Method and system for constructing refined model of stroke dysfunction based on icf theory

Abstract

The present application belongs to the technical field of medical health informatics, and specifically relates to a fine model construction method and system for stroke dysfunction based on ICF theory, aiming to construct a fine model capable of accurately describing the dysfunction area of stroke patients through dynamic optimization and adaptive adjustment mechanism, so as to improve the accuracy and efficiency of dysfunction evaluation. Through the dynamic scoring mechanism, the multi-modal evaluation data of the patient, including pathological information, imaging data and biochemical indicators, are continuously collected and analyzed. Secondly, combined with the ICF theory, the feature parameters of the dysfunction area are automatically extracted and optimized through feature matching and deep learning algorithm, and a three-dimensional model capable of accurately evaluating and positioning the dysfunction area is constructed. The model not only can update and adaptively adjust the parameters in real time, but also can provide personalized rehabilitation recommendations and treatment plans according to the individual differences of different patients.

Description

Technical Field

[0001] This invention belongs to the technical field of healthcare informatics, specifically relating to a method and system for constructing a refined model of stroke functional impairment based on ICF theory. Background Technology

[0002] Stroke is a complex disease caused by the interaction of multiple pathophysiological factors, also known as apoplexy. In severe cases, it can lead to death, and the number of deaths from stroke increases as the course of the disease lengthens. Therefore, timely detection of functional impairment areas in stroke patients and appropriate physical therapy for these areas can greatly reduce the patient's chance of death.

[0003] Among existing technologies, the ICF theory (International Classification of Functioning, Disability and Health) is the most suitable functional classification system for modern rehabilitation medicine models. By adopting internationally unified standardized language, it provides terms, definitions, and classifications related to functional health and disability. Currently, the ICF theory has been widely used in many fields such as health monitoring, functional assessment, treatment planning, health care, disability welfare allocation, vocational assessment, social policy making, universal design, education programs, income insurance, and intervention management. However, in the application of the existing ICF theory, it is impossible to provide a detailed description of stroke functional impairment, which can easily lead to inaccurate descriptions of the area and location of functional impairment. This can result in improper treatment of the patient's functional impairment area, increasing the patient's mortality rate or prolonging the recovery period. Therefore, a detailed model of stroke functional impairment area based on the ICF theory has emerged. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing a refined model of functional impairment in stroke patients based on ICF theory. The aim is to construct a refined model that can accurately describe the functional impairment areas of stroke patients through dynamic optimization and adaptive adjustment mechanisms, so as to improve the accuracy and efficiency of functional impairment assessment.

[0005] The specific technical solution adopted by this invention is as follows:

[0006] A method for constructing a refined model of stroke functional impairment based on ICF theory includes:

[0007] Acquire multimodal assessment data of stroke patients and preprocess the data, which includes patient medical records, functional assessment reports and feedback from medical staff.

[0008] Cluster analysis is performed on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information.

[0009] Feature extraction is performed based on functional impairment information, the extracted features are output, and they are summarized into a feature label set;

[0010] Each feature label in the feature label set is encoded to obtain baseline encoded data;

[0011] A mapping model is constructed based on the correspondence between the baseline encoded data and the extracted features. This model is then used to map new input data, and the model is dynamically optimized based on its application performance to continuously improve its mapping accuracy. Specifically, the steps for dynamic optimization include:

[0012] Obtain the current application effect of the mapping model, and determine the current optimization range based on this effect;

[0013] Real-time monitoring of user feedback during the application of the mapping model, and analysis and processing of the feedback to obtain optimization points;

[0014] An optimization plan is constructed based on the optimization points and the current optimization magnitude, and the mapping model is optimized according to the optimization plan.

[0015] The dynamic optimization is graded according to the application effect of the optimization plan, and the level of dynamic optimization is bound to the optimization parameters, and each optimization parameter is stored synchronously.

[0016] In a preferred embodiment, the step of acquiring and preprocessing multimodal assessment data of stroke patients includes:

[0017] Collect multimodal assessment data from stroke patients and preprocess them;

[0018] The multimodal evaluation data is standardized to unify its format;

[0019] Extract the feature information from the multimodal evaluation data, and perform deduplication and noise reduction on the feature information;

[0020] The processed feature information is paired to obtain pairing parameters, and each pairing parameter is classified into baseline input conditions and post-input conditions.

[0021] The baseline input conditions and post-input conditions are processed synchronously and output as a baseline condition parameter set.

[0022] In a preferred embodiment, when the feature information of the multimodal evaluation data is deduplicated and denoised, a noise filtering model is constructed, and the extracted feature information is input into the noise filtering model to identify invalid or erroneous data in the feature information. An anomaly monitoring module is added to the noise filtering model to monitor the feature information extraction process in real time, and the anomaly monitoring module outputs monitoring data to determine whether the feature information extraction process is abnormal.

[0023] In a preferred embodiment, before the step of synchronously processing the baseline input conditions and the subsequent input conditions and outputting them as a baseline condition parameter set, the disease course length is calculated based on the feature information, and the combination method of the baseline input conditions and the subsequent input conditions is determined based on the disease course length.

[0024] In a preferred embodiment, after the baseline condition parameter set is output, condition parameter matching is performed based on the baseline input conditions and post-input conditions in the baseline condition parameter set. Then, parameters that match the condition parameters are selected according to the disease duration and output as preprocessing results.

[0025] In a preferred embodiment, after the preprocessing result is output, it is used as input to a condition parameter correction model for correction. In the condition parameter correction model, a condition parameter correction database is constructed. The condition parameter correction database is configured with a condition pre-set subset, a condition post-set subset, and a condition parallel subset, wherein multiple condition parameters are preset in each of the condition pre-set subset, condition post-set subset, and condition parallel subset.

[0026] In a preferred embodiment, the step of extracting features based on functional impairment information, outputting the extracted features, and summarizing them into a feature label set includes:

[0027] Obtain multimodal assessment data corresponding to the functional impairment information;

[0028] Feature extraction is performed on the multimodal evaluation data to obtain multiple damage features;

[0029] The damage features are scored, and damage features with similar scores are merged and summarized into extracted features. The extracted features are then labeled and a feature label set is output.

[0030] By constructing a dynamic scoring mechanism, damage features are continuously collected, and the feature scores of the scoring mechanism are continuously updated to improve the accuracy of feature extraction.

[0031] In a preferred embodiment, after the dynamic optimization step, the application effect of the optimization plan is evaluated and analyzed to obtain an application score.

[0032] The optimization plans are ranked based on the application scores, and the historical optimization schemes of the top-K optimization plans are combined and applied to generate the optimal optimization plan.

[0033] The mapping model is optimized according to the optimal optimization plan;

[0034] The optimization effect of the optimal optimization plan on the mapping model is monitored synchronously.

[0035] In a preferred embodiment, after optimizing the mapping model according to the optimal optimization plan, a supervised learning mechanism is introduced to increase human intervention measures, adjust the output of the mapping model to better meet real-world needs, collect user feedback in real time, extract optimization suggestions, and adjust the mapping model parameters based on the optimization suggestions.

[0036] This invention also provides a system for constructing a refined model of stroke functional impairment based on ICF theory, applied to the aforementioned method for constructing a refined model of stroke functional impairment based on ICF theory, comprising:

[0037] The data acquisition module is used to acquire multimodal assessment data of stroke patients and perform preprocessing. The multimodal assessment data includes patient medical record information, functional assessment reports, and feedback information from medical staff.

[0038] The data analysis module is used to perform cluster analysis on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information.

[0039] The feature extraction module is used to extract features based on functional impairment information, output the extracted features, and summarize them into a feature label set.

[0040] An encoding module is used to encode each feature label in the feature label set to obtain baseline encoded data;

[0041] The model building module is used to build a mapping model based on the correspondence between the baseline encoded data and the extracted features, and to map new input data through the mapping model. The model is also dynamically optimized based on its application effect to continuously improve its mapping accuracy.

[0042] Beneficial Effects: This invention constructs a refined model that accurately describes the functional impairment areas of stroke patients through dynamic optimization and adaptive adjustment mechanisms. Specifically, by introducing a dynamic scoring mechanism, it continuously collects and analyzes patients' multimodal assessment data, including pathological information, imaging data, and biochemical indicators. Secondly, combining ICF theory, it automatically extracts and optimizes the feature parameters of the impairment areas through feature matching and deep learning algorithms, constructing a three-dimensional model that can accurately assess and locate the impairment areas. This model can not only update and adaptively adjust parameters in real time, but also provide personalized rehabilitation suggestions and treatment plans based on individual differences among patients. Finally, by introducing a feedback closed-loop mechanism, it continuously collects and analyzes the application effects, continuously optimizing and adjusting the model to improve the accuracy and efficiency of assessment, resulting in higher accuracy and faster efficiency in functional impairment assessment. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0044] Figure 2 This is a schematic diagram of a sub-step of step S1 of the present invention;

[0045] Figure 3 This is a schematic diagram of a sub-step of step S3 in the present invention;

[0046] Figure 4 This is a schematic diagram of a sub-step of step S6 of the present invention;

[0047] Figure 5 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0048] To more clearly illustrate the technical solutions in this invention or the prior art, the technical solutions in this invention or the prior art will be described below in conjunction with the accompanying drawings and specific embodiments.

[0049] Example 1

[0050] Please see the appendix Figure 1 , attached Figure 1 A method flowchart based on an embodiment of the present invention is provided, which includes:

[0051] S1. Acquire multimodal assessment data of stroke patients and preprocess the data. The multimodal assessment data includes patient medical records, functional assessment reports, and feedback information from medical staff.

[0052] S2. Perform cluster analysis on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information.

[0053] S3. Based on the functional impairment information, extract features, output the extracted features, and summarize them into a feature label set;

[0054] S4. Encode each feature label in the feature label set to obtain baseline encoded data;

[0055] S5. Construct a mapping model based on the correspondence between the baseline encoded data and the extracted features, and map the new input data through the mapping model. Optimize the mapping model dynamically based on its application effect to continuously improve its mapping accuracy.

[0056] Specifically, the steps corresponding to the dynamic optimization include:

[0057] Obtain the current application effect of the mapping model, and determine the current optimization range based on this effect;

[0058] Real-time monitoring of user feedback during the application of the mapping model, and analysis and processing of the feedback to obtain optimization points;

[0059] An optimization plan is constructed based on the optimization points and the current optimization magnitude, and the mapping model is optimized according to the optimization plan.

[0060] The dynamic optimization is graded according to the application effect of the optimization plan, and the level of dynamic optimization is bound to the optimization parameters, and each optimization parameter is stored synchronously.

[0061] As described in steps S1-S5, in this embodiment, the method for constructing a refined model of stroke functional impairment based on ICF theory first acquires multimodal assessment data of stroke patients, including patient medical records, functional assessment reports, and feedback from medical staff. Then, these data are standardized and deduplicated to ensure accuracy and consistency. Next, cluster analysis is performed on the preprocessed data to classify functional impairment information and non-functional impairment information. This classification result helps to better understand the patient's functional impairment status. Then, feature extraction is performed based on the functional impairment information to output a feature label set. To better process these feature labels, they are encoded to obtain baseline encoded data. This encoded data can serve as the basis for subsequent model construction. In the model construction stage, a mapping model needs to be constructed based on the correspondence between the baseline encoded data and the extracted features. Through this mapping model, new input data can be mapped to form a refined model that clearly represents the patient's functional impairment, helping doctors make corresponding diagnoses more quickly and avoiding regional identification errors caused by individual ability differences.

[0062] Ultimately, this embodiment improves the mapping accuracy of its optimized mapping model by introducing a dynamic optimization mechanism. Specifically, by monitoring user feedback in real time and analyzing user behavior and their evaluation of the mapping model, optimization points can be collected. Based on this, this embodiment constructs corresponding optimization plans to guide the improvement process of the mapping model. An important point of this embodiment is that dynamic optimization is graded according to the application effect of the optimization plan, and the corresponding dynamic optimization parameters are bound to the mapping model to coordinate the relationship between various dynamic optimizations.

[0063] Specifically, such as Figure 2 As shown, the steps for acquiring and preprocessing multimodal assessment data from stroke patients include:

[0064] S11. Collect multimodal assessment data of stroke patients and preprocess them;

[0065] S12. Standardize the multimodal evaluation data to unify its format;

[0066] S13. Extract the feature information of the multimodal evaluation data, and perform deduplication and noise reduction processing on the feature information;

[0067] S14. Perform pairing processing on the processed feature information to obtain pairing parameters, and classify each pairing parameter into baseline input conditions and post-input conditions.

[0068] S15. Synchronize the reference input conditions and post-input conditions, and output them as a reference condition parameter set.

[0069] To gain a more comprehensive understanding of the health status of stroke patients, it is necessary to first collect and preprocess relevant data. This step begins by collecting a large amount of multimodal assessment data, including basic information, medical history, physical examination reports, and various medical imaging data of stroke patients, to ensure the comprehensiveness and accuracy of the data.

[0070] Then, the multimodal evaluation data undergoes careful and systematic preprocessing to eliminate any potential data noise, redundancy, or anomalies, ensuring data purity and consistency. Next, the multimodal evaluation data is standardized to unify its format, facilitating better subsequent analysis. In this process, all data is converted to a standard format; for example, ensuring that all values ​​use consistent units, time formats, and naming conventions, preventing calculation errors due to formatting issues when processing in different environments.

[0071] Next, feature extraction is performed to extract important feature information, such as extracting different types of data and information of different properties to further simplify the data used. In addition, the feature information of the multimodal evaluation data needs to be deduplicated and denoised to ensure the uniqueness and purity of the data. Then, pairing processing is performed to obtain paired parameters, and each paired parameter is classified into baseline input conditions and post-input conditions to help clarify the relationship between data, such as determining the dependencies and constraints between different parameters. Finally, the baseline input conditions and post-input conditions are processed synchronously and summarized into a baseline condition parameter set. With the help of modern information technology, the above steps are completed automatically, forming a complete and efficient data processing system.

[0072] Specifically, in the steps of pairing parameters and classifying each pairing parameter into baseline input conditions and post-input conditions, when the feature information of the multimodal evaluation data is deduplicated and denoised, a noise filtering model is constructed, and the extracted feature information is input into the noise filtering model to identify invalid or erroneous data in the feature information. An anomaly monitoring module is added to the noise filtering model to monitor the feature information extraction process in real time. The monitoring data output by the anomaly monitoring module is used to determine whether the feature information extraction process is abnormal. During the implementation of the noise filtering model, a corresponding data filtering module is set up to help identify invalid or erroneous data, thereby avoiding the appearance of messy data and preventing misleading subsequent analysis. At the same time, an anomaly monitoring module is added to help monitor the feature information extraction process in real time, ensuring the accuracy of the extracted feature information. Furthermore, because the sources of the multimodal evaluation data used in the feature extraction process are diverse, the introduction of the anomaly monitoring module in this step is also to identify in real time whether there are any problems with the extracted data.

[0073] Further:

[0074] After synchronously processing the baseline input conditions and subsequent input conditions and outputting them as a baseline condition parameter set, the disease course length is calculated based on the feature information, and the combination method of the baseline input conditions and subsequent input conditions is determined based on the disease course length.

[0075] During the synchronous processing of baseline input conditions and subsequent input conditions, the disease course length is calculated based on the feature information, which can determine the combination method of baseline input conditions and subsequent input conditions. By taking into account the differences in the manifestation of the lesion area under different disease courses, a connection is established between the baseline input conditions and subsequent input conditions, ensuring the output order between the baseline input conditions and subsequent input conditions.

[0076] Specifically, after the baseline condition parameter set is output, condition parameter matching is performed based on the baseline input conditions and post-input conditions in the baseline condition parameter set. Then, parameters that match the condition parameters are selected according to the disease duration and output as preprocessing results.

[0077] Specifically, by importing corresponding baseline and post-input conditions, and based on the preset disease duration, the baseline and post-input conditions are filtered and matched to output key areas that have a significant impact on the lesion area, as well as clearly display the cerebrovascular circulation system of stroke patients, and locate the lesion area to help doctors respond more quickly and make the correct treatment.

[0078] Specifically, after the preprocessing result is output, it is used as input to connect to a conditional parameter correction model for correction. In this model, a conditional parameter correction database is constructed, containing conditional pre-set subsets, conditional post-set subsets, and conditional parallel subsets. Each of these subsets has multiple pre-defined conditional parameters. In this step, the preprocessing result is used as input to further adjust and optimize the preprocessing result by connecting to a variety of conditional parameter correction models. In the individual conditional parameter correction model, a conditional parameter correction database is first created. Then, based on the execution order of the conditional parameters, conditional pre-set subsets, conditional post-set subsets, and conditional parallel subsets are created. The conditional parameters are then called according to the execution order, and the corresponding post-conditional parameters are output, providing support for further optimization of the preprocessing result.

[0079] Specifically, such as Figure 3 As shown, the steps for feature extraction based on functional impairment information, outputting the extracted features, and summarizing them into a feature label set include:

[0080] S31. Obtain multimodal assessment data corresponding to the functional impairment information;

[0081] S32. Perform feature extraction on the multimodal evaluation data to obtain multiple damage features;

[0082] S33. The damage features are scored, and the damage features with similar scores are merged and summarized into extracted features. The extracted features are then labeled and output as a feature label set.

[0083] By constructing a dynamic scoring mechanism, damage features are continuously collected, and the feature scores of the scoring mechanism are continuously updated to improve the accuracy of feature extraction.

[0084] As described in steps S31-S33, when extracting features based on functional impairment information, firstly, the multimodal assessment data corresponding to the functional impairment information is acquired by the system, and features are extracted through deep learning and image processing technology. After being output, the data is matched with the impairment features to help better understand the nature and location of the impairment features. Then, further assessment and analysis of the functional impairment are carried out. Based on this, the degree of functional impairment damage to the patient caused by stroke is assessed, and doctors are given clearer treatment opinions to ensure that the patient receives timely and appropriate physical therapy.

[0085] Based on the matching relationship between the multimodal evaluation data and the damage features, the matched multimodal evaluation data will be scored and a damage feature scoring table will be output to arrange each damage feature in order of priority. Low-priority damage features will be merged into extracted features, and the extracted features will be marked as extraction labels and stored in the feature label set. The priority of the damage features will be further mapped to the priority of the extraction labels, thereby forming a clear, detailed and logically strong feature label set.

[0086] By constructing a dynamic scoring mechanism, damage features are continuously collected and converted into feature scores to update the damage feature scoring table, thereby improving the accuracy and efficiency of feature extraction. This enables doctors to quickly and accurately obtain the optimal extracted features, respond to patients' conditions in a timely manner, and provide appropriate physical therapy to patients when necessary, thus shortening the response cycle.

[0087] In the method for constructing a refined model of stroke functional impairment based on ICF theory, the average extraction accuracy of multimodal assessment data and median features of injury characteristics reached 34.07%. Among them, the extraction features with equal scores have the same priority. However, after the scoring requirements of each extraction feature are met, the priority of the current extraction feature can be further improved by ranking. Based on this, a dynamic scoring mechanism is introduced to continuously optimize the process, which increases the average extraction accuracy by nearly double to more than 6%. It can also merge multiple extraction features with low extraction accuracy into one extraction feature, increasing the comprehensiveness of stroke assessment. This allows for continuous collection and updating of the extraction label set, providing more accurate auxiliary solutions for the future, and improving the response speed and decision-making accuracy of doctors.

[0088] Specifically, such as Figure 4 As shown:

[0089] After dynamically optimizing the application effect of the mapping model, the application effect of the optimization plan is evaluated and analyzed to obtain an application score;

[0090] S61. Based on the application scores, the optimization plans are sorted, and the historical optimization schemes of the top-K optimization plans are combined and applied to generate the optimal optimization plan;

[0091] S62. Optimize the mapping model according to the optimal optimization plan;

[0092] S63. Simultaneously monitor the optimization effect of the optimal optimization plan on the mapping model.

[0093] As described in steps S61-S63, after dynamically optimizing the application effect of the mapping model, firstly, the application effect of each optimization plan is evaluated and analyzed, and summarized into a unified evaluation and analysis form. Then, the application scores in the optimization plan list are sorted, the top-K optimization plans with the best optimization effect are selected, and their historical optimization schemes are combined and applied to generate the optimal plan with high feasibility. Then, the mapping model is optimized according to the optimal plan, and the optimization effect of the optimal optimization plan on the mapping model is monitored synchronously in combination with multiple feedback methods, and further adjustments and optimizations are made as needed.

[0094] Specifically, in historical optimization schemes, the success rate and stability of each optimization scheme in historical applications are comprehensively considered. Those standardized low-score optimization schemes that have been fully validated are selected. Then, based on the priority information of doctors on lesion damage features, different standardized low-score ranges are given for damage features of different lesion types to generate the optimal optimization plan. This plan is then extracted into subsequent mapping schemes as needed to improve the efficiency of the mapping model optimization.

[0095] Specifically, after optimizing the mapping model according to the optimal optimization plan, a supervised learning mechanism is introduced to increase human intervention measures, adjust the output of the mapping model to better meet real-world needs, collect user feedback in real time, extract optimization suggestions, and adjust the mapping model parameters based on the optimization suggestions.

[0096] The mapping model continuously receives various constraint information and outputs different mapping result types, prohibiting the use of other types of mapping models without explicit authorization. However, the rapid expansion of mapping model categories can lead to abnormal matching of mapping result parameters between different types of mapping models, resulting in mapping result types that do not match the original drawings. Obviously, this deviates from the user's original intention, and the deviation is quite serious.

[0097] In this embodiment, positive and negative user feedback (such as smiley faces, sad faces, etc., and further detailed evaluation information should be applied later) under each mapping model is combined to curb the expansion of mapping model categories and avoid errors in mapping result types. Over a long period, user feedback is integrated into the dynamic scoring process, allowing it to participate in the feedback and optimization of the accuracy of damage features. Every 25 user feedbacks, the mapping model outputs TOP-K mapping results, which are further filtered by machine algorithms and human intervention, thus efficiently presenting them to users in a graphic and text format. Extensive industry practice has proven that the graphic and text display method not only improves user reading efficiency but also increases their satisfaction. Next, the output mapping results are further connected to the next stage to meet user needs. Conversely, the categories of mapping results are reduced to continuously improve user experience. While improving user satisfaction, this invention also provides researchers with a practical industry presentation tool, allowing research results to be presented in a more accessible and dynamic way, rather than being limited to dry textual descriptions in papers.

[0098] Example 2

[0099] like Figure 5 As shown, this embodiment provides a system for constructing a refined model of stroke functional impairment based on ICF theory, applied to the aforementioned method for constructing a refined model of stroke functional impairment based on ICF theory, including:

[0100] The data acquisition module is used to acquire multimodal assessment data of stroke patients and perform preprocessing. The multimodal assessment data includes patient medical record information, functional assessment reports, and feedback information from medical staff.

[0101] The data analysis module is used to perform cluster analysis on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information.

[0102] The feature extraction module is used to extract features based on functional impairment information, output the extracted features, and summarize them into a feature label set.

[0103] An encoding module is used to encode each feature label in the feature label set to obtain baseline encoded data;

[0104] The model building module is used to build a mapping model based on the correspondence between the baseline encoded data and the extracted features, and to map new input data through the mapping model. The model is also dynamically optimized based on its application effect to continuously improve its mapping accuracy.

[0105] In the aforementioned system, the data acquisition module is used to obtain multimodal assessment data of stroke patients from high-quality medical data sources, ensuring the comprehensiveness and accuracy of the data. By using advanced data acquisition tools and technologies, multimodal assessment data can be acquired from different medical facilities, equipment, and environments, covering various information such as the patient's pathology, genes, imaging, and biometrics. After data acquisition, the multimodal assessment data is preprocessed, including data cleaning, deduplication, missing value imputation, outlier handling, data transformation, and normalization, to ensure data quality and consistency. Based on specific needs and purposes, the data is mapped to an appropriate representation space or stored and transmitted in an appropriate form, increasing the efficiency of the data acquisition process and ensuring data quality and privacy security.

[0106] The data analysis module is used to identify data patterns and characteristics of functional impairment in different stroke patients, find similar functional impairment groups, sort out, summarize and compress the multidimensional and complex information in the multimodal assessment data, and classify the multimodal assessment data of stroke patients into different categories to generate classification information, including functional impairment information and non-functional impairment information. This helps to clarify the functional impairment characteristics and non-impairment characteristics of stroke patients. Combined with standardized data construction and label description matching, a visual and refined assessment map is constructed to provide intuitive assistance to doctors.

[0107] The feature extraction module is used to automatically extract key features from functional impairment information, such as extracting salient features of the patient's physical signs, symptoms, and imaging data, and outputting extracted features to improve the efficiency and accuracy of assessment and diagnosis. This enables doctors to quickly understand the patient's functional impairment, reduce time loss caused by describing features, and increase the patient's chances of receiving appropriate physical therapy.

[0108] The encoding module is used to encode each feature label in the feature label set output by the feature extraction module, converting them into digital signals that are easy to process, store, and transmit. During the encoding of feature labels, information loss is avoided, ensuring that the encoded feature labels can accurately represent the original multimodal data. These encoded feature labels are stored in the database, enabling them to be transmitted and shared between different systems and platforms, which can support subsequent data warehousing, data analysis, and data mining.

[0109] The model building module is used to construct a mapping model based on the correspondence between the baseline encoded data in the encoding module and the features extracted in the feature extraction module. After the mapping model is built, further mapping operations can be performed to help convert the patient's medical data or images into a format that is in their native language, expressing them more accurately. This allows nurses, patients, and their families to have a preliminary understanding of stroke patients without requiring doctors to provide excessive explanations, enabling doctors to focus more on confirming functional impairments.

[0110] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a refined assessment model of functional impairment regions in stroke patients based on ICF theory, characterized by: include: Acquire multimodal assessment data of stroke patients and perform preprocessing; The multimodal assessment data includes patient medical records, functional assessment reports, and feedback from medical staff. The preprocessing steps include: The multimodal evaluation data is standardized to unify its format; The feature information of the multimodal evaluation data is extracted and deduplication and noise reduction are performed. The processed feature information is paired to obtain pairing parameters, and each pairing parameter is classified into baseline input conditions and post-input conditions. The disease course length is calculated based on the feature information, and the combination method of the baseline input conditions and the post-input conditions is determined based on the disease course length. The baseline input conditions and the post-input conditions are processed synchronously and output as a baseline condition parameter set. The condition parameters are matched based on the baseline input conditions and the post-input conditions in the baseline condition parameter set. Then, the parameters that match the condition parameters are selected according to the disease course length, and the lesion area is output as the preprocessing result. After the preprocessing result is output, it is used as input to connect to the condition parameter correction model for correction. Cluster analysis is performed on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information. Feature extraction is performed based on functional impairment information, the extracted features are output, and they are summarized into a feature label set; Each feature label in the feature label set is encoded to obtain baseline encoded data; A mapping model is constructed based on the correspondence between the baseline encoded data and the extracted features. New input data is mapped through the mapping model, and the mapping model is dynamically optimized based on its application effect to continuously improve its mapping accuracy. The steps corresponding to the dynamic optimization include: Obtain the current application effect of the mapping model, and determine the current optimization range based on this effect; Real-time monitoring of user feedback during the application of the mapping model, and analysis and processing of the feedback to obtain optimization points; An optimization plan is constructed based on the optimization points and the current optimization magnitude, and the mapping model is optimized according to the optimization plan. The dynamic optimization is graded according to the application effect of the optimization plan, and the level of dynamic optimization is bound to the optimization parameters, and each optimization parameter is stored synchronously.

2. The method for constructing a refined assessment model of stroke functional impairment regions based on ICF theory according to claim 1, characterized in that: The deduplication and noise reduction process includes: constructing a noise filtering model and inputting the extracted feature information into the noise filtering model to identify invalid or erroneous data in the feature information; adding an anomaly monitoring module to the noise filtering model to monitor the feature information extraction process in real time; and combining the monitoring data output by the anomaly monitoring module to determine whether the feature information extraction process is abnormal.

3. The method for constructing a refined assessment model of stroke functional impairment regions based on ICF theory according to claim 2, characterized in that: The steps of extracting features based on functional impairment information, outputting the extracted features, and summarizing them into a feature label set include: Obtain multimodal assessment data corresponding to the functional impairment information; Feature extraction is performed on the multimodal evaluation data to obtain multiple damage features; The damage features are scored, and damage features with similar scores are merged and summarized into extracted features. The extracted features are then labeled and a feature label set is output. By constructing a dynamic scoring mechanism, damage features are continuously collected, and the feature scores of the scoring mechanism are continuously updated to improve the accuracy of feature extraction.

4. The method for constructing a refined assessment model of stroke functional impairment regions based on ICF theory according to claim 3, characterized in that: Following the dynamic optimization steps, the application effect of the optimization plan is evaluated and analyzed to obtain an application score; The optimization plans are ranked based on the application scores, and the historical optimization schemes of the top-K optimization plans are combined and applied to generate the optimal optimization plan. The mapping model is optimized according to the optimal optimization plan; The optimization effect of the optimal optimization plan on the mapping model is monitored synchronously.

5. The method for constructing a refined assessment model of stroke functional impairment regions based on ICF theory according to claim 4, characterized in that: After optimizing the mapping model according to the optimal optimization plan, a supervised learning mechanism is introduced to increase human intervention measures, adjust the output of the mapping model to better meet real-world needs, collect user feedback in real time, extract optimization suggestions, and adjust the mapping model parameters based on the optimization suggestions.

6. The method for constructing a refined assessment model of stroke functional impairment regions based on ICF theory according to claim 1, characterized in that: In the condition parameter correction model, a condition parameter correction database is constructed. The condition parameter correction database is configured with a condition pre-set subset, a condition post-set subset, and a condition parallel subset. Each of the condition pre-set subset, condition post-set subset, and condition parallel subset has multiple pre-set condition parameters.

7. A system for constructing a refined assessment model of stroke functional impairment regions based on ICF theory, used to implement the method described in any one of claims 1 to 6, characterized in that: include: The data acquisition module is used to acquire multimodal assessment data of stroke patients and perform preprocessing. The multimodal assessment data includes patient medical record information, functional assessment reports, and feedback information from medical staff. The data analysis module is used to perform cluster analysis on the preprocessed multimodal assessment data to obtain classification information, wherein the classification information includes functional impairment information and non-functional impairment information. The feature extraction module is used to extract features based on functional impairment information, output the extracted features, and summarize them into a feature label set. An encoding module is used to encode each feature label in the feature label set to obtain baseline encoded data; The model building module is used to build a mapping model based on the correspondence between the baseline encoded data and the extracted features, and to map new input data through the mapping model. The model is also dynamically optimized based on its application effect to continuously improve its mapping accuracy.

Images