Intelligent spine rehabilitation effect evaluation system

Abstract

The invention discloses an intelligent spinal rehabilitation effect evaluation system, which belongs to the technical field of intelligent medical treatment and comprises a patient data acquisition module, an individual state modeling module, a rehabilitation effect evaluation module and an intelligent feedback module. According to the method, an individual state modeling method combining clinical prior weighting and time sequence perception clustering is adopted, the individualized function benchmark and the dynamic motion state of the patient are accurately described, clinical features of the patient and time sequence changes in the rehabilitation training process are considered at the same time, and the evaluation result reflects the individual basic condition, the recovery rhythm and the improvement trend of the patient; a multi-scale attention rehabilitation effect evaluation method combined with an individualized baseline is adopted, the rehabilitation progress is quantified and the rehabilitation trend and stage are dynamically recognized based on the historical stable action state of a patient, and objective, individualized and continuous spinal rehabilitation effect evaluation is provided by integrating multi-dimensional information such as joint movement, myoelectricity activity and mechanical load.

Description

Technical Field

[0001] This invention belongs to the field of smart medical technology, specifically referring to an intelligent spinal rehabilitation effect evaluation system. Background Technology

[0002] An intelligent spinal rehabilitation effect assessment system refers to a system that continuously monitors, dynamically analyzes, and objectively quantifies the recovery of a patient's spinal function based on movements, muscle states, and clinical information generated during spinal rehabilitation training, and is used to comprehensively evaluate the rehabilitation effect. This system helps to promptly identify rehabilitation improvement trends, training fatigue, or potential degenerative risks, avoiding misjudgments caused by relying solely on subjective observation or single measurements.

[0003] However, in the process of intelligent spinal rehabilitation effect assessment, there are significant differences in the damaged parts, muscle strength, pain tolerance and daily activity ability of different patients, resulting in inconsistent rehabilitation starting points. Existing assessments usually use uniform quantitative indicators directly for evaluation, ignoring individual patient differences. This makes the rehabilitation curve overly dependent on absolute values, making it difficult to truly reflect the extent of rehabilitation. There are also technical problems such as the instability, compensatory behavior and fatigue effect of patients' movements during rehabilitation training, making it difficult to quantify continuous rehabilitation progress in single-time point assessments. Furthermore, the multimodal biomechanical characteristics are complex and traditional methods cannot be fully utilized, resulting in technical problems such as strong subjectivity and low accuracy in judging rehabilitation effects. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an intelligent spinal rehabilitation effect evaluation system. It creatively employs an individual state modeling method combining clinical prior weighting and temporal-aware clustering to accurately characterize the patient's individualized functional baseline and dynamic movement state. This method simultaneously considers the patient's clinical characteristics and the temporal changes in the rehabilitation training process, enabling the evaluation results to synchronously reflect the patient's individual baseline condition, recovery pace, and improvement trend. This provides stable and reliable individual state information for subsequent quantification of rehabilitation progress and risk assessment. Furthermore, it creatively employs a multi-scale attention rehabilitation effect evaluation method combined with an individualized baseline, achieving quantitative rehabilitation progress based on the patient's own historical stable movement state, dynamically identifying rehabilitation trends and stages, and comprehensively integrating multi-dimensional information such as joint movement, electromyographic activity, and mechanical load to provide an objective, personalized, and continuous evaluation of spinal rehabilitation effects.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent spinal rehabilitation effect evaluation system, including a patient data acquisition module, an individual state modeling module, a rehabilitation effect evaluation module and an intelligent feedback module;

[0006] The patient data acquisition module is used for patient data acquisition. Through initial data acquisition and data preprocessing, it obtains structured patient data and sends the structured patient data to the individual state modeling module and the rehabilitation effect evaluation module.

[0007] The individual state modeling module is used for individual state modeling. Based on the patient's structured data, it adopts an individual state modeling method that combines clinical prior weighting and time-aware clustering to obtain the patient's individual state information, and sends the patient's individual state information to the rehabilitation effect evaluation module.

[0008] The rehabilitation effect assessment module is used to assess the rehabilitation effect. Based on the patient's structured data and individual patient status information, it adopts a multi-scale attention rehabilitation effect assessment method combined with individualized baseline to obtain spinal rehabilitation effect assessment information, and sends the spinal rehabilitation effect assessment information to the intelligent feedback module.

[0009] The intelligent feedback module is used for intelligent feedback to obtain spinal rehabilitation feedback information.

[0010] Furthermore, the patient data collection specifically involves obtaining the patient's clinical information from the medical information system, simultaneously collecting the patient's biomechanical characteristics and action sequence data through wearable sensors and motion capture devices, and performing data preprocessing to obtain a structured patient dataset.

[0011] Furthermore, the individual state modeling includes the following steps: multimodal dimensionality reduction representation, motion state recognition, individualized baseline construction, and patient individual state information generation;

[0012] The multimodal dimensionality reduction representation is used to obtain a low-dimensional representation reflecting the individual movement and muscle characteristics of a patient. Specifically, it maps the patient's clinical information to a weight space consistent with the biomechanical feature dimension through a fully connected layer, constructs clinical prior weights, then adjusts the biomechanical features dimension by dimension based on the clinical prior weights to obtain biomechanical adjusted features, and then performs linear dimensionality reduction on the biomechanical adjusted features to generate a low-dimensional feature representation.

[0013] The motion state recognition is used to identify the patient's motion state and classify the state. Specifically, it constructs a temporal constraint distance based on low-dimensional feature representation and action acquisition time information. Then, based on the temporal constraint distance, it constructs a clustering objective function. The action samples are then clustered using a clustering algorithm to obtain the motion state. The conditional probability of each motion state is calculated, and the motion state with the highest conditional probability is taken as the final state label of the corresponding action sample to obtain the patient's motion state.

[0014] The individualized baseline construction is used to construct a dynamic characteristic baseline interval for individualized patients. Specifically, based on the state label, stable state samples are screened, and the mean and standard deviation of the stable state samples are calculated on each characteristic dimension to obtain baseline statistics. Individualized baseline intervals are then constructed for subsequent quantitative assessment of the patient's rehabilitation progress.

[0015] The generation of the patient's individual status information specifically involves integrating the patient's motion status, baseline statistics, and individualized baseline intervals to obtain the patient's individual status information.

[0016] Furthermore, the rehabilitation effect assessment, used to quantify rehabilitation progress and identify improvement trends, includes the following steps: multi-scale alignment calibration, rehabilitation progress quantification, multi-dimensional comprehensive scoring, rehabilitation stage determination, and assessment information generation;

[0017] The multi-scale alignment calibration is used to align and calibrate action sequences. Specifically, it involves calculating the feature deviation of action samples relative to steady-state samples, minimizing the difference between the feature deviation and historical feature deviations to perform baseline anchor time alignment, and then standardizing the action samples based on baseline statistics under cross-time period conditions to obtain standardized action features.

[0018] The quantification of rehabilitation progress specifically involves calculating the distance change of the standardized movement features relative to the steady-state sample, and then normalizing and mapping the baseline statistics to obtain the spatial improvement degree. Next, the smoothness of the movement sequence is calculated as the temporal stability degree. Finally, the spatial improvement degree and the temporal stability degree are weighted and fused to obtain the quantitative index of rehabilitation progress.

[0019] The multidimensional comprehensive score is used to evaluate the quality of rehabilitation from multiple dimensions based on the quantitative indicators of rehabilitation progress. Specifically, it is obtained by concatenating the features of standardized movements and the quantitative indicators of rehabilitation progress to obtain a temporal input tensor. Then, based on the Transformer temporal coding architecture, a multi-head temporal attention mechanism and a cross-feature attention mechanism are introduced, and a multidimensional score prediction head and a score fusion layer are set to construct a hybrid attention score prediction model to predict the score of the temporal input tensor and obtain a comprehensive score of rehabilitation effect.

[0020] The determination of the rehabilitation stage is specifically achieved by dynamically adjusting the rehabilitation stage boundary threshold by combining the historical statistical distribution of the comprehensive rehabilitation effect score, and then dividing the rehabilitation stage according to the rehabilitation stage boundary threshold to obtain the rehabilitation stage, which includes significant improvement, stagnant and regression.

[0021] The assessment information is generated by integrating the comprehensive score of rehabilitation effect with the rehabilitation stage to obtain spinal rehabilitation effect assessment information.

[0022] Furthermore, the intelligent feedback specifically involves providing training suggestions on training intensity and frequency based on spinal rehabilitation effect assessment information, and combining historical health score trends to indicate potential risks and generate rehabilitation feedback information to guide patients and rehabilitation personnel in optimizing subsequent rehabilitation plans.

[0023] The beneficial effects achieved by the present invention using the above solution are as follows:

[0024] (1) In the process of evaluating the effect of intelligent spinal rehabilitation, there are significant differences in the damaged parts, muscle strength, pain tolerance and daily activity ability of different patients, which leads to inconsistent rehabilitation starting points. Existing assessments usually use uniform quantitative indicators directly for evaluation, ignoring individual differences among patients, making the rehabilitation curve overly dependent on absolute values ​​and difficult to truly reflect the rehabilitation range. This solution creatively adopts an individual state modeling method that combines clinical prior weighting and temporal perception clustering to accurately depict the individualized functional benchmarks and dynamic movement states of patients. It can simultaneously consider the patient's clinical characteristics and the temporal changes in the rehabilitation training process, so that the assessment results can simultaneously reflect the patient's individual basic condition, recovery rhythm and improvement trend, providing stable and reliable individual state information for subsequent rehabilitation progress quantification and risk assessment.

[0025] (2) In the process of intelligent spinal rehabilitation effect assessment, there are technical problems such as the instability, compensatory behavior and fatigue effect of patients' movements during rehabilitation training, which makes it difficult to quantify continuous rehabilitation progress in a single time point assessment. In addition, the multimodal biomechanical characteristic information is complex and traditional methods are difficult to integrate, resulting in strong subjectivity and low accuracy in the assessment of rehabilitation effect. This solution creatively adopts a multi-scale attention rehabilitation effect assessment method combined with individualized baseline. It realizes the quantification of rehabilitation progress based on the patient's own historical stable movement state, dynamically identifies rehabilitation trends and stages, and integrates multi-dimensional information such as joint movement, electromyographic activity and mechanical load to provide objective, personalized and continuous evaluation of spinal rehabilitation effect. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of a module of an intelligent spinal rehabilitation effect evaluation system provided by the present invention;

[0027] Figure 2 A flowchart illustrating the module for modeling individual states;

[0028] Figure 3 This is a flowchart illustrating the rehabilitation effectiveness assessment module.

[0029] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0032] Example 1, see Figure 1 The present invention provides an intelligent spinal rehabilitation effect evaluation system, including a patient data acquisition module, an individual state modeling module, a rehabilitation effect evaluation module and an intelligent feedback module;

[0033] The patient data acquisition module is used for patient data acquisition. Through initial data acquisition and data preprocessing, it obtains structured patient data and sends the structured patient data to the individual state modeling module and the rehabilitation effect evaluation module.

[0034] The individual state modeling module is used for individual state modeling. Based on the patient's structured data, it adopts an individual state modeling method that combines clinical prior weighting and time-aware clustering to obtain the patient's individual state information, and sends the patient's individual state information to the rehabilitation effect evaluation module.

[0035] The rehabilitation effect assessment module is used to assess the rehabilitation effect. Based on the patient's structured data and individual patient status information, it adopts a multi-scale attention rehabilitation effect assessment method combined with individualized baseline to obtain spinal rehabilitation effect assessment information, and sends the spinal rehabilitation effect assessment information to the intelligent feedback module.

[0036] The intelligent feedback module is used for intelligent feedback to obtain spinal rehabilitation feedback information.

[0037] Example 2, see Figure 1 This embodiment is based on the above embodiment. The patient data collection specifically involves obtaining the patient's clinical information from the medical information system, synchronously collecting the patient's biomechanical characteristics and action sequence data through wearable sensors and motion capture devices, and performing data preprocessing to obtain a structured dataset of the patient.

[0038] The patient's clinical information includes, but is not limited to, the patient's age, gender, height, weight, past medical history, and pain score.

[0039] The biomechanical characteristics include kinematic characteristics, electromyographic signal characteristics, and pressure distribution characteristics;

[0040] The motion sequence data includes the patient's joint displacement, angular velocity, and mechanical load during rehabilitation training;

[0041] The data preprocessing includes denoising, normalization, and timestamp labeling.

[0042] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The individual state modeling is used to construct an individualized initial functional baseline and state representation for the patient, including the following steps: multimodal dimensionality reduction representation, motion state recognition, individualized baseline construction, and patient individual state information generation.

[0043] The multimodal dimensionality reduction representation is used to obtain a low-dimensional representation reflecting the individual movement and muscle characteristics of a patient. Specifically, it maps the patient's clinical information to a weight space consistent with the biomechanical feature dimension through a fully connected layer, constructs clinical prior weights, then adjusts the biomechanical features dimension by dimension based on the clinical prior weights to obtain biomechanical adjusted features, and then performs linear dimensionality reduction on the biomechanical adjusted features to generate a low-dimensional feature representation.

[0044] The calculation formula for the dimension-by-dimensional adjustment is as follows:

[0045] ;

[0046] In the formula, F represents the biomechanical adjustment characteristic. It is element-wise multiplication, and W is the clinical prior weight;

[0047] The linear dimensionality reduction can be achieved using methods including, but not limited to, principal component analysis, linear discriminant analysis, or enhanced linear autoencoders.

[0048] The motion state recognition is used to identify the patient's motion state and classify the state. Specifically, it constructs a temporal constraint distance based on low-dimensional feature representation and action acquisition time information. Then, based on the temporal constraint distance, it constructs a clustering objective function. The action samples are then clustered using a clustering algorithm to obtain the motion state. The conditional probability of each motion state is calculated, and the motion state with the highest conditional probability is taken as the final state label of the corresponding action sample to obtain the patient's motion state.

[0049] The motion states include steady state, compensatory state, pain-limiting state, and fatigue state;

[0050] The formula for calculating the construction time-constraint distance is as follows:

[0051] ;

[0052] In the formula, D(·) is the temporal constraint distance function, z i z is the low-dimensional feature representation of the i-th action sample. j Let be the low-dimensional feature representation of the j-th action sample, where i is the first index of the action sample, j is the second index of the action sample, and ||·||2 is the L2 norm, used to measure the Euclidean distance. It is a time-series penalty coefficient, used to control the degree to which the time difference affects the distance. It is the acquisition time of the i-th action sample. It is the acquisition time of the j-th action sample;

[0053] The formula for calculating the clustering objective function is as follows:

[0054] ;

[0055] In the formula, J is the clustering objective function, k is the first index of the motion state, K is the number of motion states, and C k It is the kth motion state, c k It is the cluster center of the kth motion state;

[0056] The clustering algorithms include, but are not limited to, K-means, hierarchical clustering, density-based clustering, and Gaussian mixture model clustering.

[0057] The formula for calculating the conditional probability is:

[0058] ;

[0059] In the formula, P(s) k │x i ) is the conditional probability that the i-th action sample belongs to the k-th motion state, s k This is the kth motion state, x i Let w be the i-th action sample, exp(·) be the natural exponential function, and w be the i-th action sample. k Here, u is the weight of the k-th motion state, and c is the second index of the motion state. u It is the cluster center of the u-th motion state, w u It is the weight of the u-th motion state;

[0060] The individualized baseline construction is used to build individualized dynamic characteristic baseline intervals for patients. Specifically, based on state labels, stable-state samples are selected, and the mean and standard deviation of the stable-state samples are calculated on each characteristic dimension to obtain baseline statistics. Individualized baseline intervals are then constructed for subsequent quantitative assessment of patient recovery progress. The calculation formula is as follows:

[0061] ;

[0062] In the formula, B v It represents the individualized baseline interval for the v-th feature dimension, where v is the feature dimension index. It is the mean of the steady-state samples along the v-th feature dimension. It is an interval coefficient, preferably 1.96. It is the standard deviation of the steady-state sample on the v-th feature dimension;

[0063] The generation of the patient's individual status information specifically involves integrating the patient's motion status, baseline statistics, and individualized baseline intervals to obtain the patient's individual status information.

[0064] By performing the above operations, this solution addresses the technical problem that, in the process of intelligent spinal rehabilitation effect assessment, there are significant differences in the injured parts, muscle strength, pain tolerance, and daily activity abilities of different patients, leading to inconsistent rehabilitation starting points. Existing assessments typically use uniform quantitative indicators directly, ignoring individual patient differences, causing rehabilitation curves to rely excessively on absolute values ​​and failing to accurately reflect the extent of rehabilitation. This solution creatively adopts an individual state modeling method that combines clinical prior weighting with temporal-aware clustering to accurately characterize the individualized functional benchmarks and dynamic movement states of patients. It can simultaneously consider the patient's clinical characteristics and the temporal changes in the rehabilitation training process, enabling the assessment results to synchronously reflect the patient's individual baseline condition, recovery pace, and improvement trend. This provides stable and reliable individual state information for subsequent quantification of rehabilitation progress and risk assessment.

[0065] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. The rehabilitation effect assessment is used to quantify rehabilitation progress and identify improvement trends, including the following steps: multi-scale alignment calibration, rehabilitation progress quantification, multi-dimensional comprehensive scoring, rehabilitation stage determination and assessment information generation;

[0066] The multi-scale alignment calibration is used to align and calibrate action sequences. Specifically, it involves calculating the feature deviation of action samples relative to steady-state samples, minimizing the difference between the feature deviation and historical feature deviations to perform baseline anchor time alignment, and then standardizing the action samples based on baseline statistics under cross-time period conditions to obtain standardized action features.

[0067] The formula for calculating the characteristic deviation is:

[0068] ;

[0069] In the formula, d i It is the feature bias of the i-th action sample. It is the feature mean vector of the stable-state samples, and ||·||2 is the L2 norm;

[0070] The formula for calculating the difference between the minimized feature bias and the historical bias is as follows:

[0071] ;

[0072] In the formula, It is the optimal time offset. It is the time offset. It is the historical feature bias of the i-th action sample;

[0073] The calculation formula for the standardization process is as follows:

[0074] ;

[0075] In the formula, z cal These are the characteristics of standardized movements. It is the characteristic standard deviation of the steady-state sample, z cur It is a low-dimensional feature representation of action samples in the current time period. It is the standard deviation of the features of the action samples in the current time period. It is the feature mean of the action samples in the current time period;

[0076] The quantification of rehabilitation progress specifically involves calculating the distance change of the standardized movement features relative to the steady-state sample, and then normalizing and mapping the baseline statistics to obtain the spatial improvement degree. Next, the smoothness of the movement sequence is calculated as the temporal stability degree. Finally, the spatial improvement degree and the temporal stability degree are weighted and fused to obtain the quantitative index of rehabilitation progress.

[0077] The formula for calculating the spatial improvement degree is as follows:

[0078] ;

[0079] In the formula, P v It represents the spatial improvement of the v-th feature dimension, and tanh(·) is the tanh activation function. It is the change in distance of the standardized action features relative to the steady-state sample. It is a minimal constant term;

[0080] The formula for calculating the time stability is:

[0081] ;

[0082] In the formula, S t This represents the time stability of time interval t, where t is the time interval index, M is the action sequence length, and z is the time stability of time interval t. i+1 It is the low-dimensional feature representation of the (i+1)th action sample. It is the scaling factor;

[0083] The formula for calculating the quantitative indicators of rehabilitation progress is as follows:

[0084] ;

[0085] In the formula, P total It is a quantitative indicator of rehabilitation progress, where V is the total number of feature dimensions, and w v It is the weight of the v-th feature dimension;

[0086] The multidimensional comprehensive score is used to evaluate the quality of rehabilitation from multiple dimensions based on the quantitative indicators of rehabilitation progress. Specifically, it is obtained by concatenating the features of standardized movements and the quantitative indicators of rehabilitation progress to obtain a temporal input tensor. Then, based on the Transformer temporal coding architecture, a multi-head temporal attention mechanism and a cross-feature attention mechanism are introduced, and a multidimensional score prediction head and a score fusion layer are set to construct a hybrid attention score prediction model to predict the score of the temporal input tensor and obtain a comprehensive score of rehabilitation effect.

[0087] The multi-head temporal attention mechanism is used to adaptively extract the action progress dependencies between different time segments and obtain a temporally enhanced representation;

[0088] The cross-feature attention mechanism is used to learn the importance weights of different feature dimensions in score prediction and obtain feature-enhanced representations;

[0089] The multidimensional scoring prediction head specifically outputs an immediate score, a single rehabilitation training score, and a long-term trend score through three parallel scoring prediction heads.

[0090] The scoring fusion layer specifically generates a comprehensive rehabilitation effect score by weighting and fusing the immediate score, the single rehabilitation training score, and the long-term trend score.

[0091] The determination of the rehabilitation stage is specifically achieved by dynamically adjusting the rehabilitation stage boundary threshold by combining the historical statistical distribution of the comprehensive rehabilitation effect score, and then dividing the rehabilitation stage according to the rehabilitation stage boundary threshold to obtain the rehabilitation stage, which includes significant improvement, stagnant and regression.

[0092] The formula for calculating the dynamically adjusted rehabilitation stage boundary threshold is as follows:

[0093] ;

[0094] ;

[0095] In the formula, T high It is the high threshold of the boundary. It is the mean of the comprehensive rehabilitation effect score under historical statistical distribution. It is the adjustment coefficient. It is the standard deviation of the comprehensive score of rehabilitation effect under historical statistical distribution;

[0096] The formula for calculating the division of rehabilitation stages is as follows:

[0097] ;

[0098] In the formula, "stage" represents the rehabilitation phase, "level 1" represents the first stage of rehabilitation, used to indicate significant improvement, and R... final It is a comprehensive score of rehabilitation effectiveness. Level 2 is the second stage of rehabilitation, which is used to indicate that the progress is at a steady level, and Level 3 is the third stage of rehabilitation, which is used to indicate that the progress is regressing.

[0099] The assessment information is generated by integrating the comprehensive score of rehabilitation effect with the rehabilitation stage to obtain spinal rehabilitation effect assessment information.

[0100] By performing the above operations, this solution addresses the technical problems in intelligent spinal rehabilitation effect assessment, such as the instability, compensatory behavior, and fatigue effects of patients' movements during rehabilitation training, making it difficult to quantify continuous rehabilitation progress at a single time point. Furthermore, the complex multimodal biomechanical characteristics make it difficult to comprehensively utilize traditional methods, leading to strong subjectivity and low accuracy in rehabilitation effect judgment. This solution creatively adopts a multi-scale attention rehabilitation effect assessment method that combines individualized baselines. It achieves the quantification of rehabilitation progress based on the patient's own historical stable movement state, dynamically identifies rehabilitation trends and stages, and integrates multi-dimensional information such as joint movement, electromyographic activity, and mechanical load to provide an objective, personalized, and continuous evaluation of spinal rehabilitation effect.

[0101] Example 5, see Figure 1 This embodiment is based on the above embodiment. Specifically, the intelligent feedback is to provide training suggestions on training intensity and frequency based on spinal rehabilitation effect assessment information, and to indicate potential risks by combining historical health score trends, thereby generating rehabilitation feedback information to guide patients and rehabilitation personnel to optimize subsequent rehabilitation plans.

[0102] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0103] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0104] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An intelligent spinal rehabilitation effect evaluation system, characterized in that: It includes a patient data acquisition module, an individual status modeling module, a rehabilitation effect evaluation module, and an intelligent feedback module; The patient data acquisition module is used for patient data acquisition. Through initial data acquisition and data preprocessing, it obtains structured patient data and sends the structured patient data to the individual state modeling module and the rehabilitation effect evaluation module. The individual state modeling module is used for individual state modeling. Based on the patient's structured data, it adopts an individual state modeling method that combines clinical prior weighting and time-aware clustering to obtain the patient's individual state information, and sends the patient's individual state information to the rehabilitation effect evaluation module. The individual state modeling includes the following steps: multimodal dimensionality reduction representation, motion state recognition, individualized baseline construction, and generation of individual patient state information; The rehabilitation effect assessment module is used to assess the rehabilitation effect. Based on the patient's structured data and individual patient status information, it adopts a multi-scale attention rehabilitation effect assessment method combined with individualized baseline to obtain spinal rehabilitation effect assessment information, and sends the spinal rehabilitation effect assessment information to the intelligent feedback module. The rehabilitation effect assessment is used to quantify rehabilitation progress and identify improvement trends, including the following steps: multi-scale alignment calibration, rehabilitation progress quantification, multi-dimensional comprehensive scoring, rehabilitation stage determination, and assessment information generation. The intelligent feedback module is used for intelligent feedback to obtain spinal rehabilitation feedback information.

2. The intelligent spinal rehabilitation effect evaluation system according to claim 1, characterized in that: The multimodal dimensionality reduction representation is used to obtain a low-dimensional representation reflecting the individual movement and muscle characteristics of a patient. Specifically, it maps the patient's clinical information to a weight space consistent with the biomechanical feature dimension through a fully connected layer, constructs clinical prior weights, then adjusts the biomechanical features dimension by dimension based on the clinical prior weights to obtain biomechanical adjusted features, and then performs linear dimensionality reduction on the biomechanical adjusted features to generate a low-dimensional feature representation. The motion state recognition is used to identify the patient's motion state and classify the state. Specifically, it constructs a temporal constraint distance based on low-dimensional feature representation and action acquisition time information. Then, based on the temporal constraint distance, it constructs a clustering objective function. The action samples are then clustered using a clustering algorithm to obtain the motion state. The conditional probability of each motion state is calculated, and the motion state with the highest conditional probability is taken as the final state label of the corresponding action sample to obtain the patient's motion state. The individualized baseline construction is used to construct a dynamic characteristic baseline interval for individualized patients. Specifically, based on the state label, stable state samples are screened, and the mean and standard deviation of the stable state samples are calculated on each characteristic dimension to obtain baseline statistics. Individualized baseline intervals are then constructed for subsequent quantitative assessment of the patient's rehabilitation progress. The generation of the patient's individual status information specifically involves integrating the patient's motion status, baseline statistics, and individualized baseline intervals to obtain the patient's individual status information.

3. The intelligent spinal rehabilitation effect evaluation system according to claim 2, characterized in that: The multi-scale alignment calibration is used to align and calibrate action sequences based on individualized baseline intervals. Specifically, it involves calculating the feature deviation of action samples relative to steady-state samples, minimizing the difference between the feature deviation and historical feature deviations to perform baseline anchor time alignment, and then standardizing the action samples based on baseline statistics under cross-time period conditions to obtain standardized action features.

4. The intelligent spinal rehabilitation effect evaluation system according to claim 3, characterized in that: The quantification of rehabilitation progress specifically involves calculating the distance change of standardized movement features relative to the steady-state sample, and then normalizing and mapping it using baseline statistics to obtain the spatial improvement degree. Next, the smoothness of the movement sequence is calculated as the temporal stability degree. Finally, the spatial improvement degree and the temporal stability degree are weighted and fused to obtain the quantitative index of rehabilitation progress.

5. The intelligent spinal rehabilitation effect evaluation system according to claim 4, characterized in that: The multidimensional comprehensive score is used to evaluate the quality of rehabilitation from multiple dimensions based on the quantitative indicators of rehabilitation progress. Specifically, it is obtained by concatenating the features of standardized movements and the quantitative indicators of rehabilitation progress to obtain a temporal input tensor. Then, based on the Transformer temporal coding architecture, a multi-head temporal attention mechanism and a cross-feature attention mechanism are introduced, and a multidimensional score prediction head and a score fusion layer are set to construct a hybrid attention score prediction model to predict the score of the temporal input tensor and obtain a comprehensive score of rehabilitation effect.

6. The intelligent spinal rehabilitation effect evaluation system according to claim 5, characterized in that: The determination of the rehabilitation stage is specifically achieved by dynamically adjusting the rehabilitation stage boundary threshold by combining the historical statistical distribution of the comprehensive rehabilitation effect score, and then dividing the rehabilitation stage according to the rehabilitation stage boundary threshold to obtain the rehabilitation stage, which includes significant improvement, stagnant and regression. The assessment information is generated by integrating the comprehensive score of rehabilitation effect with the rehabilitation stage to obtain spinal rehabilitation effect assessment information.

7. The intelligent spinal rehabilitation effect evaluation system according to claim 6, characterized in that: The intelligent feedback specifically involves providing training suggestions on training intensity and frequency based on spinal rehabilitation effect assessment information, and combining historical health score trends to indicate potential risks and generate rehabilitation feedback information to guide patients and rehabilitation personnel in optimizing subsequent rehabilitation plans.

8. The intelligent spinal rehabilitation effect evaluation system according to claim 7, characterized in that: The patient data collection specifically involves obtaining the patient's clinical information from the medical information system, synchronously collecting the patient's biomechanical characteristics and action sequence data through wearable sensors and motion capture devices, and performing data preprocessing to obtain a structured patient dataset.

Images