Stroke patient hand obstacle grading evaluation system and method of wearable ultrasonic equipment

Abstract

The invention provides a stroke patient hand obstacle grading evaluation system and method based on wearable ultrasonic equipment. The system comprises a force sensor module, the wearable ultrasonic equipment, a human-computer interaction interface and an upper computer. The force sensor module comprises a force sensor and a plurality of shells with different sizes for integrating the force sensor, and is used for collecting grip data of a patient; the wearable ultrasonic equipment comprises an equipment host and four ultrasonic probes provided with four-channel ultrasonic transducer arrays, and is used for collecting hand movement ultrasonic data of a patient; the human-computer interaction interface is used for displaying operation guidance, real-time ultrasonic signals, a force curve and the degree of hand dysfunction and supporting parameter presetting; the upper computer is connected with the force sensor module, the wearable ultrasonic device and the man-machine interaction interface, receives and processes the collected grip strength data and the hand movement ultrasonic data, and evaluates the hand dysfunction level of a patient. According to the method, the hand dysfunction level can be evaluated more objectively and accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of medical-industrial cross-rehabilitation, and in particular to a wearable ultrasonic device for grading and evaluating hand disorders in stroke patients and a method thereof. Background Art

[0002] Stroke is a common and serious neurological disease that usually causes hand dysfunction in patients, which in turn seriously affects their ability to take care of themselves in daily life and their quality of life. With the increasing incidence of stroke every year, how to scientifically and effectively evaluate hand dysfunction has become an important issue in clinical treatment. At present, the clinical evaluation of hand dysfunction in stroke patients mainly relies on the doctor's experience and some traditional scale evaluations. Although these methods can provide certain references, they often have shortcomings such as strong subjectivity, insufficient quantification, and lack of accuracy, making it difficult to fully reflect the actual functional status of patients.

[0003] In recent years, wearable devices, as an emerging technology, have been widely used in the medical field, especially in the assessment of hand dysfunction in stroke patients. Wearable devices can provide doctors with objective reference by collecting patients' motion data in real time. However, most wearable devices on the market currently rely on technologies such as surface electromyography and inertial sensors. Although these technologies can capture surface muscle activity information, they are difficult to deeply detect the specific conditions inside the muscles, resulting in limited assessment accuracy. In addition, they lack the ability to coordinate the analysis of biomechanical parameters and muscle morphology, and cannot fully reflect the patient's neural control coordination.

[0004] How to develop a multimodal evaluation system that can integrate muscle morphology and biomechanical parameters to more objectively and accurately evaluate hand dysfunction in stroke patients remains a scientific research problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to propose a wearable ultrasound device for stroke patients hand impairment grading assessment system and method to solve the problems of strong subjectivity, low quantification, poor device portability, etc. in the prior art. Stroke patients collect ultrasound signals and force signals through type A ultrasound equipment and force sensors, input them into the assessment model to synchronously predict gestures and force categories, and thereby assess the hand impairment level of stroke patients, which can more objectively and accurately assess the hand dysfunction level.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] On the one hand, the present invention provides a wearable ultrasound device for stroke patients' hand disorder grading assessment system, including a force sensor module, a wearable ultrasound device, a human-computer interaction interface and a host computer;

[0008] The force sensor module includes a force sensor and a plurality of housings of different sizes for integrating the force sensors, so as to simulate different grip force requirements; the force sensor is used to collect grip force data of patients with different grip force requirements;

[0009] The wearable ultrasound device is a wearable A-type ultrasound device, comprising a device host and four ultrasound probes provided with a four-channel ultrasound transducer array, and the four ultrasound probes are fixed to the surface projection area of ​​the target muscle of the patient's forearm, and the device host and the four ultrasound probes are used to collect the patient's hand movement ultrasound data;

[0010] The human-computer interaction interface is used to display operation instructions, real-time ultrasonic signals, force curves, and the degree of hand dysfunction, and supports parameter presets;

[0011] The host computer is respectively connected to the force sensor module, the wearable ultrasonic device, and the human-computer interaction interface, receives and processes the collected grip force data and hand movement ultrasonic data, and evaluates the hand dysfunction level of the patient.

[0012] Preferably, the four ultrasound probes are fixed to the surface projection area of ​​the target muscles of the patient's forearm by elastic straps, and the target muscles include the extensor digitorum, flexor carpi ulnaris, extensor carpi radialis longus, and flexor digitorum superficialis.

[0013] Preferably, the force sensor module adopts cylindrical shells of different diameters for integrating the force sensor, and the cylindrical shells of different diameters include cylindrical shells with diameters of 60 mm and 90 mm, and the force sensor adopts a cylindrical structure with a diameter of 25 mm; wherein different cylindrical diameters are used to simulate different grip strength requirements, only the force sensor represents that the patient can completely complete the action according to the instructions, integrating the force sensor in a cylindrical shell with a diameter of 60 mm represents that the patient can partially complete the action according to the instructions, and integrating the force sensor 2 in a cylindrical shell with a diameter of 90 mm represents that the patient cannot complete the instructed action at all.

[0014] Preferably, the parameter preset includes a preset of a gesture and a force guidance curve, and the force guidance curve adopts a step curve or a sine curve.

[0015] On the other hand, the present invention provides a wearable ultrasound device for stroke patients hand disorder grading assessment method, the assessment method is based on any of the above-mentioned wearable ultrasound device for stroke patients hand disorder grading assessment system, specifically comprising the following steps:

[0016] Step 1: The tester wears the wearable ultrasound device and uses different gestures to grasp force sensor modules of different sizes with different force levels according to the operation instructions of the human-computer interaction interface, and simultaneously collects type A ultrasound data and grip force data;

[0017] Step 2: The host computer performs preprocessing operations on the collected data;

[0018] Step 3: Extract features from the preprocessed A-type ultrasound data;

[0019] Step 4: Build a fine gesture classification model to classify fine gestures according to the input A-type ultrasound data feature values; wherein the fine gesture category is a combination of grasping gesture and gesture completion, and each gesture completion degree under each grasping gesture is regarded as a type of fine gesture; for each fine gesture category, a force level classification model or regression model is established to identify the grasping force level; the grip force data obtained in steps 2 and 3 and the A-type ultrasound data feature values ​​are divided into a training set and a test set, and the grip force data of different gestures, different gesture completion degrees and different force levels are used as labels for the A-type ultrasound data feature values ​​to perform model training;

[0020] Step 5: The patient wears the wearable ultrasound device and performs different gestures according to the operation instructions of the human-computer interaction interface. The wearable ultrasound device synchronously collects type A ultrasound data, and the host computer completes preprocessing and feature extraction operations; the extracted features are input into the trained fine gesture classification model to obtain the patient's fine gesture category; based on the classification model or regression model of the force level corresponding to the identified fine gesture category, the patient's force level category is obtained; the completion of the patient's hand movements and the degree of muscle strength damage are measured by the identified fine gesture category and force level category, so as to rate the patient's hand dysfunction, and output the patient's hand dysfunction degree, which is divided into normal, mild, moderate, severe, and complete disorder.

[0021] Preferably, the step 2 specifically includes:

[0022] When the force-guiding curve adopts a step curve, for the collected step force curve data, the data before n seconds and after n seconds for each force level are discarded, and only the middle time is retained as valid data; when the force-guiding curve adopts a sine curve, the collected sine force curve data is retained;

[0023] For type A ultrasonic signals, m data points are discarded at the head and tail of the raw data collected in each frame of each channel, and the data points in the middle of the raw data are used as valid data points for the next step of preprocessing, which includes Gaussian filtering, Hilbert transform and logarithmic compression.

[0024] Preferably, the feature extraction of the preprocessed ultrasonic data specifically includes: for the preprocessed ultrasonic signal, the valid data points of each frame are cut according to a preset fixed window size, and the time domain or frequency domain feature calculation is performed on the data of each window obtained after cutting, and after the feature extraction is completed, the principal component analysis method is used for dimensionality reduction.

[0025] Preferably, the time domain feature calculation includes the calculation of mean standard deviation MSD, root mean square RMS, and mean value ME; the frequency domain feature calculation includes the calculation of spectrum feature S(k), mean frequency MF, and root mean square frequency RMSF.

[0026] Preferably, the gesture completion degree is quantified according to different sizes of the force sensor module and divided into three completion degrees: complete bending, able to bend but not fully, and completely unable to bend; the fine gesture classification model adopts a support vector machine SVM.

[0027] Preferably, the force classification model uses an SVM classification model to classify force levels, and the labels 1, 2, 3, and 4 of the force classification model correspond to 0%, 10%, 20%, and 30% of the maximum voluntary contraction force, respectively; the regression model uses a GPR Gaussian process regression model to regress the force level, and the label of the regression model is the true force value corresponding to the sample.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention provides a stroke patient hand disorder grading assessment system and method based on a wearable ultrasound device, which uses a wearable ultrasound device and a force sensor module to collect signals, and classifies the patient's hand movements through the collected signals to judge the patient's hand disorder condition, thereby avoiding the influence of subjective factors on the assessment results, and can objectively, accurately and efficiently assess the degree of hand dysfunction in stroke patients, providing a scientific basis for clinical diagnosis and rehabilitation treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a schematic diagram of a stroke patient hand disorder grading assessment system according to the present invention;

[0031] Figure 2 A schematic diagram of the position of the wearable ultrasound device probe of the present invention;

[0032] Figure 3 Schematic diagram of the force sensor of the present invention and the integration of the force sensor with cylindrical shells of different sizes;

[0033] Figure 4 It is a schematic diagram of the force guiding curve type of the present invention;

[0034] Figure 5 This is a schematic diagram of the process of grading and evaluating hand disorders in stroke patients according to the present invention;

[0035] Figure 6 It is a framework diagram of the algorithm model adopted by the present invention;

[0036] Figure 7 It is a schematic diagram of the original ultrasonic signal and the signal after Gaussian filtering, Hilbert transform and logarithmic compression processing of the present invention;

[0037] Figure 8 The feature is a schematic diagram of ultrasonic data feature extraction of the present invention.

[0038] In the figure, 1- wearable ultrasound device; 2- force sensor module; 3- human-computer interaction interface; 4- host computer; 5- extensor digitorum; 6- flexor carpi ulnaris; 7- extensor carpi radialis longus; 8- flexor digitorum superficialis. DETAILED DESCRIPTION

[0039] The following is combined with Figure 1-8 , the technical solution of the present invention is specifically described.

[0040] refer to Figure 1 , the present invention proposes a wearable ultrasound device for stroke patients hand disorder grading assessment system, including a force sensor module 2, a wearable ultrasound device 1, a human-computer interaction interface 3 and a host computer 4;

[0041] The force sensor module 2 includes a force sensor and a plurality of housings of different sizes for integrating the force sensors, so as to simulate different grip force requirements; the force sensor 2 is used to collect grip force data of patients with different grip force requirements;

[0042] The wearable ultrasound device 1 is a wearable A-type ultrasound device, which includes a device host and four ultrasound probes (working frequency is 2.5 MHZ) provided with a four-channel ultrasound transducer array. The four ultrasound probes are fixed to the surface projection area of ​​the target muscle of the patient's forearm. Under the action of the high-frequency electrical signal of the device host, the piezoelectric crystal in the probe vibrates and emits ultrasound waves. The ultrasound waves propagate inside the human muscle. The probe receives the ultrasound echo signal and converts it into an electrical signal, which is transmitted to the device host for processing. The device host and the four ultrasound probes are used to collect the patient's hand movement ultrasound data;

[0043] The human-machine interaction interface 3 is used to display operation instructions, real-time ultrasonic signals, force curves, and the degree of hand dysfunction, and supports parameter presets;

[0044] The host computer 4 is respectively connected to the force sensor module 2, the wearable ultrasonic device 1, and the human-computer interaction interface 3, and the platform is developed based on the C++ language. The collected grip strength data and hand movement ultrasonic data are received and processed, and the patient's hand dysfunction level is evaluated by classifying and evaluating the patient's fine gestures and force levels.

[0045] In this embodiment, the four ultrasound probes are fixed to the patient's forearm target muscle surface projection area (about 3 / 5 of the forearm length from the wrist joint) by elastic straps. Figure 2 The target muscles include extensor digitorum 5, flexor carpi ulnaris 6, extensor carpi radialis longus 7, and flexor digitorum superficialis 8; after putting on the device, place the arm on a fixed bracket to keep the arm movement unchanged; apply medical ultrasonic coupling agent between the probe and the skin to reduce the influence of air and external environment and optimize signal quality.

[0046] In this embodiment, the force sensor module 2 uses cylindrical shells with different diameters for integrating force sensors. Figure 3 The cylindrical shells with different diameters include cylindrical shells with diameters of 60 mm and 90 mm, and the force sensor adopts a cylindrical structure with a diameter of 25 mm; wherein different cylindrical diameters are used to simulate different grip force requirements, only the force sensor represents that the patient can completely complete the action according to the instruction, integrating the force sensor in a cylindrical shell with a diameter of 60 mm represents that the patient can partially complete the action according to the instruction, and integrating the force sensor 2 in a cylindrical shell with a diameter of 90 mm represents that the patient cannot complete the instructed action at all.

[0047] In this embodiment, the parameter presets include the presets of gestures and force guidance curves, see Figure 4 The force guidance curve adopts a step curve or a sine curve. Before the experiment begins, the required grasping gesture and force guidance curve are added manually to evaluate the degree of fit between the prediction curve and the guidance curve to assess the percentage of the patient's output muscle force that can reach the normal muscle force, so as to measure the degree of damage to the hand muscle strength:

[0048] The present invention also proposes a wearable ultrasound device for stroke patients hand disorder grading assessment method, the assessment method is based on any of the above-mentioned wearable ultrasound device for stroke patients hand disorder grading assessment system, reference Figure 5 , specifically including the following steps:

[0049] Step 1: The tester wears the wearable ultrasound device and uses different gestures to grasp force sensor modules of different sizes with different force levels according to the operation instructions of the human-computer interaction interface, and simultaneously collects type A ultrasound data and grip force data;

[0050] Step 2: The host computer 4 performs preprocessing operations on the collected data;

[0051] Step 3: Extract features from the preprocessed A-type ultrasound data;

[0052] Step 4: Build a fine gesture classification model to classify fine gestures according to the input A-type ultrasound data feature values; wherein the fine gesture category is a combination of grasping gesture and gesture completion, and each gesture completion degree under each grasping gesture is regarded as a type of fine gesture; for each fine gesture category, a force level classification model or regression model is established to identify the grasping force level; the grip force data obtained in steps 2 and 3 and the A-type ultrasound data feature values ​​are divided into a training set and a test set, and the grip force data of different gestures, different gesture completion degrees and different force levels are used as labels for the A-type ultrasound data feature values ​​to perform model training;

[0053] Step 5: The patient wears the wearable ultrasound device and performs different gestures according to the operation instructions of the human-computer interaction interface. The wearable ultrasound device synchronously collects type A ultrasound data, and the host computer completes preprocessing and feature extraction operations; the extracted features are input into the trained fine gesture classification model to obtain the patient's fine gesture category; based on the classification model or regression model of the force level corresponding to the identified fine gesture category, the patient's force level category is obtained; the completion of the patient's hand movements and the degree of muscle strength damage are measured by the identified fine gesture category and force level category, so as to rate the patient's hand dysfunction, and output the patient's hand dysfunction degree, which is divided into normal, mild, moderate, severe, and complete disorder.

[0054] In this embodiment, the step 1 specifically includes:

[0055] The maximum voluntary contraction force MVC of the tester is obtained through the test, and the force level is divided according to the maximum voluntary contraction force MVC of the tester, for example, 0% MVC, 10% MVC, 20% MVC, 30% MVC; this is used as a force guidance curve to guide the tester to perform grasping actions using different force levels, thereby obtaining corresponding type A ultrasound data and grip strength data as a training data set.

[0056] In this embodiment, the step 2 specifically includes:

[0057] In order to eliminate the impact of instantaneous changes when force levels are switched, when the force guidance curve adopts a step curve, for the collected step force curve data, the first n seconds and the last n seconds of each force level are discarded, and only the middle time is retained as valid data; when the force guidance curve adopts a sine curve, the collected sinusoidal force curve data is retained.

[0058] by Figure 4For example, when the force guidance curve adopts a step curve, the data of the first 1 second and the last 1 second are discarded for each force level, and only the middle 3 seconds are retained as valid data; when the force guidance curve adopts a sine curve, the entire 20 seconds of data are retained;

[0059] For type A ultrasound signals, m data points are discarded at the head and tail of the raw data collected in each channel and each frame, and the data points in the middle of the raw data are used as valid data points for the next step of preprocessing. Figure 7 The next step of preprocessing includes Gaussian filtering, Hilbert transform and logarithmic compression.

[0060] For type A ultrasound signals, the four probes correspond to the four channels of the ultrasound machine, and the size of the data matrix generated per second is 40×1000. Ultrasonic signals will produce invalid information in the surface layer of the skin and deep muscle. In order to avoid the interference of these invalid data, 20 data points at the head and tail of the raw data collected in each frame of each channel will be discarded, and the 960 data points in the middle of the data will be used as valid data points for the next step of preprocessing.

[0061] In this embodiment, the feature extraction of the pre-processed ultrasonic data specifically includes: for the pre-processed ultrasonic signal, the valid data points of each frame are cut according to the preset fixed window size, the time domain or frequency domain feature calculation is performed on the data of each window obtained after cutting, and the principal component analysis method is used for dimensionality reduction after the feature extraction is completed. The time domain feature calculation includes the calculation of the mean standard deviation MSD, the root mean square RMS, and the mean value ME; the frequency domain feature calculation includes the calculation of the spectrum feature S(k), the mean frequency MF, and the root mean square frequency RMSF. Among them, the grip strength data is the label of the ultrasonic data predicting the force level, and no feature extraction is required.

[0062] Among them, after preprocessing, each frame of the ultrasonic signal has 960 valid data points. According to the fixed window size w = 20, the total number of output feature data is k = 960 / w = 48. There is no overlap between the windows. The time domain or frequency domain feature calculation is performed on the data of each window. Figure 8 ; In this embodiment, the gesture completion degree is quantified according to the different sizes of the force sensor module, and is divided into three completion degrees: complete bending, able to bend but not fully, and completely unable to bend; specifically, the gesture completion degree is quantified according to the cylinder diameters of 25, 60, and 90 (unit: mm), respectively regarded as cur1, cur2, and cur3 (cur1 represents complete bending, cur2 represents able to bend but not fully, and cur3 represents completely unable to bend); The fine gesture classification model adopts support vector machine SVM.

[0063] In this embodiment, the force classification model uses the SVM classification model to classify the force level, and the labels 1, 2, 3, and 4 of the force classification model correspond to 0%, 10%, 20%, and 30% of the maximum voluntary contraction force, respectively; the regression model uses the GPR Gaussian process regression model to regress the force level, and the label of the regression model is the true force value corresponding to the sample.

[0064] Take the finger movement assessment in the FMA scale (total score 14 points) as an example:

[0065] For finger movements: 0 points for failure to complete, 1 point for incomplete completion, and 2 points for complete and active completion;

[0066] For grip strength level: 0 points if the force level fits within the range of 0-10% MVC, 1 point if it fits within the range of 10%-20%, and 2 points if it fits within the range of 20%-30% MVC;

[0067] The patient completes several actions in sequence. After each action is completed, the upper computer displays the score and records the score of each action. The level of the disorder is assessed based on the total score, as shown in the following table:

[0068]

[0069] In summary, the stroke patient hand disorder grading assessment system and method based on wearable ultrasound equipment proposed in the present invention has the following advantages:

[0070] 1. Strong objectivity: Wearable ultrasound equipment is used to obtain muscle morphological structure information, and force sensors are used to obtain muscle force information, avoiding the influence of subjective factors on the evaluation results;

[0071] 2. High degree of quantification: Extracting multiple parameters reflecting muscle morphology and mechanics can more comprehensively and accurately assess the degree of hand dysfunction;

[0072] 3. Good portability: The wearable design makes it easy for patients to wear and use, and is suitable for various scenarios such as home and community;

[0073] 4. High evaluation efficiency: Use machine learning algorithms to automatically analyze data and quickly output evaluation results, thereby improving evaluation efficiency.

[0074] This not only has important clinical significance for improving the treatment effect of stroke patients, but also provides a new direction for the development of personalized medicine. Therefore, the research and development of new technologies and methods for the assessment of hand dysfunction has become an important task that cannot be ignored in medical research and clinical treatment.

[0075] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A wearable ultrasound device for stroke patients hand disorder grading assessment system, characterized in that: It includes a force sensor module, a wearable ultrasound device, a human-computer interaction interface and a host computer; The force sensor module includes a force sensor and a plurality of housings of different sizes for integrating the force sensors, so as to simulate different grip force requirements; the force sensor is used to collect grip force data of patients with different grip force requirements; The wearable ultrasound device is a wearable type A ultrasound device, comprising a device host and four ultrasound probes provided with an ultrasound transducer array, and the four ultrasound probes are fixed to the projection area of ​​the target muscle surface of the patient's forearm, and the device host and the four ultrasound probes are used to collect the patient's hand movement ultrasound data; The human-computer interaction interface is used to display operation instructions, real-time ultrasonic signals, force curves, and the degree of hand dysfunction, and supports parameter presets; The host computer is respectively connected to the force sensor module, the wearable ultrasonic device, and the human-computer interaction interface, receives and processes the collected grip force data and hand movement ultrasonic data, and evaluates the hand dysfunction level of the patient.

2. The wearable ultrasound device-based stroke patient hand impairment grading assessment system according to claim 1, characterized in that: The four ultrasonic probes are fixed to the surface projection area of ​​the target muscles of the patient's forearm through elastic bands. The target muscles include the extensor digitorum, the flexor carpi ulnaris, the extensor carpi radialis longus, and the flexor digitorum superficialis.

3. The wearable ultrasound device-based stroke patient hand impairment grading assessment system according to claim 1, characterized in that: The force sensor module adopts cylindrical shells of different diameters for integrating force sensors, and the cylindrical shells of different diameters include cylindrical shells with diameters of 60mm and 90mm. The force sensor adopts a cylindrical structure with a diameter of 25mm. The different cylindrical diameters are used to simulate different grip strength requirements. Only the force sensor represents that the patient can complete the action completely according to the instructions. Integrating the force sensor in a cylindrical shell with a diameter of 60mm represents that the patient can partially complete the action according to the instructions. Integrating the force sensor 2 in a cylindrical shell with a diameter of 90mm represents that the patient cannot complete the instructed action at all.

4. The wearable ultrasound device-based stroke patient hand impairment grading assessment system according to claim 1, characterized in that: The parameter presets include presets of gestures and force guidance curves, and the force guidance curve adopts a step curve or a sine curve.

5. A wearable ultrasound device for evaluating hand impairment in stroke patients, characterized in that: The evaluation method is implemented based on the stroke patient hand disorder grading evaluation system of the wearable ultrasound device according to any one of claims 1 to 4, and specifically comprises the following steps: Step 1: The tester wears the wearable ultrasound device and uses different gestures to grasp force sensor modules of different sizes with different force levels according to the operation instructions of the human-computer interaction interface, and simultaneously collects type A ultrasound data and grip force data; Step 2: The host computer performs preprocessing operations on the collected data; Step 3: Extract features from the preprocessed A-type ultrasound data; Step 4: Build a fine gesture classification model to classify fine gestures according to the input A-type ultrasound data feature values; wherein the fine gesture category is a combination of grasping gesture and gesture completion, and each gesture completion degree under each grasping gesture is regarded as a type of fine gesture; for each fine gesture category, establish a force level classification model or regression model to identify the grasping force level according to the input A-type ultrasound data feature values; divide the grip force data obtained in steps 2 and 3 and the A-type ultrasound data feature values ​​into a training set and a test set, and use the grip force data of different gestures, different gesture completion degrees and different force levels as labels for the A-type ultrasound data feature values ​​to perform model training; Step 5: The patient wears the wearable ultrasound device and performs different gestures according to the operation instructions of the human-computer interaction interface. The wearable ultrasound device synchronously collects type A ultrasound data, and the host computer completes preprocessing and feature extraction operations; the extracted features are input into the trained fine gesture classification model to obtain the patient's fine gesture category; based on the classification model or regression model of the force level corresponding to the identified fine gesture category, the patient's force level category is obtained; the completion of the patient's hand movements and the degree of muscle strength damage are measured by the identified fine gesture category and force level category, so as to rate the patient's hand dysfunction, and output the patient's hand dysfunction degree, which is divided into normal, mild, moderate, severe, and complete disorder.

6. The method for grading hand impairment of stroke patients using a wearable ultrasound device according to claim 5, characterized in that: The step 2 specifically includes: When the force-guiding curve adopts a step curve, for the collected step force curve data, the data before n seconds and after n seconds for each force level are discarded, and only the middle time is retained as valid data; when the force-guiding curve adopts a sine curve, the collected sine force curve data is retained; For type A ultrasonic signals, m data points are discarded at the head and tail of the raw data collected in each frame of each channel, and the data points in the middle of the raw data are used as valid data points for the next step of preprocessing, which includes Gaussian filtering, Hilbert transform and logarithmic compression.

7. The method for grading hand impairment of stroke patients using a wearable ultrasound device according to claim 6, characterized in that: The feature extraction of the preprocessed ultrasonic data specifically includes: for the preprocessed ultrasonic signal, the valid data points of each frame are cut according to a preset fixed window size, the time domain or frequency domain feature calculation is performed on the data of each window obtained after cutting, and after the feature extraction is completed, the principal component analysis method is used for dimensionality reduction.

8. The method for grading hand impairment of stroke patients using a wearable ultrasound device according to claim 7, characterized in that: The time domain feature calculation includes the calculation of the mean standard deviation MSD, the root mean square RMS, and the mean value ME; the frequency domain feature calculation includes the calculation of the spectrum feature S(k), the mean frequency MF, and the root mean square frequency RMSF.

9. The method for grading hand impairment of stroke patients using a wearable ultrasound device according to claim 5, characterized in that: The gesture completion degree is quantified according to different sizes of the force sensor module and is divided into three completion degrees: complete bending, able to bend but not fully, and completely unable to bend; the fine gesture classification model adopts a support vector machine SVM.

10. The method for grading hand impairment of stroke patients using a wearable ultrasound device according to claim 5, characterized in that: The force classification model uses the SVM classification model to classify the force level. The labels 1, 2, 3, and 4 of the force classification model correspond to 0%, 10%, 20%, and 30% of the maximum voluntary contraction force, respectively. The regression model uses the GPR Gaussian process regression model to regress the force level. The label of the regression model is the true force value corresponding to the sample.

Images