Portable rehabilitation training and intelligent evaluation system

Abstract

The invention discloses a portable rehabilitation training and intelligent evaluation system, and relates to the technical field of computer-aided diagnosis. The portable rehabilitation training and intelligent evaluation system is mainly composed of wearable motion sensing equipment, a wireless receiver, a portable suitcase, a network communication module and rehabilitation training and intelligent rehabilitation evaluation software. The system has rehabilitation training, rehabilitation evaluation and remote rehabilitation cloud platform interaction functions. According to a patient rehabilitation evaluation result and clinical data, an individualized rehabilitation prescription model is established, a patient individualized rehabilitation prescription is generated, multi-source feature fusion is carried out through a feature fusion formula provided by the invention, a virtual reality scene interactive game is controlled by fused features, and individualized and targeted rehabilitation training is carried out; an artificial intelligence quantitative evaluation model is established through deep learning, a nonlinear dynamic tool is designed to visualize an evaluation result, and a quantitative evaluation result consistent with a clinical Fugl-Meyer scale and an intelligent staging result consistent with a Brunnstrom staging are realized.

Description

Technical Field

[0001] The invention relates to the technical field of computer-aided diagnosis, in particular to a portable rehabilitation training and intelligent assessment system. Background Art

[0002] With the improvement of people's living standards and the aging of the population in many countries, cardiovascular and cerebrovascular diseases have become the world's number one health killer, and stroke is the leading cause of death and lasting disability. Stroke has the characteristics of high morbidity, high mortality, high disability rate, and high burden. It has become the second leading cause of death and the third leading cause of disability in the world. According to statistics from the World Health Organization, as of 2019, there are about 25 million existing stroke patients in my country, and 3.5 million new cases occur each year. In addition, more than 50 million people suffer from brain trauma every year in the world. It is estimated that by 2030, nerve damage caused by brain trauma will still be the most important cause of disability caused by neurological diseases (2 to 3 times higher than Alzheimer's disease or cerebrovascular disease). Among them, about 75% of patients have varying degrees of limb motor dysfunction, which brings a heavy burden to society and families.

[0003] Based on the currently recognized theory of neural remodeling, clinical research and practice have confirmed that after timely and effective rehabilitation training, about 90% of patients can recover a certain degree of daily movement and self-care ability. In clinical practice, 0-6 months is called the golden rehabilitation period, but rehabilitation is a long process and most of them should last for more than 1 year.

[0004] However, the number of stroke patients in my country is very large, rehabilitation medical resources are very scarce, and the distribution of rehabilitation resources in different regions and between urban and rural areas is extremely uneven. These conditions result in stroke patients being unable to receive effective rehabilitation after discharge, missing the golden rehabilitation period, and even leading to lifelong disability. Therefore, there is a great demand for rehabilitation of hospital patients.

[0005] Currently, there is no portable rehabilitation training and intelligent assessment system for discharged patients. The core technical difficulty lies in: how to accurately achieve Fugl-Meyer quantitative assessment results consistent with clinical assessment results and intelligent staging results consistent with Brunnstrom staging results, and how to conduct individualized and targeted rehabilitation training.

[0006] The portable rehabilitation training and intelligent rehabilitation assessment system proposed in the present invention is based on wearable technology, Internet of Things technology, and artificial intelligence technology to realize intelligent rehabilitation training, accurate and objective rehabilitation assessment, and individualized rehabilitation training. It is an intelligent rehabilitation system urgently needed for clinical and discharge rehabilitation. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] In view of the shortcomings of existing technologies, wearable intelligent sensing devices are used to collect human rehabilitation movement data with high precision, and an artificial intelligence quantitative assessment model is established to design individualized rehabilitation training prescriptions, thereby realizing intelligent quantitative rehabilitation assessment and individualized interactive rehabilitation training.

[0009] (II) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a portable rehabilitation training and intelligent assessment system, including a portable rehabilitation training system and an intelligent rehabilitation assessment system. The portable rehabilitation training system is mainly composed of a wearable motion sensor device, a wireless receiver, a portable suitcase, a network communication module and rehabilitation training and intelligent rehabilitation assessment software.

[0011] Preferably, the wearable motion sensing device mainly includes: a wearable motion capture device, a wearable rehabilitation glove, and a wearable electromyography ring;

[0012] The wearable motion capture device includes 4 wireless nodes, which are worn on the upper arm, forearm, thigh, and calf of the affected limb, or 5 wireless nodes, which are worn on the upper arm, forearm, thigh, calf, and chest of the affected limb. Each node has a built-in high-precision 9-axis motion sensor and wireless module to capture the motion trajectory, acceleration, angular velocity, angle, joint range of motion, and other limb motion information of the upper and lower limbs in real time during rehabilitation; the chest sensor is used to collect trunk motion information, and the data is used to correct the patient's limb compensation during rehabilitation assessment.

[0013] The rehabilitation glove is worn on the affected hand. Each finger has a built-in bending sensor and a micro vibration motor. The back of the hand is a data acquisition and wireless module with a built-in high-precision 9-axis motion sensor. The rehabilitation glove is used for finger and wrist motion signal collection, human-computer interaction rehabilitation training and physical therapy during the rehabilitation process.

[0014] The wearable EMG ring is worn on the forearm and / or upper arm, and has a built-in 6- or 8-channel surface EMG acquisition module and wireless module to collect surface EMG signals of the upper limb muscles in real time for movement intention recognition, muscle strength assessment, and motor function assessment of more serious patients such as Brunnstrom stage I or II patients.

[0015] Preferably, the wireless receiver is a ZigBee, Wifi or Bluetooth receiver, connected to a computer via USB, or a built-in Wifi or Bluetooth module in a computer or smart TV. The wireless receiver acts as a wireless communication master, sending data acquisition and control instructions to each wearable motion sensor device for synchronous data acquisition.

[0016] Preferably, the portable suitcase has built-in charging slots for wearable motion sensor devices, rehabilitation training props, and a tablet computer. The portable suitcase is directly connected to a household socket or connected through a computer adapter to power the portable suitcase and charge the wearable motion sensor device. The rehabilitation training props include a rehabilitation training ball and / or a gripping cylinder. The tablet computer is optional. The tablet computer is fixed to the inner side of the upper cover of the portable suitcase through the slots, and the suitcase is opened to face the patient. The opening angle of the upper cover of the suitcase can be adjusted and fixed.

[0017] Preferably, the network communication module is connected to the Internet via a wired and / or 4G / 5G wireless module and / or Wifi module, and the rehabilitation training data and results are uploaded to the rehabilitation cloud platform for review and analysis by doctors in the hospital, who then provide feedback and guidance on the patient's rehabilitation and confirm rehabilitation prescriptions.

[0018] Preferably, the rehabilitation training and intelligent rehabilitation assessment software provides versions for various platforms, and can work on optional embedded tablet computers, ordinary home computers, smart TVs and smart phones, and other smart display terminals;

[0019] Preferably, the training and intelligent assessment system and individualized rehabilitation method are aimed at patients with limb movement dysfunction caused by stroke, brain trauma, etc., based on the high-precision collection of human rehabilitation movement data by wearable intelligent sensing equipment, and the establishment of an artificial intelligence quantitative assessment model, and the design of individualized rehabilitation training prescriptions, to achieve intelligent quantitative rehabilitation assessment and individualized interactive rehabilitation training.

[0020] Patients perform rehabilitation training according to the individualized rehabilitation exercise prescription updated on the software interface and based on standard training videos.

[0021] Preferably, the software preprocesses and extracts features from the collected patient rehabilitation movement data, controls daily life scenario interactive games to conduct targeted feedback training, automatically generates a rehabilitation assessment report after all rehabilitation movements are completed, and uploads the data and results to the rehabilitation cloud platform through the network communication module.

[0022] Preferably, the limb motion information collected by the multi-source motion capture system cannot directly drive (control) the tasks (elements) in the virtual reality scene. Starting from the original limb motion information, a series of processing is required before the interface between the multi-source motion capture system and the virtual reality scene task can be connected. Specifically, the limb motion signal collected by the multi-source motion capture system needs to be preprocessed to eliminate the interference and influence of system noise; then, the original motion signal needs to be feature extracted. The present invention extracts features from the amplitude, direction, dynamic energy, smoothness and force of the motion signal and surface electromyography signal of the patient's rehabilitation action. The detailed definition of each feature and its corresponding physical meaning are listed in Table 1; finally, it is necessary to use a multi-source fusion algorithm to fuse the features of multiple dimensions to obtain fused features, and the fused features are used to drive (control) the tasks and roles in the virtual reality scene and provide human-computer interaction feedback;

[0023] The feature fusion method is as follows: first, the feature weight of each wearable device is determined by the information entropy of each feature, and then the fused feature value is calculated by the features and weights, and quantified to 0-100 points or other range scores;

[0024] Information entropy is used to describe the probability and uncertainty of possible events of information sources. Assume that the random variable X has n different values, denoted as x1, x2, ..., x n , the probability of each value is p1,p2,...,p n , then the information entropy H of random variable X can be expressed as:

[0025]

[0026] The size of H describes the uncertainty of the random variable. The larger the value, the greater the uncertainty of the random variable, and the smaller the value, the smaller the uncertainty of the random variable.

[0027] Conditional entropy H(X|Y) represents the uncertainty of random variable X under the condition of known random variable Y. The expression of H(X|Y) is:

[0028]

[0029] For random variable X, there are K different features D: D1, D2, ..., D K , when the features are determined, there are K conditional entropies: H(X|D1),H(X|D2),...,H(X|D k ), and define the information gain as:

[0030] G(X|D i )=H(X)-H(X|D i ) (3)

[0031] Information gain can be understood as the reduction in the uncertainty of random variable X due to the determination of a certain feature of random variable X. If the information gain is larger, it means that the feature D i The greater the information entropy determined;

[0032] In the rehabilitation action assessment, this paper takes the score S (S = 0, 1, 2) of a single action as a random variable and the extracted features as the features D of the random variable: D1, D2, ..., D K , through the patient rehabilitation action data samples collected in the early stage, the information gain G corresponding to each feature extracted from the rehabilitation action is calculated: G(X|D1),G(X|D2),...,G(X|D k ), the smaller the information gain of the feature, the smaller the weight of determining the S value, and vice versa. Finally, the k information gains are normalized to determine the weight δ of each feature: δ1,δ2,...,δ k , whose expression is:

[0033]

[0034] The characteristics after multi-source information fusion are:

[0035]

[0036] Among them, δ is the feature weight, f is the feature value after normalization to 0-100, n is the number of positive contribution score features, and m is the number of negative contribution score features.

[0037] The present invention discloses a portable rehabilitation training and intelligent assessment system, which has the following beneficial effects:

[0038] This portable rehabilitation training and intelligent assessment system is aimed at patients with limb movement dysfunction caused by stroke, brain trauma, etc. It collects human rehabilitation movement data with high precision based on wearable intelligent sensing equipment, establishes an artificial intelligence quantitative assessment model, designs individualized rehabilitation training prescriptions, and realizes intelligent quantitative rehabilitation assessment and individualized interactive rehabilitation training.

[0039] This portable rehabilitation training and intelligent assessment system and individualized rehabilitation method collects patient rehabilitation training data based on wearable devices and transmits it wirelessly to embedded computers, personal computers or smart TVs. The software analyzes and processes the data to establish intelligent interaction and feedback between virtual reality task scenes and patient rehabilitation training. Through machine learning and artificial intelligence algorithms, a mapping model is established with clinical Fugl-Meyer and other assessment scales to achieve quantitative rehabilitation assessment of limb movement.

[0040] This portable rehabilitation training and intelligent assessment system and individualized rehabilitation method develops rehabilitation training and assessment human-computer interaction virtual reality interactive games based on individualized exercise prescriptions and daily life scenarios, which enhances the enthusiasm of patients for rehabilitation training and the accuracy and objectivity of rehabilitation assessment. At the same time, it enables patients to understand their rehabilitation training process and rehabilitation results more conveniently and intuitively, and provides doctors with an auxiliary intelligent rehabilitation training and rehabilitation assessment system and equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 The structure of the present invention is a diagram of an individualized rehabilitation training and intelligent rehabilitation assessment method deployed on a cloud platform;

[0043] Figure 2 A diagram of the individualized rehabilitation training and intelligent rehabilitation assessment method deployed locally by the present invention;

[0044] Figure 3 This is a model diagram of the individualized rehabilitation prescription of the present invention;

[0045] Figure 4 This is a flow chart of the virtual reality scenario interactive rehabilitation of the present invention;

[0046] Figure 5 This is a diagram of the rehabilitation assessment model and method of the present invention;

[0047] Figure 6 A flow chart for designing visual quantitative assessment parameters of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are 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.

[0049] The portable rehabilitation training and intelligent rehabilitation assessment system is mainly composed of wearable motion sensor equipment, wireless receiver, portable suitcase, network communication module, rehabilitation training and intelligent rehabilitation assessment software, etc.

[0050] Wearable motion sensing devices mainly include: wearable motion capture devices, wearable rehabilitation gloves, and wearable electromyography rings. The wearable motion capture device includes 4 wireless nodes, which are worn on the upper arm, forearm, thigh, and calf of the affected limb respectively. Each node has a built-in high-precision 9-axis motion sensor and wireless module, which is used to capture the motion trajectory, acceleration, angular velocity, angle and joint range of motion of the upper and lower limbs in real time during rehabilitation. The rehabilitation gloves are worn on the affected hand, and each finger has a built-in bending sensor and a micro vibration motor. The back of the hand is a data acquisition and wireless module with a built-in high-precision 9-axis motion sensor. The rehabilitation gloves are used for finger and wrist motion signal acquisition, interaction and physical therapy during rehabilitation. The wearable electromyography ring is worn on the forearm or upper arm, and has a built-in 6 or 8-channel surface electromyography acquisition module and wireless module to collect surface electromyography signals of the upper limb muscle groups in real time, which is used for movement intention recognition, muscle strength assessment, and motor function assessment of patients with more severe symptoms.

[0051] The wireless receiver is a ZigBee, Wifi or Bluetooth receiver, which is connected to the computer via USB, or it can be a built-in Wifi or Bluetooth module in the computer or smart TV.

[0052] The portable suitcase has built-in charging slots for wearable motion sensor devices, rehabilitation training props, and a tablet computer (optional). The portable suitcase is directly connected to a household socket or connected through a computer adapter to power the portable suitcase and charge the wearable motion sensor devices. The rehabilitation training props include rehabilitation training balls and / or gripping cylinders. The tablet computer is optional and is fixed to the inner side of the upper cover of the portable suitcase through the slots, so that the suitcase is opened to face the patient.

[0053] The network communication module is connected to the Internet via a wired and / or 4G / 5G wireless module and / or Wifi module, and the rehabilitation training data and results are uploaded to the rehabilitation cloud platform for review and analysis by doctors in the hospital, who also provide feedback, guidance and rehabilitation prescriptions.

[0054] Rehabilitation training and intelligent rehabilitation assessment software provides versions for various platforms and can work on optional embedded tablets, ordinary home computers, smart TVs, and smart phones. Patients perform rehabilitation training according to the individualized rehabilitation exercise prescriptions updated on the software interface and based on standard training videos. The software preprocesses and extracts features from the collected patient rehabilitation exercise data, controls interactive games in daily life scenarios to conduct targeted feedback training, and automatically generates a rehabilitation assessment report after all rehabilitation movements are completed, and uploads the data and results to the rehabilitation cloud platform through the network communication module.

[0055] Individualized rehabilitation training and intelligent rehabilitation assessment method, Example 1: Individualized rehabilitation training and intelligent rehabilitation assessment method such as Figure 1As shown, after the patient puts on the device and opens the software, the software automatically updates the individualized rehabilitation prescription. The patient follows the individualized rehabilitation prescription and follows the standard action video for situational interactive rehabilitation training (patients who need physical therapy on their hands also receive vibration physical therapy by micro motors in the finger area during the rehabilitation training). After the training is completed, the rehabilitation assessment results and assessment report are automatically generated, and all data and results are uploaded to the cloud platform, where an individualized prescription model is deployed to generate individualized rehabilitation strategies and prescriptions. After confirmation by the doctor, a new individualized rehabilitation prescription is formed.

[0056] Example 2: Individualized rehabilitation training and intelligent rehabilitation assessment method Figure 2 As shown, after the patient puts on the device and opens the software, the software automatically updates the individualized rehabilitation prescription. The patient performs situational interactive rehabilitation training according to the individualized rehabilitation prescription and follows the standard action video (patients who need physical therapy on their hands also receive vibration physical therapy by micro motors in the finger area during the rehabilitation training). After the training is completed, the rehabilitation assessment results and assessment report are automatically generated. The rehabilitation assessment results are used to generate individualized rehabilitation strategies and prescriptions through an individualized prescription model. After the prescription is updated, a new individualized rehabilitation prescription is formed.

[0057] Individualized prescription model Figure 3 As shown, based on the rehabilitation assessment results, combined with clinical diagnosis data and rehabilitation rules, the rehabilitation prescription library and exercise prescriptions issued by clinicians are used as labels, and a machine learning model is established through support vector machines or other conventional machine learning algorithms. The machine learning model is then corrected through logical optimization. For example: if there are problems with the upper limbs, upper limb training is strengthened, and if there are problems with movement coordination, coordinated movement training is strengthened. Finally, the corrected machine learning model provides individualized rehabilitation strategies and rehabilitation paradigms.

[0058] The limb motion information collected by the multi-source motion capture system cannot directly drive (control) the tasks (elements) in the virtual reality scene. Starting from the original limb motion information, it needs to go through a series of processing before the interface between the multi-source motion capture system and the virtual reality scene task can be connected.

[0059] Specifically, firstly, the limb motion signals collected by the multi-source motion capture system need to be preprocessed to eliminate the interference and influence of system noise; then, the original motion signals need to be feature extracted. The present invention extracts features from the amplitude, direction, dynamic energy, smoothness and force of the motion signals and surface electromyography signals of the patient's rehabilitation movements. The detailed definition of each feature and its corresponding physical meaning are listed in Table 1; finally, the multi-source fusion algorithm needs to be used to fuse the features of multiple dimensions to obtain fused features, and the fused features are used to drive (control) the tasks and roles in the virtual reality scene and provide human-computer interaction feedback;

[0060] Virtual reality scenario interactive rehabilitation process Figure 4 shown.

[0061] The feature fusion method is as follows: first, the feature weight of each wearable device is determined by the information entropy of each feature, and then the fused feature value is calculated by the features and weights, and quantified to 0-100 points;

[0062] Information entropy is used to describe the probability and uncertainty of possible events of information sources. Assume that the random variable X has n different values, denoted as x1, x2, ..., x n , the probability of each value is p1,p2,...,p n , then the information entropy H of random variable X can be expressed as:

[0063]

[0064] The size of H describes the uncertainty of the random variable. The larger the value, the greater the uncertainty of the random variable, and the smaller the value, the smaller the uncertainty of the random variable.

[0065] Conditional entropy H(X|Y) represents the uncertainty of random variable X under the condition of known random variable Y. The expression of H(X|Y) is:

[0066]

[0067] For random variable X, there are K different features D: D1, D2, ..., D K , when the features are determined, there are K conditional entropies: H(X|D1),H(X|D2),...,H(X|D k ). Define information gain as:

[0068] G(X|D i )=H(X)-H(X|D i ) (3)

[0069] Information gain can be understood as the reduction in the uncertainty of random variable X due to the determination of a certain feature of random variable X. The larger the information gain, the greater the reduction in the uncertainty of random variable X due to the determination of a certain feature of random variable X. i The greater the information entropy determined.

[0070] In the rehabilitation action assessment, this paper takes the score S (S = 0, 1, 2) of a single action as a random variable and the extracted features as the features D of the random variable: D1, D2, ..., D K ;

[0071] Through the patient rehabilitation action data samples collected in the early stage, the information gain G corresponding to each feature extracted from the rehabilitation action is calculated: G(X|D1), G(X|D2),..., G(X|Dk ), the smaller the information gain of the feature, the smaller the weight of determining the S value, and vice versa. Finally, the k information gains are normalized to determine the weight δ of each feature: δ1,δ2,...,δ k , whose expression is:

[0072]

[0073] The characteristics after multi-source information fusion are:

[0074]

[0075] Among them, δ is the feature weight, f is the feature value after normalization to 0-100, n is the number of positive contribution score features, and m is the number of negative contribution score features.

[0076] Table 1 Extracted features and their physical meanings

[0077]

[0078] Note: (1) The x in the table is the original signal; N is the length of x.

[0079] Based on the information fusion of multi-node and multi-source wearable devices, a one-dimensional feature vector of rehabilitation action is obtained, and a one-dimensional CNN network is proposed to establish an intelligent rehabilitation assessment model (such as Figure 5 shown).

[0080] The steps to establish the intelligent rehabilitation assessment model are as follows:

[0081] 1) The original training data is first preprocessed by windowing and framing: a complete training cycle lasts 100 seconds, divided into 100 short cycles, each lasting 1 second;

[0082] 2) Feature parameter extraction: Extract multiple parameters such as resonance peak, fundamental frequency, fundamental frequency perturbation, average energy, etc. for each short-period signal;

[0083] 3) Build a one-dimensional CNN input signal model, and concatenate 10 short-period signals into a long-period signal, and repeat the above step (2);

[0084] 4) Data normalization, which is used to eliminate the influence of scale and dimension on calculation;

[0085] 5) Establishment of training data standard score mapping: This process is carried out under the guidance of clinicians, according to the Fugl-Meyer, Brunnstrom and other commonly used clinical scales, to give each training action a score (full score 100) and stage (mainly based on the Brunnstrom scale, stage 1-6);

[0086] 6) Establishment of one-dimensional convolutional network. This process is the same as the establishment process of one-dimensional action phased ResNet network mentioned above.

[0087] 7) Network training and deep learning;

[0088] 8) Action evaluation and model self-learning: Different from the one-dimensional action recognition model, in action evaluation, the prediction process is the evaluation process (giving action scores), and the data is included in the training set to update the training model, so that the model has the function of self-learning;

[0089] 9) Establish an evaluation model for the evaluation effect of one-dimensional CNN network: Evaluate the established one-dimensional CNN network based on parameters such as consistency between the CNN network evaluation results and clinical results, stability of the evaluation, specificity, and optimize and improve the network accordingly.

[0090] Visual quantitative evaluation parameter design based on nonlinear dynamics tools such as Figure 6 As shown in the figure, based on the establishment of an artificial intelligence quantitative rehabilitation evaluation model, nonlinear dynamics tools such as Poincare difference scatter plots, maximized Lyapunov exponents, and distribution entropy are used to design visualized quantitative assessment parameters, so that the quantitative evaluation model can give visualized quantitative scores and intelligent staging results for training movements.

[0091] A portable rehabilitation training and intelligent rehabilitation assessment system is proposed, which is mainly composed of a wearable motion sensor device, a wireless receiver, a portable suitcase, a network communication module, and rehabilitation training and intelligent rehabilitation assessment software. It is used for discharged patients and clinical rehabilitation training and rehabilitation assessment.

[0092] A method for generating individualized rehabilitation training prescriptions is proposed. Based on rehabilitation assessment results, clinical diagnosis data, and rehabilitation rules, an artificial intelligence individualized rehabilitation prescription model is established and modified to generate individualized rehabilitation prescriptions for individualized and targeted rehabilitation training of patients.

[0093] A multi-source information feature fusion method is proposed. The feature weight is determined by the information entropy of each feature of multi-node multi-source information, and the feature fusion calculation formula is extracted. The fused feature value is calculated by the features and weights.

[0094] Through wearable device data and artificial intelligence algorithms, a deep learning intelligent rehabilitation assessment model is established to achieve rehabilitation assessment and intelligent staging consistent with clinical Fugl-Meyer, Brunnstrom and other scales. Visual quantitative assessment parameters of nonlinear dynamics tools are designed to obtain visualized quantitative rehabilitation assessment results.

[0095] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0096] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. Portable rehabilitation training and intelligent assessment system, characterized by: It is mainly composed of wearable motion sensor devices, wireless receivers, portable suitcases, network communication modules and rehabilitation training and intelligent rehabilitation assessment software.

2. The portable rehabilitation training and intelligent assessment system according to claim 1, characterized in that: The wearable motion sensing device mainly includes: a wearable motion capture device, a wearable rehabilitation glove, and a wearable electromyography ring; The wearable motion capture device includes 4 wireless nodes, which are worn on the upper arm, forearm, thigh, and calf of the affected limb, or 5 wireless nodes, which are worn on the upper arm, forearm, thigh, calf, and chest of the affected limb; each node has a built-in high-precision 9-axis motion sensor and wireless module to capture the motion trajectory, acceleration, angular velocity, angle, joint range of motion and other limb movement information of the upper and lower limbs during rehabilitation in real time; the chest sensor is used to collect trunk motion information, and the data is used to correct the patient's limb compensation during rehabilitation assessment. The rehabilitation glove is worn on the affected hand. Each finger has a built-in bending sensor and a micro vibration motor. The back of the hand is a data acquisition and wireless module with a built-in high-precision 9-axis motion sensor. The rehabilitation glove is used for finger and wrist motion signal collection, human-computer interaction rehabilitation training and physical therapy during the rehabilitation process. The wearable EMG ring is worn on the forearm and / or upper arm, and has a built-in 6- or 8-channel surface EMG acquisition module and wireless module to collect surface EMG signals of the upper limb muscles in real time for movement intention recognition, muscle strength assessment, and motor function assessment of more serious patients such as Brunnstrom stage I or II patients.

3. The portable rehabilitation training and intelligent assessment system according to claim 1, characterized in that: The wireless receiver is a ZigBee, Wifi or Bluetooth receiver, which is connected to the computer via USB, or can be a built-in Wifi or Bluetooth module in the computer or smart TV. The wireless receiver acts as a wireless communication master, sending data collection and control instructions to each wearable motion sensor device for synchronous data collection.

4. The portable rehabilitation training and intelligent assessment system according to claim 1, characterized in that: The portable suitcase has built-in charging slots for wearable motion sensor devices, rehabilitation training props, and a tablet computer. The portable suitcase is directly connected to a household socket or other power source or is connected through a computer adapter to power the portable suitcase and charge the wearable motion sensor device. The rehabilitation training props include a rehabilitation training ball and / or a gripping cylinder. The tablet computer is optional. The tablet computer is fixed to the inner side of the upper cover of the portable suitcase through the slots. When the suitcase is opened, it faces the patient. The opening angle of the upper cover of the suitcase can be adjusted and fixed.

5. The portable rehabilitation training and intelligent assessment system according to claim 1, characterized in that: The network communication module is connected to the Internet via a wired and / or 4G / 5G wireless module and / or Wifi module, and the rehabilitation training data and results are uploaded to the rehabilitation cloud platform for review and analysis by doctors in the hospital, who then provide feedback and guidance on the patient's rehabilitation and confirm rehabilitation prescriptions.

6. The portable rehabilitation training and intelligent assessment system according to claim 1, characterized in that: Rehabilitation training and intelligent rehabilitation assessment software provides various platform versions, which can work on optional embedded tablet computers, ordinary home computers, smart TVs, smart phones and other smart display terminals; Patients perform rehabilitation training according to the individualized rehabilitation exercise prescription updated on the software interface and based on standard training videos.

7. The portable rehabilitation training and intelligent assessment system according to claim 6, characterized in that: The software pre-processes and extracts features from the collected patient rehabilitation movement data, controls daily life scenario interactive games for targeted feedback training, automatically generates a rehabilitation assessment report after all rehabilitation movements are completed, and uploads the data and results to the rehabilitation cloud platform through the network communication module.

8. The individualized rehabilitation method according to claim 1, characterized in that: Including individualized rehabilitation prescriptions and human-computer interactive feedback control. The method for generating individualized rehabilitation prescriptions is: based on the rehabilitation assessment results, combined with clinical diagnostic data and rehabilitation rules, using the rehabilitation prescription library and the exercise prescriptions issued by clinicians as labels, a machine learning model is established through support vector machines or other conventional machine learning algorithms, and the machine learning model is corrected through logical optimization, such as: if there are problems with the upper limbs, strengthen upper limb training, and if there are problems with movement coordination, strengthen coordinated movement training. Finally, the corrected machine learning model gives individualized rehabilitation strategies and rehabilitation paradigms. The human-computer interaction feedback control method is as follows: the limb motion information collected by the multi-source motion capture system cannot directly drive (control) the tasks (elements) in the virtual reality scene. Starting from the original limb motion information, a series of processing is required before the interface between the multi-source motion capture system and the virtual reality scene task can be connected. Specifically, the limb motion signal collected by the multi-source motion capture system needs to be preprocessed to eliminate the interference and influence of system noise; then, the original motion signal needs to be feature extracted. The present invention extracts features from the amplitude, direction, dynamic energy, smoothness and force of the motion signal and surface electromyography signal of the patient's rehabilitation action. The detailed definition of each feature and its corresponding physical meaning are listed in Table 1; finally, a multi-source fusion algorithm needs to be used to fuse the features of multiple dimensions to obtain fused features, and the fused features are used to drive (control) the tasks and roles in the virtual reality scene and provide human-computer interaction feedback.

9. The intelligent quantitative rehabilitation assessment method according to any one of claims 1 to 7, characterized in that: The feature weight of each wearable device is determined by the information entropy of each feature. Then, the fused feature values ​​are calculated through features and weights, and the score of a single rehabilitation action is finely quantified to 0-100 points or other score ranges; Information entropy is used to describe the probability and uncertainty of possible events of information sources. Assume that the random variable X has n different values, recorded as x1, x2, ..., x n , the probability of each value is p1,p2,...,p n , then the information entropy H of random variable X can be expressed as: The size of H describes the uncertainty of the random variable. The larger the value, the greater the uncertainty of the random variable, and the smaller the value, the smaller the uncertainty of the random variable. Conditional entropy H(X|Y) represents the uncertainty of random variable X under the condition of known random variable Y. The expression of H(X|Y) is: For random variable X, there are K different features D: D1, D2, ..., D K , when the features are determined, there are K conditional entropies: H(X|D1),H(X|D2),...,H(X|D k ), and define the information gain as: G(X|D i )=H(X)-H(X|D i ) (3) Information gain can be understood as the reduction in the uncertainty of random variable X due to the determination of a certain feature of random variable X. If the information gain is larger, it means that the feature D i The greater the information entropy determined; In the rehabilitation action assessment, this paper takes the score S (S = 0, 1, 2) of a single action as a random variable and the extracted features as the features D of the random variable: D1, D2, ..., D K , through the patient rehabilitation action data samples collected in the early stage, the information gain G corresponding to each feature extracted from the rehabilitation action is calculated: G(X|D1),G(X|D2),...,G(X|D k ), the smaller the information gain of the feature, the smaller the weight of determining the S value, and vice versa. Finally, the k information gains are normalized to determine the weight δ of each feature: δ1,δ2,...,δ k , whose expression is: The characteristics after multi-source information fusion are: Among them, δ is the feature weight, f is the feature value after normalization to 0-100, n is the number of positive contribution score features, and m is the number of negative contribution score features.

10. The intelligent quantitative rehabilitation assessment method according to any one of claims 1 to 7, characterized in that: Through wearable motion sensor device data and artificial intelligence algorithms, a machine learning or deep learning intelligent rehabilitation assessment model is established to achieve rehabilitation assessment and intelligent staging consistent with clinical Fugl-Meyer, Brunnstrom and other scales. The visual quantitative assessment parameters of the nonlinear dynamics tool are designed to obtain visual quantitative rehabilitation assessment results.

Images