation of cognitive performance information. Simultaneously, cognitive state values ​​from multimodal data and cognitive performance values ​​from training are weighted according to cognitive weights (e.g., cognitive state weights and cognitive performance weights) to calculate a unified intervention index. This index measures whether intervention is currently needed and its urgency or intensity. The intervention index is mapped to a predefined intervention table, and specific intervention measures (such as prompts, brief training, medication assessment prompts, professional intervention recommendations, etc.) are selected for the corresponding intervals during the feedback period. The system re-collects cognitive training or performance data to assess immediate effects, calculates the difference in cognitive performance values ​​before and after intervention, and obtains the cognitive intervention deviation (which can be determined by dividing the post-intervention cognitive performance value by the pre-intervention cognitive performance value). Based on a scoring table, the deviation is mapped to an intervention score, and a multidimensional evaluation report is generated (including multimodal indicators, training performance, intervention index, deviation, and whether a threshold was reached). This supports decision-making by clinical or caregiver personnel. By assessing cognitive status and assigning training tasks based on real-life data from users in natural life scenarios, it avoids the bias of relying solely on laboratory or one-time assessments. This makes training and intervention more closely aligned with users' real needs and life situations, through multi-dimensional analysis of behavior, physiology, environment, and training performance. Source data fusion reduces the impact of single signal noise or occasional fluctuations on judgment, improves the robustness and objectivity of assessment, and quantifies indicators such as cognitive performance, intervention index, and bias, facilitating long-term tracking, horizontal comparison of the effects of different interventions, and providing interpretable quantitative evidence to clinicians. It can calculate the intervention index in real time and automatically adjust or reselect interventions based on feedback results, forming a rapid response closed-loop system, thereby shortening the time from problem discovery to effective intervention and improving the intervention conversion rate. Through modular configuration of task tables, intervention tables, duration compensation tables, and scoring tables, it is easy to quickly adjust strategies according to different populations (age, disease course, comorbidities) or different scenarios, and has good potential for engineering implementation and large-scale promotion.

[0023] In a preferred embodiment, the steps of acquiring multimodal data of the target user in natural life scenarios and obtaining cognitive training tasks based on the multimodal data include: S101. Acquire multimodal data of the target user in natural life scenarios, including behavioral data, physiological data and environmental data; S102. Obtain behavioral, physiological, and environmental values ​​based on multimodal data; S103. Obtain multimodal weights, including behavioral weights, physiological weights, and environmental weights. S104. Obtain cognitive state values ​​based on behavioral values, physiological values, environmental values, and multimodal weights; S105. Obtain the task table, which includes multiple cognitive state value intervals and the cognitive training task corresponding to each cognitive state value interval. S106. Obtain the corresponding cognitive training task from the task table based on the cognitive state value interval corresponding to the cognitive state value.

[0024] As described in steps S101 to S106 above, in a user's natural life scenario, raw data is automatically collected through various terminals and sensors: behavioral data (such as activity type, movement trajectory, touch, interaction logs, voice, facial expression events), physiological data (such as heart rate, heart rate variability, skin conductance, sleep quality, respiratory rate), and environmental data (such as noise, light intensity, temperature, location, indoor and outdoor conditions). Collection methods may include wearable devices, smartphone sensors, home IoT devices, and environmental sensors. Collection supports real-time or time-segmented batch uploads, and timestamps and collection contexts are annotated. Various raw signals are preprocessed (denoising, interpolation, time alignment, normalization) and features are extracted: behavioral values ​​(which can be extracted from daily activity frequency), physiological values ​​(which can be extracted from average heart rate), and environmental values ​​(which can be extracted from average noise level). These are then converted into a unified data set according to preset rules. Numerical values ​​(such as behavioral value lookup tables, which include the frequency of multiple daily activities and the corresponding behavioral values ​​converted to the same dimension for each daily activity frequency; physiological and environmental values ​​can also be obtained through corresponding lookup tables, or through model training or dynamic learning to obtain corresponding values ​​converted to the same dimension) are used. Based on user profiles (age, disease course, previous medical records), scenarios (such as early morning, mealtime), historical effect evaluations, and model training results, the importance weights of each modality are determined or dynamically learned. These weights can be fixed or adaptively adjusted through online learning (Bayesian update, weighted regression, or neural network attention mechanisms) to ensure that key modalities are given higher weights for different users or at different times. Weighted fusion is used to combine behavioral, physiological, and environmental values ​​with multimodal weights to obtain a comprehensive cognitive state value. The formula for calculating the cognitive state value is as follows: In the formula, R represents the cognitive state value. Let W represent the behavior weight, and W represent the behavior value. This is represented by physiological weight, and S represents physiological value. Let H represent the environmental weight, and H represent the environmental value. The task table is a discrete mapping table or rule base, containing several cognitive state value ranges (which can be numerical ranges) and a set of cognitive training tasks (task type, difficulty, duration, feedback strategy, etc.) corresponding to each range. The task table supports expert setting (preset by clinical and rehabilitation experts) and data-driven generation (optimization based on historical efficacy), and supports storage by user group or individual. The current cognitive state value is mapped to the corresponding cognitive state value range in the task table, and suitable cognitive training tasks are retrieved from the range. Task retrieval can be combined with strategy optimization (e.g., prioritizing those with the best historical performance or prioritizing user-preferred tasks). Through multimodal perception in natural life scenarios, task allocation is directly associated with the user's current real cognitive state, avoiding single standardized training and improving the relevance and acceptability of training. Combining behavioral, physiological, and environmental signals can reduce the risk of misjudgment caused by single sensor failure or noise interference, making cognitive state estimation more reliable and stable. Multimodal weights can be adjusted online according to individual differences and historical efficacy, thereby optimizing the identification and task distribution strategies for different disease stages or different individuals and improving long-term efficacy.

[0025] In a preferred embodiment, the steps of acquiring cognitive training data of the target user based on a cognitive training task, and acquiring cognitive performance information based on the cognitive training data, include: S201. Perform a cognitive training task and obtain cognitive training data, wherein the cognitive training data includes operational feature information, memory matching information and attention duration information; S202. Obtain the corresponding operational feature values, memory matching values, and attention duration values ​​based on cognitive training data; S203. Obtain training weights, including operational feature weights, memory matching weights, and attention duration weights. S204. Obtain cognitive performance values ​​based on operational feature values, memory matching values, attention duration values, and cognitive training weights, and label them as cognitive performance information.

[0026] As described in steps S201 to S204 above, during the execution of the issued cognitive training task, raw interaction data and sensor data related to the task are collected in real time or periodically to form a cognitive training data set. This cognitive training data includes operational feature information (interaction response time between the user and the task interface), memory matching information (test item accuracy in memory-based tasks), and attention duration information (attention-related indicators, such as gaze duration). The raw training data is preprocessed (denoising, outlier handling, interpolation, and time alignment), and structured feature operation feature values ​​(such as average reaction time), memory matching values ​​(such as immediate recall accuracy), and attention duration values ​​(such as average gaze duration) are extracted. Training weights (operation feature weights, memory matching values, and attention duration) are then obtained. Memory matching weight and attention duration weight are used. Weights can be determined through multiple methods: expert rule setting (clinical or rehabilitation experts set initial weight configurations based on task type and goals, e.g., memory matching has a higher weight in memory tasks), data-driven learning (using historical training data and efficacy labels to learn optimal weights through machine learning methods such as linear regression, weighted logistic regression, random forest, neural networks, or attention mechanisms), and online adaptive updating (using Bayesian updating, exponentially weighted moving average, or reinforcement learning strategies to dynamically adjust weights based on recent task performance and long-term efficacy feedback). The cognitive performance value is calculated by combining operational feature values, memory matching values, and attention duration values ​​with the training weights. The formula for calculating the cognitive performance value is... In the formula, Z represents the cognitive performance value, a represents the operational feature weight, T represents the operational feature value, b represents the memory matching weight, J represents the memory matching value, c represents the attention duration weight, and Y represents the attention duration value, a+b+c=1. In the actual calculation process, in order to ensure the scientificity and comparability of the cognitive performance value, each indicator was standardized. Since the original data of each indicator have different dimensions (such as reaction time in milliseconds and accuracy in percentage), they were unified to the same numerical range through normalization before weighted calculation, thereby avoiding the weight offset problem caused by the difference in dimensions between different indicators. This ensures that the final cognitive performance value has cross-task and cross-individual consistency. By simultaneously quantifying the three dimensions of operational ability, memory ability and attention duration, the assessment of cognitive performance is expanded from a single indicator to a multi-angle and fine-grained quantitative description, which helps to accurately identify the type and severity of functional deficits. The training weights support both expert configuration and data-driven learning and online adaptation. They can dynamically optimize the assessment model according to different individual characteristics and efficacy feedback, improving the targeted accuracy and long-term efficacy.

[0027] In a preferred embodiment, the step of obtaining the intervention index based on multimodal data and cognitive performance information includes: S301. Obtain the corresponding cognitive state value based on multimodal data; S302. Obtain the corresponding cognitive performance value based on cognitive performance information; S303. Obtain cognitive weights based on cognitive training tasks, wherein the cognitive weights include cognitive state weights and cognitive performance weights. S304. Obtain the intervention index based on cognitive state value, cognitive performance value and cognitive weight.

[0028] As in steps S301 to S304 above Multimodal features, including behavioral, physiological, and environmental data, were preprocessed (denoising, time alignment, and imputation of missing values). Cognitive state values ​​were extracted, and user multimodal data and clinical standard scale scores (including MMSE, MoCA, and the Digit Span Test (DST)) were collected. The consistency between the two was verified using correlation analysis methods (such as Pearson correlation coefficient and Spearman rank correlation coefficient). The verification results showed a significant positive correlation between the calculated cognitive state values ​​and the scores of the aforementioned scales, accurately reflecting the user's true cognitive function level. Cognitive performance values ​​were extracted based on the cognitive performance information obtained from training, and cognitive weights were defined. The cognitive weight vector can be further subdivided into cognitive state weights and cognitive performance weights. Weights can be obtained through expert rules (initial configuration based on clinical and rehabilitation experience), data-driven learning (training using historical efficacy labels, such as regression, ranking, neural networks to obtain optimal weights, or attention mechanisms), or online adaptive methods (dynamically adjusting weights based on real-time feedback using Bayesian updates or reinforcement learning to adapt to individual differences and situational changes). An intervention index was calculated based on the cognitive state values, cognitive performance values, and cognitive weights. The formula for calculating the intervention index is as follows: In the formula, G represents the intervention index, r represents the cognitive state weight, R represents the cognitive state value, z represents the cognitive performance weight, and Z represents the cognitive performance value, r+z=1. To ensure the clinical effectiveness of the intervention index, a clinical validation step was introduced during the indicator system design stage. The cognitive state value was correlated with internationally recognized cognitive function assessment scales (such as MMSE, MoCA, etc.) to establish a statistically significant correspondence, so that the output cognitive state results can be consistent with traditional clinical assessment results. In addition, the intervention index is divided into several standardized intervals, each interval corresponding to a clear clinical meaning, such as "mild intervention required", "moderate intervention preferred", "severe cognitive risk", etc., so that the intervention index not only has algorithmic computability, but also clinical interpretability and application guidance value. Life situations and actual training performance are included in the decision-making simultaneously to avoid misjudgment caused by a single source. For example, short-term performance decline caused by environmental interference will not be automatically judged as requiring severe intervention.

[0029] In a preferred embodiment, the steps of obtaining intervention measures based on intervention indices, implementing intervention measures, and re-obtaining post-intervention cognitive performance information include: S401. Obtain the intervention table, which includes multiple intervention index intervals and the corresponding intervention measures for each intervention index interval; S402. Obtain the corresponding intervention measures from the intervention table based on the intervention index interval corresponding to the intervention index; S403. Implement intervention measures for the target users; S404, Feedback period based on intervention measures; S405. Obtain cognitive performance information after intervention during the feedback period.

[0030] As described in steps S401 to S405 above, the intervention table is a discrete mapping library, and its structure can be represented by several entries. The intervention index, after standardization, ranges from 0 to 100, and is used to represent the degree of cognitive intervention needs. A mapping relationship between the intervention index and intervention measures is established through statistical fitting. For example, 0 ≤ intervention index < 40 corresponds to severe cognitive impairment, belonging to the "severe need for intervention" state, which can be addressed by professional intervention (appointment assessment, reporting to medical staff) or emergency alarm (triggering hotline or on-site intervention); 40 ≤ intervention index < 70 corresponds to mild to moderate cognitive decline, belonging to the "moderate intervention priority" state, which can be addressed by remote guidance (video, voice guidance, remote supervision); 70 ≤ intervention index ≤ 100 corresponds to mild cognitive decline. The patient's cognitive state is normal or at a low level, which falls under the "mild intervention" category. Behavioral guidance (short-term cognitive exercises, attention training) can be provided. The above interval divisions are based on the normalized mapping of standard clinical score segments of MMSE and MoCA, thus giving the intervention index clear clinical interpretability. Furthermore, each state can be further subdivided into Level 1, Level 2, Level 3, etc., according to actual needs. The intervention table supports two sources: expert presets (clinical rules) and data-driven approaches (strategies selected based on historical efficacy). It records the version, effective date, and applicable population tags (age group, cognitive stage, comorbidities, etc.). The real-time calculated intervention index is mapped to the corresponding interval in the intervention table, and a candidate measure set is retrieved. An intervention is generated when the measure is issued. The pre-task sheet includes the type of measure, implementation timeline, implementing equipment, responsible party, expected duration, feedback sampling requirements, and failure fallback strategy. It can be implemented via terminal push notifications (mobile phone / tablet prompts and training initiation), wearable device vibration, lighting activation, smart home execution (automatic adjustment of lighting / volume), or remote intervention by call centers or medical personnel. During implementation, status monitoring is performed to confirm arrival and initiation (receipts), record user interactions (execution status, completion rate), and reasons for anomalies and rejections. For measures that cannot be implemented immediately (user offline, device not connected), a retry or downgraded alternative strategy is triggered. Feedback periods are obtained from pre-set procedures based on the intervention measures, and within these periods, feedback is collected according to sampling rules. Knowing performance information (training performance indicators, operation logs, attention indicators, etc.) and comparing it with the baseline before intervention, key steps include: data cleaning, normalization to the same scale, short-term trend detection and statistical testing (such as significance tests or threshold comparisons). The intervention table and automatic mapping mechanism connect assessment and intervention into a repeatable and auditable process, reducing the burden of manual judgment and accelerating response time. Through clear feedback periods and standardized sampling, short-term and medium-term effects can be judged more reliably, avoiding misjudgments caused by inconsistent sampling. Sampling and re-evaluation within the feedback period reduces false alarms and over-intervention caused by instantaneous fluctuations. At the same time, the contextual adaptation and implementation confirmation of the measures can improve user compliance.

[0031] In a preferred embodiment, the step of obtaining feedback time periods based on intervention measures includes: S4041. Obtain the time points for implementing intervention measures and mark them as the start time of the feedback period; S4042. Obtain the duration compensation table, wherein the duration compensation table includes multiple intervention measures and the duration coefficient corresponding to each intervention measure; S4043. Obtain the corresponding duration coefficient from the duration compensation table based on the intervention measures; S4044. Obtain the benchmark intervention index and benchmark duration; S4045. Obtain the feedback duration based on the intervention index, the baseline intervention index, the duration coefficient, and the baseline duration. S4046. Obtain the end time of the feedback period based on the feedback duration and the start time of the feedback period, and construct the feedback period.

[0032] As described in steps S4041 to S4046 above, when implementing an intervention, the implementation time node is recorded on the event bus. This timestamp also carries metadata (intervention ID, execution subject, such as automatic, remote, manual, retry, compensation flag, device context, and environment label). This implementation time node is directly used as the start time of the feedback period and written to the log for traceability and subsequent comparison. The duration compensation table is a configuration table or rule base, including multiple interventions and the duration coefficient corresponding to each intervention. The duration compensation table supports multi-channel expert experience configuration, clinical guidelines, and historical efficacy data statistics (recommended confidence interval). Using the intervention identifier as the key, matching entries are retrieved from the duration compensation table to obtain the duration coefficient. The preset baseline intervention index and baseline duration are obtained from the database (which can be set by expert experience configuration, clinical guidelines, and historical efficacy data statistics). The feedback duration is calculated based on the intervention index, baseline intervention index, duration coefficient, and baseline duration. The formula for calculating the feedback duration is as follows: In the formula, t represents the feedback duration, and G represents the intervention index. This is represented as the benchmark intervention index. Represented as duration coefficient, The baseline duration is represented by the start time plus the corresponding feedback duration to obtain the end time, and this is used to construct the feedback period. The clear start time and the feedback duration based on the intervention type ensure that the effect evaluation is carried out within an appropriate time window, which can capture both immediate effects and measure delayed and sustained effects, avoiding premature judgment or delayed response. The duration compensation table can adjust the duration coefficient according to user characteristics and risk level to realize population-based and individualized monitoring strategies, avoiding under-evaluation of vulnerable groups or over-monitoring of low-risk users.

[0033] In a preferred implementation, the steps of obtaining the cognitive intervention shift before and after the intervention and generating a multidimensional evaluation report include: S501. Obtain the corresponding pre-intervention cognitive performance value based on pre-intervention cognitive performance information; S502. Obtain the corresponding cognitive performance value after intervention based on the cognitive performance information after intervention; S503. Obtain the cognitive intervention deviation based on the cognitive performance values ​​before and after the intervention. S504. Obtain the scoring table, which includes multiple cognitive intervention deviation intervals and the intervention score corresponding to each cognitive intervention deviation interval. S505. Based on the cognitive intervention offset interval corresponding to the cognitive intervention offset, obtain the corresponding intervention score from the scoring table and generate a multidimensional evaluation report.

[0034] As described in steps S501 to S505 above, comparative indicators are extracted from pre-intervention cognitive performance information to obtain pre-intervention cognitive performance values. Post-intervention cognitive performance values ​​are extracted from cognitive performance information collected during the feedback period. Cognitive intervention deviation is calculated based on the pre-intervention and post-intervention cognitive performance values, which can be determined by dividing the post-intervention cognitive performance value by the pre-intervention cognitive performance value. To avoid assessment bias caused by relying solely on the "ratio of pre- and post-intervention performance," a baseline correction mechanism is introduced when calculating the cognitive intervention deviation. For users with low cognitive baseline, the deviation calculation is automatically adjusted to prevent low baseline values ​​from being exaggerated into high-increase results. For users with high cognitive baseline, the deviation threshold is dynamically adjusted to ensure that small but clinically significant improvements are not overlooked. This mechanism can accurately reflect the individual's cognitive improvement trend after intervention, ensuring that the intervention effect assessment is consistent with clinical reality. (For example, a low cognitive rating scale is introduced for users with low cognitive baseline, and a high cognitive rating scale is introduced for users with high cognitive baseline. The specific rating scale is matched according to the actual user.) The rating scale is a configurable rule base, including multiple cognitive intervention deviation intervals. The report includes intervention scores corresponding to each cognitive intervention deviation range, mapping the deviation to a scoring table to obtain intervention scores, and generating a multidimensional evaluation report. The evaluation report includes basic information (user ID, intervention ID, intervention type, start and end times, sample size, and data integrity indicators), numerical results, dimensional analysis (previous and subsequent values ​​and deviations of each sub-indicator, contribution rate of each sub-item to the overall deviation), evaluation conclusions (intervention scores based on the scoring table, whether the target was met, whether reassessment or escalation of intervention is needed), and visualizations (time series trend charts, before-and-after comparison bar charts, contribution breakdown pie charts or tables for clinical review). The system provides feedback, suggestions, and follow-up actions (recommendation to continue current measures, adjust or upgrade measures, conduct manual review, or terminate the intervention). It presents the intervention effect in easily understandable numerical values ​​using offset, allowing clinicians, caregivers, and users to intuitively see the changes brought about by the intervention. Configurable scoring tables and version management mechanisms support clinical experts in adjusting judgment rules to meet the personalized threshold needs of different populations. Simultaneously, it saves rule versions to meet audit and regulatory compliance requirements. Reports and scoring data can be used as training data to feed back into the model and strategy library to evaluate which interventions are more effective in which situations, driving continuous optimization of the intervention table and scoring thresholds.

[0035] In a preferred embodiment, after obtaining the cognitive intervention shift before and after the intervention and generating a multidimensional evaluation report, the method further includes: S506. Obtain the cognitive intervention deviation threshold; S507. Determine whether the cognitive intervention deviation is lower than the cognitive intervention deviation; If the cognitive intervention deviation is not less than the cognitive intervention deviation, the intervention is deemed effective. If the cognitive intervention deviation is lower than the cognitive intervention deviation, the intervention is deemed ineffective, an alarm is issued, the intervention is reacquired and implemented, until the intervention is effective.

[0036] As described in steps S506 to S507 above, the cognitive intervention deviation threshold is obtained. The threshold can be a single value or a stratified rule (configured separately by overall, sub-dimension, population, and risk level). The sources include clinical guidelines, expert experience, historical efficacy statistics (selecting significant improvement thresholds based on the effect distribution of previous intervention data), or automatic model recommendations. If the cognitive intervention deviation is not lower than the cognitive intervention deviation, it is considered effective; otherwise, it is considered ineffective. If effective, it continues to be implemented; if ineffective, an alarm message is issued, and intervention measures are re-obtained and implemented until the intervention measures are effective. The threshold judgment and automatic re-intervention form a closed loop, which can directly transform the evaluation results into subsequent actions, improving the response speed and closed-loop management capabilities of intervention.

[0037] To facilitate understanding of the implementation effect of the above-mentioned evaluation method applied to cognitive impairment intervention, the following describes the actual implementation process of this method using a specific application scenario. For example, for a 65-year-old user A with mild cognitive impairment, who wears a smart bracelet, uses a smart home system, and a mobile terminal application in a home environment, multimodal data is continuously collected. Behavioral data includes the number of daily activity switching, physiological data includes average heart rate, and environmental data includes the noise level of the living environment. After denoising and normalizing the above data, a behavioral value of 0.65, a physiological value of 0.72, and an environmental value of 0.58 are obtained. These values ​​are then weighted and fused according to multimodal weights (behavioral weight 0.4, physiological weight 0.4, environmental weight 0.2) to obtain a comprehensive cognitive state value of 0.67. According to the task table, the training task corresponding to this interval (0.6-0.7) is "moderate intensity memory association task + attention duration task". This training content is sent to user A, and the results are recorded. The user's cognitive training data was recorded, with an average reaction time of 1.8 seconds (operational feature information), an accuracy rate of 75% (memory matching information), and an average fixation duration of 6.5 seconds (attention duration information). After standardization, these data yielded operational feature values ​​of 0.62, memory matching values ​​of 0.75, and attention duration values ​​of 0.68. These values, combined with training weights (operational feature weight 0.3, memory matching weight 0.5, attention duration weight 0.2), resulted in a cognitive performance value of 0.69. A weighted fusion with cognitive state weight of 0.4 and cognitive performance weight of 0.6 yielded an intervention index of 0.68. The score fell within the "mild to moderate intervention priority" range (0.4–0.7) in the intervention table. The automatically recommended behavioral intervention was "add one memory training task daily and adjust based on sleep monitoring." The intervention was implemented, and post-training performance data was re-acquired during the 24-hour feedback period. Reaction time was reduced to 1.5 seconds, accuracy improved to 82%, and fixation duration increased to 7.1 seconds. A new cognitive performance score of 0.76 was calculated. Comparing the differences before and after the intervention, the cognitive intervention deviation was approximately 1.10 (a 10% performance improvement). A multidimensional evaluation report was generated, including the pre-intervention cognitive state score of 0.67 and the intervention... The post-cognitive performance score was 0.76, within the intervention index range, with mild to moderate intervention prioritized. The deviation was 1.10. The clinical recommendation was to maintain the existing training program and to reassess after one week. This example demonstrates that the system can automatically collect multi-source data from users in natural scenarios, quantify their cognitive state and training performance, and correlate intelligent assessment with clinical significance through intervention indices and intervention tables. It automatically generates appropriate intervention programs and verifies efficacy through a feedback phase. Compared with traditional manual questionnaires or single-scale assessments, this method achieves continuous monitoring, personalized intervention, and data-driven closed-loop optimization, significantly improving the scientific rigor and effectiveness of cognitive impairment intervention.

[0038] Please see the appendix Figure 2As shown, the present invention also provides an evaluation system for cognitive impairment intervention, used in the aforementioned evaluation method for cognitive impairment intervention, comprising: The cognitive training module is used to acquire multimodal data of target users in natural life scenarios, and to acquire cognitive training tasks based on the multimodal data; The cognitive performance module acquires the cognitive training data of the target user based on the cognitive training task, and obtains cognitive performance information based on the cognitive training data; The intervention numerical module obtains intervention indices based on multimodal data and cognitive performance information; The cognitive intervention module obtains intervention measures based on the intervention index, implements the intervention measures, and re-acquires cognitive performance information after the intervention. The intervention feedback module is used to obtain the cognitive intervention deviation before and after the intervention and generate a multidimensional evaluation report.

[0039] The cognitive training module is responsible for collecting multimodal raw data (behavioral, physiological, and environmental) in natural life scenarios through wearable devices, mobile phone sensors, and home IoT, outputting behavioral, physiological, and environmental values, and calling the task table service to select and generate the currently suitable cognitive training task, which is then sent to the terminal for execution and to initiate training data collection. The cognitive performance module collects interaction logs, response records, and attention-related signals while the user is performing training tasks, extracting operational feature values, memory matching values, and attention duration values, and extracting cognitive performance information. The intervention value module receives multimodal cognitive state values ​​and the cognitive performance values ​​obtained from training, calculates and outputs an intervention index based on configurable cognitive weights (state weights, performance weights), and the cognitive intervention module searches the intervention table for candidate intervention measures based on the intervention index, selects the final measure according to priority, feasibility (device, user status), and safety strategies, generates an intervention task list, and sends it out for execution in real time through appropriate channels (mobile push, smart home control, remote call), records the implementation time as the start point of the feedback period, and... The duration compensation table calculates the feedback period. Some interventions require manual confirmation or pose a high risk, triggering a manual review workflow. The intervention feedback module re-collects cognitive performance information within the feedback period, calculates the deviation before and after intervention, maps the deviation to an intervention score based on the scoring table, and generates a multi-dimensional evaluation report (overall score, sub-dimension score, time series diagram, contribution breakdown, judgment conclusion and suggestions). If the target is not met, an alarm is triggered and a re-intervention request is sent back to the cognitive intervention module, forming a closed-loop iteration. From life scenario data to task issuance and intervention evaluation, an automatic closed loop is formed, enabling rapid problem identification and timely intervention, improving response speed and timeliness. Based on multimodal and individual baseline assessment, the training tasks and intervention measures are highly matched with the user's current state, improving hit rate and compliance. By integrating behavior, physiology, environment, and training performance, misjudgments caused by single signal noise or occasional interference are reduced, resulting in more stable and reliable results. Intervention effect data is fed back for model training and strategy updates. With data accumulation, the intervention table and scoring thresholds can gradually evolve from experience-driven to data-driven, thereby continuously improving the overall efficacy.

[0040] And, an assessment terminal for cognitive impairment intervention, comprising: One or more processors; A storage device on which one or more programs are stored; When one or more programs are executed by one or more processors, the one or more processors implement evaluation methods applicable to cognitive impairment interventions.

[0041] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. An evaluation method applied to cognitive impairment intervention, characterized in that, include: Acquire multimodal data of target users in natural life scenarios, and acquire cognitive training tasks based on the multimodal data; Cognitive training data of target users is obtained based on cognitive training tasks, and cognitive performance information is obtained based on cognitive training data; Intervention indices are obtained based on multimodal data and cognitive performance information; Intervention measures are obtained based on the intervention index, the intervention measures are implemented, and cognitive performance information after the intervention is re-obtained; Obtain the cognitive shift before and after the intervention and generate a multidimensional evaluation report.

2. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps for acquiring multimodal data of target users in natural life scenarios and obtaining cognitive training tasks based on this multimodal data include: Acquire multimodal data of target users in natural life scenarios, including behavioral data, physiological data, and environmental data; Behavioral, physiological, and environmental values ​​are obtained based on multimodal data; Obtain multimodal weights, which include behavioral weights, physiological weights, and environmental weights; Cognitive state values ​​are obtained based on behavioral values, physiological values, environmental values, and multimodal weights. Obtain the task table, which includes multiple cognitive state value intervals and the cognitive training task corresponding to each cognitive state value interval; The corresponding cognitive training task is obtained from the task table based on the cognitive state value range corresponding to the cognitive state value.

3. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps of acquiring cognitive training data of target users based on cognitive training tasks, and acquiring cognitive performance information based on cognitive training data, include: Perform cognitive training tasks and acquire cognitive training data, which includes operational feature information, memory matching information, and attention duration information. Based on cognitive training data, obtain corresponding operational feature values, memory matching values, and attention duration values; Obtain the training weights, which include operational feature weights, memory matching weights, and attention persistence weights. Cognitive performance values ​​are obtained based on operational feature values, memory matching values, attention duration values, and cognitive training weights, and are then labeled as cognitive performance information.

4. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps for obtaining intervention indices based on multimodal data and cognitive performance information include: Obtain the corresponding cognitive state value based on multimodal data; Obtain the corresponding cognitive performance value based on cognitive performance information; Cognitive weights are obtained based on cognitive training tasks, where cognitive weights include cognitive state weights and cognitive performance weights. An intervention index is obtained based on cognitive state value, cognitive performance value, and cognitive weight.

5. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps of obtaining intervention measures based on intervention indices, implementing intervention measures, and re-obtaining post-intervention cognitive performance information include: Obtain the intervention table, which includes multiple intervention index ranges and the corresponding intervention measures for each intervention index range; Based on the intervention index range corresponding to the intervention index, obtain the corresponding intervention measures from the intervention table; Implement intervention measures for target users; Feedback period based on intervention measures; Obtain cognitive performance information after intervention during the feedback period.

6. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps for obtaining feedback based on intervention measures include: Obtain the time points for implementing intervention measures and mark them as the start time of the feedback period; Obtain the duration compensation table, which includes multiple intervention measures and the duration coefficient corresponding to each intervention measure; Obtain the corresponding duration coefficient from the duration compensation table based on the intervention measures; Obtain the benchmark intervention index and benchmark duration; The feedback duration is obtained based on the intervention index, the baseline intervention index, the duration coefficient, and the baseline duration. The end time of the feedback period is obtained based on the feedback duration and the start time of the feedback period, and the feedback period is constructed.

7. The evaluation method for cognitive impairment intervention according to claim 1, characterized in that, The steps for obtaining the cognitive shift before and after the intervention and generating a multidimensional evaluation report include: Obtain the corresponding pre-intervention cognitive performance value based on pre-intervention cognitive performance information; Obtain the corresponding cognitive performance value after intervention based on the cognitive performance information after intervention; Cognitive intervention offset is obtained based on cognitive performance values ​​before and after intervention. Obtain the scoring table, which includes multiple cognitive intervention offset intervals and the intervention score corresponding to each cognitive intervention offset interval; Based on the cognitive intervention offset interval corresponding to the cognitive intervention offset, the corresponding intervention score is obtained from the scoring table, and a multidimensional evaluation report is generated.

8. The evaluation method for cognitive impairment intervention according to claim 7, characterized in that, After obtaining the cognitive shift before and after the intervention and generating a multidimensional evaluation report, the following steps are also included: Obtain the cognitive intervention offset threshold; Determine whether the cognitive intervention offset is lower than the cognitive intervention offset; If the cognitive intervention deviation is not less than the cognitive intervention deviation, the intervention is deemed effective. If the cognitive intervention deviation is lower than the cognitive intervention deviation, the intervention is deemed ineffective, an alarm is issued, the intervention is reacquired and implemented, until the intervention is effective.

9. An evaluation system for cognitive impairment intervention, applied to the evaluation method for cognitive impairment intervention as described in any one of claims 1 to 8, characterized in that, include: The cognitive training module is used to acquire multimodal data of target users in natural life scenarios, and to acquire cognitive training tasks based on the multimodal data; The cognitive performance module acquires the cognitive training data of the target user based on the cognitive training task, and obtains cognitive performance information based on the cognitive training data; The intervention numerical module obtains intervention indices based on multimodal data and cognitive performance information; The cognitive intervention module obtains intervention measures based on the intervention index, implements the intervention measures, and re-acquires cognitive performance information after the intervention. The intervention feedback module is used to obtain the cognitive intervention deviation before and after the intervention and generate a multidimensional evaluation report.

10. An assessment terminal for cognitive impairment intervention, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When one or more programs are executed by one or more processors, the one or more processors implement the evaluation method for cognitive impairment intervention as described in any one of claims 1 to 8.

Images