Orthopedic rehabilitation system and rehabilitation evaluation method based on data feedback

Abstract

The invention relates to the technical field of rehabilitation assessment, in particular to an orthopedic rehabilitation system and a rehabilitation assessment method based on data feedback, comprising the following steps: training a classifier by using a first single-mode data set formed by myoelectricity feedback data and corresponding rehabilitation status labels to obtain a first single data model of orthopedic rehabilitation assessment; a classifier is trained according to a second single-mode data set formed by the bone image data and the corresponding rehabilitation condition labels, and a second single-data model of orthopedic rehabilitation evaluation is obtained; and respectively carrying out transfer learning on the orthopedic rehabilitation assessment second single data model and the orthopedic rehabilitation assessment first single data model to obtain an orthopedic rehabilitation assessment multi-data model. According to the method, the orthopedic rehabilitation assessment multi-data model is constructed, objective assessment of orthopedic rehabilitation conditions is completed according to objective data such as the myoelectricity feedback data and the skeleton image data of the assessment object, the intelligent degree of assessment can be improved, and the assessment efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation assessment, and in particular to a rehabilitation assessment method and system based on data feedback. Background Art

[0002] Among common diseases, orthopedic diseases have a high incidence rate, which has a great impact on people's physical health and quality of life. Rehabilitation medicine is an important part of modern medicine and plays a great role in the functional recovery of musculoskeletal diseases. Therefore, strengthening the clinical rehabilitation treatment of bone diseases is of great significance for restoring patients' body functions and improving their quality of life.

[0003] At present, the bone recovery is mainly evaluated through patient self-assessment questionnaires, observation and direct measurement. Commonly used self-assessment questionnaires include the Netherlands Musculoskeletal Injury Questionnaire (DMQ) and the Maastricht Upper Limb Musculoskeletal Questionnaire (MUEQ). The patient's bone recovery is obtained through self-assessment based on the items on the questionnaire and through the doctor's assessment of the patient's physical condition.

[0004] However, since the self-assessment method of the patient is subjective, it is impossible to accurately assess one's own condition. Therefore, in the process of assessing the patient's bone condition, the existing technology cannot accurately judge the patient's bone recovery based on objective data, and has the defects of low intelligence and low assessment accuracy. Summary of the invention

[0005] The purpose of the present invention is to provide an orthopedic rehabilitation system and a rehabilitation assessment method based on data feedback, so as to solve the technical problems in the prior art that it is impossible to accurately judge the patient's bone recovery condition based on objective data, and there is a low level of intelligence in the assessment of bone recovery and a low assessment accuracy.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: A rehabilitation assessment method based on data feedback comprises the following steps: Collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; Labeling the electromyographic feedback data and the bone imaging data to obtain rehabilitation status labels for the electromyographic feedback data and the orthopedic imaging data; Using the first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels, a classifier is trained to obtain a first single-data model for orthopedic rehabilitation assessment, and using the second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels, a classifier is trained to obtain a second single-data model for orthopedic rehabilitation assessment; Using the first single-modality data set and the second single-modality data set, transfer learning is performed on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain an orthopedic rehabilitation assessment multi-data model.

[0007] As a preferred embodiment of the present invention, the rehabilitation status label includes: a good label and a bad label.

[0008] As a preferred solution of the present invention, the method for labeling the electromyographic feedback data and the skeletal image data includes: The orthopedic rehabilitation status is judged based on electromyographic feedback data, including: When the orthopedic rehabilitation condition is judged to be good, the corresponding electromyographic feedback data is assigned a good label; When the orthopedic rehabilitation condition is judged as poor, the corresponding electromyographic feedback data is assigned a poor label; The orthopedic rehabilitation status is judged based on the bone imaging data, including: When the orthopedic rehabilitation condition is judged to be good, the corresponding bone image data is assigned a good label; When the orthopedic rehabilitation condition is judged to be poor, the corresponding bone image data is assigned a poor label.

[0009] As a preferred embodiment of the present invention, the method for constructing the first single data model for orthopedic rehabilitation assessment includes: Dividing the first unimodal dataset into a training set and a test set; Based on the training set of the first single-modal data set, the classifier is trained by taking the electromyographic feedback data as the input of the classifier and the rehabilitation status label of the electromyographic feedback data as the output of the classifier to obtain the first single-data model of orthopedic rehabilitation assessment; Based on the test set of the first unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The first data model of orthopedic rehabilitation assessment is: Labela=SVMa(Da); Wherein, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation evaluation.

[0010] As a preferred embodiment of the present invention, the method for constructing the second single data model for orthopedic rehabilitation assessment includes: Dividing the second unimodal dataset into a training set and a test set; Based on the training set of the second single-modality data set, the bone imaging data is used as the input of the classifier, and the rehabilitation status label of the bone imaging data is used as the output of the classifier, and the classifier is trained to obtain the second single-modality data model of orthopedic rehabilitation assessment; Based on the test set of the second unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The second data model of orthopedic rehabilitation assessment is: Labelg=SVMg(Dg); In the formula, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model of orthopedic rehabilitation evaluation, Dg is the bone imaging data, and SVMg is the classifier in the second single data model of orthopedic rehabilitation evaluation.

[0011] As a preferred embodiment of the present invention, the method for constructing the orthopedic rehabilitation assessment multi-data model includes: The bone imaging data is used as the input of the first single data model of orthopedic rehabilitation assessment, and the rehabilitation status label of the bone imaging data is used as the output of the first single data model of orthopedic rehabilitation assessment; The electromyographic feedback data is used as the input of the second single data model of orthopedic rehabilitation evaluation, and the rehabilitation status label of the electromyographic feedback data is used as the output of the second single data model of orthopedic rehabilitation evaluation; Constructing a retraining loss function for the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment, and retraining the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment based on the retraining loss function to obtain a multi-data model of orthopedic rehabilitation assessment; The orthopedic rehabilitation assessment multi-data model includes: Labelgc=SVMa(Dg); Wherein, Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model of orthopedic rehabilitation assessment, Dg is the bone imaging data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation assessment; Labelac=SVMg(Da); Wherein, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the second single data model of orthopedic rehabilitation evaluation.

[0012] As a preferred embodiment of the present invention, the loss function of the first single data model of orthopedic rehabilitation assessment is: ; Where La is the output loss value of the first single data model for orthopedic rehabilitation evaluation, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation evaluation, and Labelar is the true value of the rehabilitation status label of the electromyographic feedback data. for and The L2 norm between .

[0013] As a preferred embodiment of the present invention, the loss function of the second single data model for orthopedic rehabilitation assessment is: ; Where Lg is the output loss value of the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgr is the true value of the rehabilitation status label of the bone imaging data. for and The L2 norm between .

[0014] As a preferred embodiment of the present invention, the loss function of the orthopedic rehabilitation assessment multi-data model is: ; Wherein, Lag is the output loss of the multi-data model for orthopedic rehabilitation assessment, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation assessment, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model for orthopedic rehabilitation assessment. for and The L2 norm between for and The L2 norm between .

[0015] As a preferred embodiment of the present invention, the present invention provides an orthopedic rehabilitation system, which is applied to a rehabilitation evaluation method based on data feedback. The system includes: A data acquisition unit, used to collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; A data processing unit, used for labeling the electromyographic feedback data and the bone image data to obtain rehabilitation status labels for the electromyographic feedback data and the orthopedic image data; A deep learning unit, for training a classifier using a first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels to obtain a first single-data model for orthopedic rehabilitation assessment, and a second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels to obtain a second single-data model for orthopedic rehabilitation assessment; The transfer learning unit is used to use the first single-modality data set and the second single-modality data set to perform transfer learning on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain a multi-data model of orthopedic rehabilitation assessment.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention constructs a multi-data model for orthopedic rehabilitation assessment, and realizes objective assessment of orthopedic rehabilitation status based on objective data such as electromyographic feedback data and bone imaging data of the assessment object. In addition, modeling the assessment of rehabilitation status can improve the intelligence level of the assessment and improve the efficiency of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0018] Figure 1 A flow chart of a rehabilitation assessment method based on data feedback provided by an embodiment of the present invention; Figure 2 A block diagram of an orthopedic rehabilitation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, the present invention provides a rehabilitation assessment method based on data feedback, comprising the following steps: Collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; Labeling the electromyographic feedback data and the bone imaging data to obtain rehabilitation status labels of the electromyographic feedback data and the orthopedic imaging data; Using the first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels, a classifier is trained to obtain a first single-data model for orthopedic rehabilitation assessment, and using the second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels, a classifier is trained to obtain a second single-data model for orthopedic rehabilitation assessment; Using the first single-modality data set and the second single-modality data set, transfer learning is performed on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain an orthopedic rehabilitation assessment multi-data model.

[0021] The present invention collects electromyographic feedback data and bone imaging data reflecting the orthopedic rehabilitation situation, which are used to achieve an objective evaluation of the orthopedic rehabilitation situation, thereby ensuring the objectivity of the orthopedic rehabilitation situation evaluation and being more in line with the actual rehabilitation situation.

[0022] The present invention utilizes the classifier model in the machine learning model to analyze the electromyographic feedback data and the bone imaging data, so that the electromyographic feedback data, the bone imaging data and the rehabilitation status are correlated and mapped, thereby directly obtaining the rehabilitation status of the evaluation object based on the electromyographic feedback data and the bone imaging data of the evaluation object according to the trained classifier.

[0023] The present invention uses data sets composed of electromyographic feedback data, bone imaging data and corresponding rehabilitation status labels to train classifiers respectively, and obtains a first single data model of orthopedic rehabilitation assessment that obtains rehabilitation status based on electromyographic feedback data, and a second single data model of orthopedic rehabilitation assessment that obtains rehabilitation status based on bone imaging data. Both models evaluate rehabilitation status based on single-modality data.

[0024] Rehabilitation status assessment based on single-modal data has the advantages of small amount of model processing data and high efficiency, but there is also a defect that single-modal data has poor assessment accuracy compared to rehabilitation status assessment based on multi-modal data. Among them, rehabilitation status assessment based on multi-modal data refers to inputting electromyographic feedback data and bone imaging data at the same time, and the classifier outputs the classification result of the rehabilitation status. Therefore, compared with only using one of the electromyographic feedback data and bone imaging data to obtain the classification result of the rehabilitation status, more data features need to be processed. It is precisely because more data features can be obtained that the classifier has richer information for judging the classification result of the rehabilitation status, and the accuracy of the output classification result of the rehabilitation status is higher.

[0025] Based on the above situation, the present invention utilizes transfer learning (a machine learning method, which refers to a pre-trained model being reused in another task by using a model suitable for a different but related task to narrow the data search range of possible models in a favorable manner), and retrains the first single data model of orthopedic rehabilitation assessment using bone imaging data, and retrains the second single data model of orthopedic rehabilitation assessment using electromyography feedback data to obtain a multi-data model of orthopedic rehabilitation assessment.

[0026] The retraining of the first single data model for orthopedic rehabilitation assessment can realize the classification results of rehabilitation conditions obtained through electromyographic feedback data, and train it to obtain classification results of rehabilitation conditions through bone imaging data. The classification accuracy inherits the model accuracy of the rehabilitation condition classification results obtained through electromyographic feedback data, so that the first single data model for orthopedic rehabilitation assessment after retraining integrates the classification accuracy of both electromyographic feedback data and bone imaging data, and only needs to input bone imaging data. In other words, after retraining, only single-modal data input is needed to obtain multi-modal data output results. The data processing amount is reduced, but the accuracy remains at the level of rehabilitation condition assessment based on multi-modal data.

[0027] Similarly, the retraining of the second single data model for orthopedic rehabilitation assessment can realize the classification results of rehabilitation conditions obtained through bone imaging data, and train it to obtain the classification results of rehabilitation conditions through electromyographic feedback data, and the classification accuracy inherits the model accuracy of the rehabilitation status classification results obtained through bone imaging data, so that the second single data model for orthopedic rehabilitation assessment after retraining integrates the classification accuracy of both bone imaging data and electromyographic feedback data, and only needs to input electromyographic feedback data. In other words, after retraining, only single-modal data input is required to obtain multi-modal data output results, and the data processing amount is reduced, but the accuracy remains at the level of rehabilitation status assessment based on multi-modal data.

[0028] Therefore, the efficiency of the retrained multi-data model for orthopedic rehabilitation assessment is improved and the accuracy is maintained at a high level.

[0029] The recovery status labels include: good labels and bad labels.

[0030] Methods for labeling electromyographic feedback data and skeletal imaging data include: The orthopedic rehabilitation status is judged based on electromyographic feedback data (based on expert experience), including: When the orthopedic rehabilitation condition is judged to be good, the corresponding electromyographic feedback data is assigned a good label; When the orthopedic rehabilitation condition is judged as poor, the corresponding electromyographic feedback data is assigned a poor label; The orthopedic rehabilitation status is judged based on the bone imaging data (based on expert experience), including: When the orthopedic rehabilitation condition is judged to be good, the corresponding bone image data is assigned a good label; When the orthopedic rehabilitation condition is judged to be poor, the corresponding bone image data is assigned a poor label.

[0031] The construction method of the first data model for orthopedic rehabilitation assessment includes: Divide the first unimodal dataset into a training set and a test set; Based on the training set of the first single-modal data set, the electromyographic feedback data is used as the input of the classifier, and the rehabilitation status label of the electromyographic feedback data is used as the output of the classifier, and the classifier is trained to obtain the first single-data model for orthopedic rehabilitation assessment; Based on the test set of the first unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The first data model for orthopedic rehabilitation assessment is: Labela=SVMa(Da); Wherein, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation evaluation.

[0032] The construction method of the second data model for orthopedic rehabilitation assessment includes: Divide the second unimodal dataset into a training set and a test set; Based on the training set of the second single modality data set, the bone imaging data is used as the input of the classifier, and the rehabilitation status label of the bone imaging data is used as the output of the classifier, and the classifier is trained to obtain the second single data model of orthopedic rehabilitation assessment; Based on the test set of the second unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The second data model for orthopedic rehabilitation assessment is: Labelg=SVMg(Dg); In the formula, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model of orthopedic rehabilitation evaluation, Dg is the bone imaging data, and SVMg is the classifier in the second single data model of orthopedic rehabilitation evaluation.

[0033] The present invention uses data sets composed of electromyographic feedback data, bone imaging data and corresponding rehabilitation status labels to train classifiers respectively, and obtains a first single data model of orthopedic rehabilitation assessment that obtains rehabilitation status based on electromyographic feedback data, and a second single data model of orthopedic rehabilitation assessment that obtains rehabilitation status based on bone imaging data. Both models evaluate rehabilitation status based on single-modality data.

[0034] The construction method of the multi-data model for orthopedic rehabilitation assessment includes: The bone imaging data is used as the input of the first single data model of orthopedic rehabilitation assessment, and the rehabilitation status label of the bone imaging data is used as the output of the first single data model of orthopedic rehabilitation assessment; The electromyographic feedback data is used as the input of the second single data model of orthopedic rehabilitation evaluation, and the rehabilitation status label of the electromyographic feedback data is used as the output of the second single data model of orthopedic rehabilitation evaluation; Constructing a retraining loss function for the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment, and retraining the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment based on the retraining loss function to obtain a multi-data model of orthopedic rehabilitation assessment; The multi-data model for orthopedic rehabilitation assessment includes: Labelgc=SVMa(Dg); Wherein, Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model of orthopedic rehabilitation assessment, Dg is the bone imaging data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation assessment; The retraining of the first single data model for orthopedic rehabilitation assessment can realize the classification results of rehabilitation conditions obtained through electromyographic feedback data, and train it to obtain classification results of rehabilitation conditions through bone imaging data. The classification accuracy inherits the model accuracy of the rehabilitation condition classification results obtained through electromyographic feedback data, so that the first single data model for orthopedic rehabilitation assessment after retraining integrates the classification accuracy of both electromyographic feedback data and bone imaging data, and only needs to input bone imaging data. In other words, after retraining, only single-modal data input is needed to obtain multi-modal data output results. The data processing amount is reduced, but the accuracy remains at the level of rehabilitation condition assessment based on multi-modal data.

[0035] Labelac=SVMg(Da); Wherein, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the second single data model of orthopedic rehabilitation evaluation.

[0036] Similarly, the retraining of the second single data model for orthopedic rehabilitation assessment can realize the classification results of rehabilitation conditions obtained through bone imaging data, and train it to obtain the classification results of rehabilitation conditions through electromyographic feedback data, and the classification accuracy inherits the model accuracy of the rehabilitation status classification results obtained through bone imaging data, so that the second single data model for orthopedic rehabilitation assessment after retraining integrates the classification accuracy of both bone imaging data and electromyographic feedback data, and only needs to input electromyographic feedback data. In other words, after retraining, only single-modal data input is required to obtain multi-modal data output results, and the data processing amount is reduced, but the accuracy remains at the level of rehabilitation status assessment based on multi-modal data.

[0037] Therefore, the efficiency of the retrained multi-data model for orthopedic rehabilitation assessment is improved and the accuracy is maintained at a high level.

[0038] The loss function of the first data model for orthopedic rehabilitation evaluation is: ; Where La is the output loss value of the first single data model for orthopedic rehabilitation evaluation, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation evaluation, and Labelar is the true value of the rehabilitation status label of the electromyographic feedback data. for and The L2 norm between .

[0039] The loss function of the first single data model for orthopedic rehabilitation assessment is the difference between the model output and the true value. This loss function is used to ensure that the trained first single data model for orthopedic rehabilitation assessment can achieve the highest accuracy performance in the case of single modal data.

[0040] The loss function of the second data model for orthopedic rehabilitation evaluation is: ; Where Lg is the output loss value of the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgr is the true value of the rehabilitation status label of the bone imaging data. for and The L2 norm between .

[0041] The loss function of the second single data model for orthopedic rehabilitation evaluation is the difference between the model output and the true value. This loss function is used to ensure that the trained second single data model for orthopedic rehabilitation evaluation can achieve the highest accuracy performance in the case of single modal data.

[0042] The loss function of the multi-data model for orthopedic rehabilitation evaluation is: ; Wherein, Lag is the output loss of the multi-data model for orthopedic rehabilitation assessment, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation assessment, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model for orthopedic rehabilitation assessment. for and The L2 norm between for and The L2 norm between .

[0043] The loss function of the multi-data model for orthopedic rehabilitation evaluation consists of two parts. The first part is , is the difference between the output of the first single data model of orthopedic rehabilitation assessment after retraining and the output of the second single data model of orthopedic rehabilitation assessment, so that the first single data model of orthopedic rehabilitation assessment after retraining can integrate the accuracy performance of the second single data model of orthopedic rehabilitation assessment, combined with its own accuracy performance, that is, the first single data model of orthopedic rehabilitation assessment after retraining has both high-precision performance in the case of single-modal data of electromyographic feedback data and high-precision performance in the case of single-modal data of bone imaging data, so that the first single data model of orthopedic rehabilitation assessment after retraining can integrate the high-precision performance in the two-modal data cases of electromyographic feedback data and bone imaging data, and realizes the high-precision classification results in the two-modal data cases of electromyographic feedback data and bone imaging data according to the input processing of the single-modal data of bone imaging data.

[0044] The second part is , is the difference between the output of the second single data model of orthopedic rehabilitation assessment after retraining and the output of the first single data model of orthopedic rehabilitation assessment, so that the second single data model of orthopedic rehabilitation assessment after retraining can integrate the accuracy performance of the first single data model of orthopedic rehabilitation assessment, combined with its own accuracy performance, that is, the second single data model of orthopedic rehabilitation assessment after retraining has both high-precision performance in the case of single-modal data of electromyographic feedback data and high-precision performance in the case of single-modal data of bone imaging data, so that the second single data model of orthopedic rehabilitation assessment after retraining can integrate the high-precision performance in the two-modal data of bone imaging data and electromyographic feedback data, and realizes the high-precision classification results in the two-modal data of bone imaging data and electromyographic feedback data according to the input processing of the single-modal data of electromyographic feedback data.

[0045] like Figure 2 As shown, the present invention provides an orthopedic rehabilitation system, which is applied to a rehabilitation evaluation method based on data feedback, and the system includes: A data acquisition unit, used to collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; A data processing unit, used for labeling the electromyographic feedback data and the bone imaging data to obtain rehabilitation status labels for the electromyographic feedback data and the orthopedic imaging data; A deep learning unit, for training a classifier using a first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels to obtain a first single-data model for orthopedic rehabilitation assessment, and a second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels to obtain a second single-data model for orthopedic rehabilitation assessment; The transfer learning unit is used to use the first single-modality data set and the second single-modality data set to perform transfer learning on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain a multi-data model of orthopedic rehabilitation assessment.

[0046] The present invention constructs a multi-data model for orthopedic rehabilitation assessment, and realizes objective assessment of orthopedic rehabilitation status based on objective data such as electromyographic feedback data and bone imaging data of the assessment object. In addition, modeling the assessment of rehabilitation status can improve the intelligence level of the assessment and improve the efficiency of the assessment.

[0047] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A rehabilitation assessment method based on data feedback, characterized in that: The following steps are involved: Collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; Labeling the electromyographic feedback data and the bone imaging data to obtain rehabilitation status labels for the electromyographic feedback data and the orthopedic imaging data; Using the first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels, a classifier is trained to obtain a first single-data model for orthopedic rehabilitation assessment, and using the second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels, a classifier is trained to obtain a second single-data model for orthopedic rehabilitation assessment; Using the first single-modality data set and the second single-modality data set, transfer learning is performed on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain an orthopedic rehabilitation assessment multi-data model.

2. A rehabilitation assessment method based on data feedback according to claim 1, characterized in that: The recovery status labels include: a good label and a bad label.

3. A rehabilitation assessment method based on data feedback according to claim 2, characterized in that: The method for labeling the electromyographic feedback data and the skeletal image data includes: The orthopedic rehabilitation status is judged based on electromyographic feedback data, including: When the orthopedic rehabilitation condition is judged to be good, the corresponding electromyographic feedback data is assigned a good label; When the orthopedic rehabilitation condition is judged as poor, the corresponding electromyographic feedback data is assigned a poor label; The orthopedic rehabilitation status is judged based on the bone imaging data, including: When the orthopedic rehabilitation condition is judged to be good, the corresponding bone image data is assigned a good label; When the orthopedic rehabilitation condition is judged to be poor, the corresponding bone image data is assigned a poor label.

4. A rehabilitation assessment method based on data feedback according to claim 3, characterized in that: The method for constructing the first single data model for orthopedic rehabilitation assessment includes: Dividing the first unimodal dataset into a training set and a test set; Based on the training set of the first single-modal data set, the classifier is trained by taking the electromyographic feedback data as the input of the classifier and the rehabilitation status label of the electromyographic feedback data as the output of the classifier to obtain the first single-data model of orthopedic rehabilitation assessment; Based on the test set of the first unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The first data model of orthopedic rehabilitation assessment is: Labela=SVMa(Da); Wherein, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation evaluation.

5. A rehabilitation assessment method based on data feedback according to claim 4, characterized in that: The method for constructing the second single data model for orthopedic rehabilitation assessment includes: Dividing the second unimodal dataset into a training set and a test set; Based on the training set of the second single-modality data set, the bone imaging data is used as the input of the classifier, and the rehabilitation status label of the bone imaging data is used as the output of the classifier, and the classifier is trained to obtain the second single-modality data model of orthopedic rehabilitation assessment; Based on the test set of the second unimodal dataset, the performance of the classifier is evaluated using the mean square error as the evaluation indicator; The second data model of orthopedic rehabilitation assessment is: Labelg=SVMg(Dg); In the formula, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model of orthopedic rehabilitation evaluation, Dg is the bone imaging data, and SVMg is the classifier in the second single data model of orthopedic rehabilitation evaluation.

6. A rehabilitation assessment method based on data feedback according to claim 5, characterized in that: The method for constructing the orthopedic rehabilitation assessment multi-data model includes: The bone imaging data is used as the input of the first single data model of orthopedic rehabilitation assessment, and the rehabilitation status label of the bone imaging data is used as the output of the first single data model of orthopedic rehabilitation assessment; The electromyographic feedback data is used as the input of the second single data model of orthopedic rehabilitation evaluation, and the rehabilitation status label of the electromyographic feedback data is used as the output of the second single data model of orthopedic rehabilitation evaluation; Constructing a retraining loss function for the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment, and retraining the first single data model of orthopedic rehabilitation assessment and the second single data model of orthopedic rehabilitation assessment based on the retraining loss function to obtain a multi-data model of orthopedic rehabilitation assessment; The orthopedic rehabilitation assessment multi-data model includes: Labelgc=SVMa(Dg); Wherein, Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model of orthopedic rehabilitation assessment, Dg is the bone imaging data, and SVMa is the classifier in the first single data model of orthopedic rehabilitation assessment; Labelac=SVMg(Da); Wherein, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model of orthopedic rehabilitation evaluation, Da is the electromyographic feedback data, and SVMa is the classifier in the second single data model of orthopedic rehabilitation evaluation.

7. A rehabilitation assessment method based on data feedback according to claim 6, characterized in that: The loss function of the first single data model of orthopedic rehabilitation assessment is: ； Where La is the output loss value of the first single data model for orthopedic rehabilitation evaluation, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation evaluation, and Labelar is the true value of the rehabilitation status label of the electromyographic feedback data. for and The L2 norm between .

8. A rehabilitation assessment method based on data feedback according to claim 7, characterized in that: The loss function of the second single data model for orthopedic rehabilitation assessment is: ； Where Lg is the output loss value of the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgr is the true value of the rehabilitation status label of the bone imaging data. for and The L2 norm between .

9. A rehabilitation assessment method based on data feedback according to claim 8, characterized in that: The loss function of the orthopedic rehabilitation evaluation multi-data model is: ； Wherein, Lag is the output loss of the multi-data model for orthopedic rehabilitation assessment, Labela is the rehabilitation status label of the electromyographic feedback data output by the first single data model for orthopedic rehabilitation assessment, Labelac is the rehabilitation status label of the electromyographic feedback data output by the second single data model for orthopedic rehabilitation assessment, Labelg is the rehabilitation status label of the bone imaging data output by the second single data model for orthopedic rehabilitation assessment, and Labelgc is the rehabilitation status label of the bone imaging data output by the first single data model for orthopedic rehabilitation assessment. for and The L2 norm between for and The L2 norm between .

10. An orthopedic rehabilitation system, characterized in that: A rehabilitation assessment method based on data feedback applied to claims 1-9, the system comprising: A data acquisition unit, used to collect electromyographic feedback data and bone imaging data of orthopedic rehabilitation assessment subjects; A data processing unit, used for labeling the electromyographic feedback data and the bone image data to obtain rehabilitation status labels for the electromyographic feedback data and the orthopedic image data; A deep learning unit, for training a classifier using a first single-modal data set consisting of electromyographic feedback data and corresponding rehabilitation status labels to obtain a first single-data model for orthopedic rehabilitation assessment, and a second single-modal data set consisting of skeletal imaging data and corresponding rehabilitation status labels to obtain a second single-data model for orthopedic rehabilitation assessment; The transfer learning unit is used to use the first single-modality data set and the second single-modality data set to perform transfer learning on the second single-data model of orthopedic rehabilitation assessment and the first single-data model of orthopedic rehabilitation assessment, respectively, to obtain a multi-data model of orthopedic rehabilitation assessment.

Images