Auxiliary robot control method and system for improving walking function of lower limbs after stroke

Abstract

The invention provides an auxiliary robot control method and system for improving the walking function of lower limbs after stroke, and relates to the field of health data processing, and the system comprises a data collection module which is used for collecting multi-modal test data of a patient, comprising key head cortex point location test data, bilateral joint coordination test data and walking balance obstacle test data; the function analysis module is used for determining lower limb dysfunction characteristics of the patient according to the multi-modal test data of the patient; the data acquisition module is also used for acquiring lower limb characteristics of the patient; the decision making module is used for determining walking function rehabilitation training parameters including training frequency, period duration, walking speed and gait tracks according to the lower limb dysfunction characteristics of the patient, the lower limb characteristics and the sample database; the training control module is used for controlling the auxiliary robot to assist the patient in lower limb walking function training according to the walking function rehabilitation training parameters, and the advantage of improving the lower limb walking function rehabilitation training effect after stroke is achieved.

Description

Technical Field

[0001] This invention relates to the field of health data processing, and in particular to an assistive robot control method and system for improving lower limb walking function after stroke. Background Technology

[0002] Stroke is an acute cerebrovascular disease caused by the sudden rupture or blockage of blood vessels in the brain, resulting in brain tissue damage. It includes ischemic stroke (cerebral infarction) and hemorrhagic stroke (cerebral hemorrhage, subarachnoid hemorrhage, etc.). It is characterized by high incidence, high disability rate, high mortality rate, and high recurrence rate. Stroke can lead to motor dysfunction such as hemiplegia, balance disorders, gait abnormalities, and postural abnormalities. Long-term rehabilitation training is necessary to restore lower limb function.

[0003] In existing technologies, manual weight-bearing gait training involves suspending the patient on a movable platform with a weight-bearing device. This reduces the load on the lower limbs while maintaining balance, requiring two or more therapists or rehabilitation therapists to stand around the patient and guide them through various manual techniques to perform gait-like movements. This manual weight-bearing training is characterized by short training time, low efficiency, and high workload. The training effect is closely related to the skill and experience of the therapist or rehabilitation therapist, making it difficult to maintain precise consistency in rehabilitation movements. Therefore, assistive robots have been developed, but they still require the assistance of a physician for rehabilitation training, resulting in low treatment efficiency. Furthermore, the intensity of training cannot be set for patients with different conditions, failing to meet the rehabilitation needs of diverse patients.

[0004] Therefore, there is a need to provide assistive robot control methods and systems for improving lower limb walking function after stroke, in order to enhance the rehabilitation training effect of lower limb walking function after stroke. Summary of the Invention

[0005] This invention provides an assistive robot control system for improving lower limb walking function after stroke, comprising: a data acquisition module for acquiring multimodal test data of the patient, wherein the multimodal test data includes key head cortical point test data, bilateral joint coordination test data, and walking balance disorder test data; a functional analysis module for determining the patient's lower limb functional impairment characteristics based on the patient's multimodal test data; the data acquisition module is also used to acquire the patient's lower limb characteristics; a decision-making module for determining the patient's walking function rehabilitation training parameters based on the patient's lower limb functional impairment characteristics, lower limb characteristics, and a sample database, wherein the walking function rehabilitation training parameters include training frequency, cycle duration, walking speed, and gait trajectory; and a training control module for controlling the assistive robot to assist the patient in lower limb walking function training based on the walking function rehabilitation training parameters.

[0006] Furthermore, the data acquisition module collects test data from key scalp cortex points of the patient, including: determining multiple test scalp cortex points; determining multiple test movements; for each test movement, acquiring EEG signals from multiple test scalp cortex points of multiple test patients during the test movement, wherein any two test patients have different lower limb functional impairments; based on the EEG signals from multiple test scalp cortex points of multiple test patients during the test movement, determining the β-wave power vector corresponding to the test movement for each test patient; and calculating the β-wave power vector corresponding to the test movement based on the β-wave power vector corresponding to the test movement for each test patient. Beta wave power difference value; based on the beta wave power difference value corresponding to each test movement, determine multiple key test movements from multiple test movements; based on the EEG signals of multiple test scalp cortex points of multiple test patients during each key test movement, determine multiple left key scalp cortex points and multiple right key scalp cortex points; based on the multiple left key scalp cortex points and multiple right key scalp cortex points, collect the key scalp cortex point test data of the patients, wherein the key scalp cortex point test data of the patients include the EEG signal of each left key scalp cortex point and the EEG signal of each right key scalp cortex point.

[0007] Furthermore, the data acquisition module acquires bilateral joint coordination test data of the patient, including: determining key adjacent motor segments based on the patient's key head cortex point test data; and obtaining continuous relative phase curves of each pair of key adjacent motor segments on both sides of the patient.

[0008] Furthermore, the data acquisition module collects the patient's walking balance disorder test data, including: determining multiple key gait cycle percentiles; collecting the patient's joint kinematic characteristics in one gait cycle; and extracting multiple key joint angles on the left and right sides of the patient corresponding to each key gait cycle percentile from the patient's joint kinematic characteristics in one gait cycle, based on the multiple key gait cycle percentiles.

[0009] Furthermore, the data acquisition module determines the patient's lower limb dysfunction characteristics based on the patient's multimodal test data, including: calculating the lateralization index based on the EEG signals of each key left cephalic cortex point and each key right cephalic cortex point; calculating the bilateral joint coordination index based on the continuous relative phase curves of each pair of key adjacent motor segments on both sides of the patient; and calculating the balance index based on multiple key joint angles on the left and right sides corresponding to each key gait cycle percentile. The patient's lower limb dysfunction characteristics include the lateralization index, the bilateral joint coordination index, and the balance index.

[0010] Furthermore, the data acquisition module acquires the patient's lower limb features, including: acquiring images of the patient's lower limbs; and determining the patient's bone length based on the images of the patient's lower limbs.

[0011] Furthermore, the sample database includes multiple rehabilitation training samples, wherein the rehabilitation training samples include the lower limb dysfunction characteristics, lower limb characteristics, and gait function rehabilitation training parameters of the patients in the rehabilitation training samples; the decision determination module determines the gait function rehabilitation training parameters of the patients based on the lower limb dysfunction characteristics, lower limb characteristics, and the sample database, including: grouping the multiple rehabilitation training sample patients according to the lower limb dysfunction characteristics and lower limb characteristics of the patients in the rehabilitation training samples, generating grouping results; and determining the gait function rehabilitation training parameters of the patients based on the grouping results and the lower limb dysfunction characteristics and lower limb characteristics of the patients.

[0012] Furthermore, based on the lower limb dysfunction characteristics and lower limb features of the rehabilitation training sample patients, multiple rehabilitation training sample patients are grouped to generate grouping results, including: calculating the lower limb feature similarity between any two rehabilitation training sample patients; performing a first grouping of multiple rehabilitation training sample patients based on the lower limb feature similarity between any two rehabilitation training sample patients to determine multiple first sample groups; for each first sample group, calculating the lower limb dysfunction feature similarity between any two rehabilitation training sample patients included in the first sample group, and performing a second grouping of multiple rehabilitation training sample patients included in the first sample group based on the lower limb dysfunction feature similarity between any two rehabilitation training sample patients included in the first sample group to determine multiple second sample groups.

[0013] Furthermore, the system also includes: a heart rate monitoring module, used to acquire the patient's heart rate data during the process of the assistive robot assisting the patient in lower limb walking function training; the decision-making module is also used to determine whether to adjust parameters based on the patient's heart rate data, and if so, to adjust the patient's walking function rehabilitation training parameters based on the patient's heart rate data, generating adjusted walking function rehabilitation training parameters; the training control module is also used to control the assistive robot to assist the patient in lower limb walking function training based on the adjusted walking function rehabilitation training parameters.

[0014] This invention provides an assistive robot control method for improving lower limb walking function after stroke, comprising: collecting multimodal test data of the patient, wherein the multimodal test data includes key head cortical point test data, bilateral ankle joint coordination test data, and balance disorder test data; determining the patient's lower limb functional impairment characteristics based on the patient's multimodal test data; collecting the patient's lower limb characteristics; determining the patient's walking function rehabilitation training parameters based on the patient's lower limb functional impairment characteristics, lower limb characteristics, and a sample database, wherein the walking function rehabilitation training parameters include training frequency, cycle duration, walking speed, and gait trajectory; and controlling the assistive robot to assist the patient in lower limb walking function training based on the walking function rehabilitation training parameters.

[0015] Compared with existing technologies, the assistive robot control method and system for improving lower limb walking function after stroke provided by this invention have at least the following beneficial effects: By comprehensively quantifying the characteristics of lower limb dysfunction through multimodal data (cerebral cortical activation, joint coordination, and balance ability), the limitations of single indicators are overcome, providing a scientific basis for personalized training. Training parameters (frequency / duration / speed / trajectory) are intelligently matched with the patient's individual characteristics and sample database to achieve dynamic adaptation of training intensity to the patient's recovery stage, thereby improving rehabilitation efficiency. Attached Figure Description

[0016] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein: Figure 1 These are schematic diagrams of the auxiliary robot shown in some embodiments of this specification; Figure 2 This is a schematic diagram of a module of an assistive robot control system for improving lower limb walking function after stroke, according to some embodiments of this specification. Figure 3 This is a schematic diagram of the process for collecting test data of key scalp cortex points of a patient, as shown in some embodiments of this specification; Figure 4 This is a flowchart illustrating an assistive robot control method for improving lower limb walking function after stroke, according to some embodiments of this specification.

[0017] In the figure, 111 is the support frame; 112 is the lower limb exoskeleton device. Detailed Implementation

[0018] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0019] The assistive robot control method and system described in this specification for improving lower limb walking function after stroke can be applied to existing assistive robots. This is merely an example. Figure 1 These are schematic diagrams of the auxiliary robot shown in some embodiments of this specification, such as... Figure 1 As shown, the assistive robot includes support frames 111 symmetrically arranged on the left and right sides, and a lower limb exoskeleton device 112 is arranged between the support frames 111. The lower limb exoskeleton device 112 includes a hip support corresponding to the human hip bone. Two thigh supports corresponding to the human thighs are movably connected to both sides of the hip support. A first flexible actuator for adjusting the bending angle between each thigh support and the hip support is arranged between them. A lower leg support corresponding to the human calf is movably connected to the lower end of the thigh support. A second flexible actuator for adjusting the bending angle between the lower leg support and the thigh support is arranged between them. A foot pedal support corresponding to the human foot is movably connected to the lower end of the lower leg support. A third flexible actuator for adjusting the bending angle between the foot pedal support and the lower leg support is arranged between them. The lower limb rehabilitation training robot also includes a control device, which is electrically connected to the first flexible actuator, the second flexible actuator, and the third flexible actuator. The first, second, and third flexible actuators are pneumatic cylinders. Each pneumatic cylinder has three rotatable connections: the first cylinder is rotatably connected to the hip and thigh supports, the second cylinder is rotatably connected to the thigh and calf supports, and the third cylinder is rotatably connected to the calf and foot pedal supports. The left and right sides of the hip support are rotatably connected to corresponding support frames 111 via pivot shafts. A hip fixation strap is provided on the hip support, a thigh fixation strap is provided on the thigh support, a calf fixation strap is provided on the calf support, and a foot fixation strap is provided on the foot pedal support. Each leg's exoskeleton structure allows for adjustment of the hip, knee, and ankle joint's degrees of freedom in flexion and extension. These adjustments utilize a first, second, and third flexible actuator, respectively, making the lower limb rehabilitation robot more similar to normal human lower limb movements. This meets the requirements for flexible actuation, increases operational stability, avoids injury, and achieves better rehabilitation outcomes. The support frame 111 bears most of the patient's weight, reducing pressure on the upper limbs and ensuring complete freedom of movement. Using a control device, different walking function rehabilitation training parameters can be used to assist patients in lower limb walking function training.

[0020] Understandable. Figure 1 The assistive robot shown is for illustrative purposes only and does not limit the scope of application of the assistive robot control methods and systems for improving lower limb walking function after stroke described in this specification.

[0021] Figure 2 This is a schematic diagram of a module of an assistive robot control system for improving lower limb walking function after stroke, as shown in some embodiments of this specification. Figure 2 As shown, the assistive robot control system for improving lower limb walking function after stroke may include a data acquisition module, a function analysis module, a decision-making module, and a training control module.

[0022] The data acquisition module is used to collect multimodal test data from patients.

[0023] The multimodal test data includes key head cortical point test data, bilateral joint coordination test data, and walking balance disorder test data.

[0024] Figure 3 This is a schematic flowchart illustrating the collection of key scalp cortex test data from a patient according to some embodiments of this specification. In some embodiments, the data acquisition module collects the key scalp cortex test data from the patient, including: Multiple cortical sites for testing can be identified, including those closely related to lower limb motor control, which can be determined based on motor neuroscience. For example, multiple cortical sites can be determined manually. These sites may include left M1 (C3), right M1 (C4), left SMA (FC3), right SMA (FC4), left parietal lobe (CP3), and right parietal lobe (CP4), etc. Identify various test movements, such as single-joint movements of the left lower limb (e.g., ankle dorsiflexion), single-joint movements of the right lower limb (e.g., ankle dorsiflexion), compound movements (e.g., from sitting to standing), and imaginary movements (e.g., imagining walking). For each test exercise, EEG signals from multiple test cortical points on the head of multiple test patients are acquired during the test exercise. The lower limb dysfunction of any two test patients is different. The multiple test patients can be divided into multiple groups, such as a healthy group, a right hemiplegic group, and a left hemiplegic group. The right hemiplegic group and the left hemiplegic group can include stroke patients with different degrees of hemiplegia. Based on the EEG signals from multiple test cortical points on the head of multiple test patients during the test exercise, the β-wave power vector corresponding to the test exercise for each test patient is determined. Based on the β-wave power vector corresponding to the test exercise for each test patient, the β-wave power difference value corresponding to the test exercise is calculated. Based on the difference in β-wave power corresponding to each test motion, several key test motions are determined from a variety of test motions; Based on the EEG signals of multiple test cerebrocortical points from multiple test patients during each key test movement, multiple left-side key cerebrocortical points and multiple right-side key cerebrocortical points were identified. Based on multiple key left scalp cortex points and multiple key right scalp cortex points, test data of key scalp cortex points of the patient were collected. The test data of key scalp cortex points of the patient included EEG signals of each key left scalp cortex point and EEG signals of each key right scalp cortex point.

[0025] Specifically, the β-wave power vector corresponding to the test exercise for the test patient can be determined according to the following procedure: S11. For each test scalp cortex point, the low-frequency drift and high-frequency noise of the EEG signal of the test scalp cortex point are removed by bandpass filtering (0.5-70Hz) to generate a denoised EEG signal of the test scalp cortex point. The denoised EEG signal of the test scalp cortex point is then subjected to fast Fourier transform to obtain the power spectral density. The power integral of the 13-30Hz frequency band is extracted to obtain the β wave power of the test scalp cortex point. S12. The β-wave power of each test point on the head cortex is spliced ​​together to generate the β-wave power vector corresponding to the test patient's test motion.

[0026] For any two test patients, the cosine similarity of the beta-wave power vectors of the corresponding test movements can be calculated. The mean of the cosine similarity of the beta-wave power vectors of the corresponding test movements for any two test patients is obtained. The reciprocal of the mean cosine similarity yields the difference in beta-wave power corresponding to the test movements.

[0027] Test motions with a β-wave power difference value greater than a β-wave power difference value threshold (e.g., 2) can be used as key test motions, where the β-wave power difference value threshold can be determined by human experience or experimental data.

[0028] For each test point on the cephalic cortex, the beta-wave power of the test point during single-joint movement of the left lower limb in each test patient can be determined. The mean beta-wave power of the test point during single-joint movement of the left lower limb in each test patient is calculated to obtain the mean beta-wave power of the corresponding cephalic cortex point in the left lower limb. Similarly, the beta-wave power of the test point during single-joint movement of the right lower limb in each test patient can be determined. The mean beta-wave power of the test point during single-joint movement of the right lower limb in each test patient is calculated to obtain the mean beta-wave power of the corresponding cephalic cortex point in the right lower limb. The mean difference between the mean beta-wave power of the corresponding cephalic cortex point in the right lower limb and the mean beta-wave power of the corresponding left lower limb is calculated to obtain the mean difference of beta-wave power of the corresponding cephalic cortex point.

[0029] The test scalp points where the mean difference of β-wave power is greater than the β-wave power mean difference threshold (e.g., 0.2) and the mean β-wave power of the corresponding right lower limb is greater than the mean β-wave power of the corresponding left lower limb can be used as key scalp points on the right side. The mean difference threshold of β-wave power can be determined by human experience or experimental data.

[0030] The test scalp cortex points where the mean difference of β-wave power is greater than the threshold of the mean difference of β-wave power, and where the mean β-wave power of the corresponding left lower limb is greater than the mean β-wave power of the corresponding right lower limb, can be used as the key scalp cortex points on the left side.

[0031] Multiple key left and right scalp cortex points can also be determined by other means, such as by using human experience to determine multiple key left and right scalp cortex points based on EEG signals from multiple test scalp cortex points of multiple test patients during each key test movement.

[0032] The patient can perform single-joint movements of the left lower limb to acquire EEG signals at each key left scalp cortex location. Similarly, the patient can perform single-joint movements of the right lower limb to acquire EEG signals at each key right scalp cortex location.

[0033] Understandably, by screening cortical sites strongly correlated with lower limb movement based on motor neuroscience and combining this with β-wave power difference analysis, key brain regions with abnormal motor control after stroke can be precisely identified, providing a neurophysiological basis for personalized neurorehabilitation. By calculating β-wave power differences under various motor tasks, key test movements that best reflect lower limb dysfunction are selected, improving the sensitivity and specificity of subsequent data acquisition. Using EEG data from multiple groups of patients (healthy / left hemiplegic / right hemiplegic) ensures that the identified cortical sites and key motor tasks are applicable to patients with different types and degrees of hemiplegia, enhancing the generalization ability of the system for clinical application. By replacing traditional invasive testing with EEG signal analysis, the activation levels of brain regions related to lower limb motor function can be quantitatively assessed non-invasively, reducing patient risk and improving the acceptability of rehabilitation assessments.

[0034] In some embodiments, the data acquisition module acquires bilateral joint coordination test data of the patient, including: Based on the patient's key cephalometric data, key adjacent motor segments are identified, such as the hip-knee segment, knee-ankle segment, and ankle-toe segment. Obtain continuous relative phase curves for each pair of key adjacent motor segments on both sides of the patient.

[0035] Specifically, multiple second samples can be obtained. These second samples include the key beta-wave power vectors for single joint movements of the left lower limb and the right lower limb, as well as key adjacent motor segments. The key adjacent motor segments of the second sample patients can be determined manually or through diagnostic data. These key adjacent motor segments are adjacent motor segments that are prone to incoordination. The key beta-wave power vector for single joint movements of the left lower limb can be composed of the beta-wave power of multiple key left-sided cephalic cortex points, and the key beta-wave power vector for single joint movements of the right lower limb can be composed of the beta-wave power of multiple key right-sided cephalic cortex points.

[0036] Similarly, determine the key β-wave power vector for the corresponding single joint movement of the left lower limb and the key β-wave power vector for the corresponding single joint movement of the right lower limb.

[0037] For each second sample patient, the cosine similarity between the key β-wave power vector of the second sample patient's corresponding left lower limb single joint movement and the patient's corresponding left lower limb single joint movement (referred to as the left β-wave power similarity) can be calculated, and the cosine similarity between the key β-wave power vector of the second sample patient's corresponding right lower limb single joint movement and the patient's corresponding right lower limb single joint movement (referred to as the right β-wave power similarity) can be calculated. The left β-wave power similarity and the right β-wave power similarity are then weighted and summed to obtain the β-wave power similarity between the second sample patient and the patient.

[0038] A second sample patient whose β-wave power similarity is greater than a β-wave power similarity threshold (e.g., 0.6) can be considered a similar second sample patient. Key adjacent motor segments can be determined based on the key adjacent motor segments of the similar second sample patient with the highest β-wave power similarity. Alternatively, duplicate key adjacent motor segments can be removed from the similar second sample patients to obtain the key adjacent motor segments.

[0039] Reflection markers can be set at corresponding positions based on key adjacent motion segments. Images can be acquired using an infrared camera, and joint angles can be calculated using the coordinates of the reflection markers. A Hilbert transform is performed on the angular velocity signal of each joint to obtain an analytical signal. The instantaneous phase angle is calculated based on the analytical signal. For each pair of key adjacent segments, the relative phase is calculated. For example, the relative phase of hip-knee is the instantaneous phase angle of hip minus the instantaneous phase angle of knee. A sliding window (window length = 0.5 seconds, step size = 0.1 seconds) is applied to the relative phase, and the mean value within the window is calculated to reduce instantaneous noise and generate continuous relative phase curves corresponding to key adjacent motion segments.

[0040] Understandably, screening similar second-sample patients using key head cortical point test data (β-wave power vector) and correlating neurophysiological characteristics (brain region activation patterns) with motor segment coordination reveals joint movement incoordination caused by abnormal neural control after stroke (e.g., hip-knee segment phase difference disorder may be caused by reduced β-wave power in the M1 region). Based on the neurological characteristics of similar patients, key adjacent motor segments (e.g., knee-ankle segments) are identified, avoiding misjudgments caused by standardized criteria and improving the accuracy of coordination disorder assessment. Continuous relative phase curves can visually display intersegmental movement delays (e.g., failure of knee flexion after hip extension), providing specific goals for rehabilitation training (e.g., adjusting hip-knee segment phase synchronization).

[0041] In some embodiments, the data acquisition module collects walking balance disorder test data from patients, including: Determine multiple key gait cycle percentiles; Collect joint kinematic characteristics of the patient during one gait cycle; Based on multiple key gait cycle percentiles, multiple key joint angles on the left and right sides of the patient are extracted from the joint kinematic features of the patient in one gait cycle, corresponding to each key gait cycle percentile.

[0042] Specifically, the gait cycle is the complete process of walking from the heel touching the ground on one side to the heel touching the ground again on the same side, and is usually divided into the standing phase (60%) and the swing phase (40%). Percentiles are used to divide the gait cycle into standardized time points.

[0043] For example, examples of each percentile are shown in Table 1.

[0044]

[0045] If the patient's gait cycle is 1.5 seconds, then the 50th percentile corresponds to 0.75 seconds. At this point, the angles of the hip, knee, and ankle joints need to be extracted to determine the joint stability during the mid-standing phase.

[0046] The second sample may also include multiple key gait cycle percentiles and multiple key joints corresponding to the second sample patients. Among them, the key gait cycle percentiles may be gait cycle percentiles that are likely to show walking balance disorders.

[0047] Based on the multiple key gait cycle percentiles and multiple key joints of the similar second sample patient with the highest β-wave power similarity, the multiple key gait cycle percentiles and multiple key joints of the current patient can be determined. Alternatively, the multiple key gait cycle percentiles of the similar second sample patient can be deduplicated to obtain the multiple key gait cycle percentiles of the current patient, and the multiple key joints of the similar second sample patient can also be deduplicated to obtain the multiple key joints of the current patient.

[0048] Understandably, by dividing the gait cycle into multiple percentiles (e.g., 10%, 30%, 50%, 70%, 90%), it is possible to accurately pinpoint periods of high-incidence balance disturbances (e.g., excessive center of gravity shift at 50% of the mid-stance phase), overcoming the limitations of traditional gait analysis that only focuses on discrete events (e.g., heel strike / lift). Simultaneously, key joint angles on both sides (e.g., flexion angles of the hip / knee / ankle joints at 30% of the gait cycle) are extracted, quantifying bilateral movement differences (e.g., a 200ms delay in knee extension on the affected side). Based on joint angle data from consecutive gait cycle percentiles, angle-time dynamic curves can be constructed (e.g., the rate of change of ankle dorsiflexion angle with the gait cycle), identifying balance disturbance-related features (e.g., excessive plantar flexion of the ankle at the end of the swing phase leading to toe dragging). Combined with percentile markers, joint angle stability in specific gait phases (e.g., the single-leg stance phase at 70% of the cycle) can be analyzed, assessing the patient's ability to control center of gravity shift.

[0049] The functional analysis module is used to determine the characteristics of lower limb dysfunction in patients based on their multimodal test data.

[0050] Specifically, it includes: The lateralization index was calculated based on the EEG signals of each key left scalp cortex point and each key right scalp cortex point. The bilateral joint coordination index is calculated based on the continuous relative phase curves of each pair of key adjacent motor segments on both sides of the patient. The balance index is calculated based on the left and right key joint angles corresponding to the percentile of each key gait cycle. The patient's lower limb dysfunction characteristics include the lateralization index, bilateral joint coordination index, and balance index.

[0051] Specifically, for multiple key left scalp cortical points, the β-wave power of each key left scalp cortical point can be calculated based on the EEG signal of each key left scalp cortical point. The mean β-wave power of each key left scalp cortical point can then be calculated to obtain the mean left β-wave power of the patient.

[0052] For multiple key right-sided scalp cortex sites, the β-wave power of each site can be calculated based on the EEG signal of each site. The mean β-wave power of each site can then be calculated to obtain the mean right-sided β-wave power of the patient.

[0053] The lateralization index was calculated based on the mean power of the left and right beta waves of the patient.

[0054] For example, the lateralization index can be calculated using the following formula: , in, The side-effect index, The mean power of the right-side beta wave. This represents the average power of the left-side beta wave.

[0055] Understandably, by using the ratio of the mean power of the left and right β waves, the asymmetry of the brain's neural control over lower limb movement can be quantified, providing an objective basis for the analysis of brain region compensation mechanisms (e.g., LI>0 suggests overactivation of the right cortex).

[0056] For any critical adjacent motion segment, the Euclidean distance between the continuous relative phase curve of the critical adjacent motion segment on the left and the continuous relative phase curve of the critical adjacent motion segment on the right can be calculated as the Euclidean distance corresponding to the critical adjacent motion segment.

[0057] The mean of the Euclidean distances corresponding to each key adjacent motion segment can be calculated to obtain the mean Euclidean distance. The reciprocal of the mean Euclidean distance can be used as the bilateral joint coordination index.

[0058] A standard sample can be obtained, which can be multiple key joint angles on the left and right sides of each key gait cycle percentile corresponding to patients with similar lower limb characteristics to the current patient and without lower limb dysfunction. The balance index can be calculated based on the differences between the multiple key joint angles on the left and right sides of each key gait cycle percentile corresponding to the current patient and those of patients with similar lower limb characteristics and without lower limb dysfunction.

[0059] For example, the balance index can be calculated using the following formula: , in, As a balance index, Let m be the angle of the m-th critical joint to the left of the current patient at the percentile of the nth critical gait cycle. The angle of the m-th critical joint to the left of the percentile of the n-th critical gait cycle in the standard sample. Let m be the angle of the m-th critical joint to the right of the current patient at the percentile of the nth critical gait cycle. The angle of the m-th critical joint to the right of the percentile of the n-th critical gait cycle in the standard sample. The total number of percentiles of critical gait cycles. This represents the total number of critical joints.

[0060] Understandably, for each gait percentile (n) and each joint (m), the difference in left and right joint angles between the patient and the standard sample is calculated to reflect the degree to which the bilateral angles deviate from the standard values. The sum of squares of the differences for all gait percentiles and joints is double-summed to obtain the total difference value. The total difference is then mapped to an index (BI) between 0 and 1 using 1 / (1 + total difference). The smaller the total difference, the closer the BI is to 1, indicating better balance function.

[0061] The data acquisition module is also used to collect the patient's lower limb characteristics.

[0062] Specifically, it includes: Acquire images of the patient's lower limbs; The patient's bone length was determined based on images of the lower limbs.

[0063] Specifically, deep learning models (such as YOLO and U-Net) can be used to identify landmarks in the patient's lower limb images (such as the center of the femoral head, the center of the knee joint, and the center of the ankle joint) to determine the length of each bone in the patient, such as the length of the femur and the length of the tibia.

[0064] The decision-making module is used to determine the rehabilitation training parameters for patients' walking function based on the characteristics of their lower limb dysfunction, lower limb characteristics, and sample database.

[0065] Among them, the parameters for gait function rehabilitation training include training frequency, cycle duration, walking speed, and gait trajectory.

[0066] Specifically, the sample database includes multiple rehabilitation training samples, which include the lower limb dysfunction characteristics, lower limb characteristics, and walking function rehabilitation training parameters of the patients in the rehabilitation training samples.

[0067] In some embodiments, the decision determination module determines the patient's walking function rehabilitation training parameters based on the patient's lower limb dysfunction characteristics, lower limb characteristics, and a sample database, including: Based on the lower limb dysfunction characteristics and lower limb features of patients in the rehabilitation training sample, multiple rehabilitation training sample patients were grouped to generate grouping results; Based on the grouping results and the patients' lower limb dysfunction characteristics and lower limb features, the parameters for rehabilitation training of patients' walking function were determined.

[0068] In some embodiments, multiple rehabilitation training sample patients are grouped according to their lower limb dysfunction characteristics and lower limb features, generating grouping results, including: Calculate the similarity of lower limb features between any two rehabilitation training sample patients; Based on the similarity of lower limb features between any two rehabilitation training sample patients, multiple rehabilitation training sample patients are first grouped to determine multiple first sample groups; For each first sample group, calculate the similarity of lower limb dysfunction characteristics between any two rehabilitation training sample patients included in the first sample group. Based on the similarity of lower limb dysfunction characteristics between any two rehabilitation training sample patients included in the first sample group, perform a second grouping of multiple rehabilitation training sample patients included in the first sample group to determine multiple second sample groups.

[0069] Specifically, for each rehabilitation training sample patient, a skeletal feature sequence can be generated based on the length of each bone in the patient, for example, (L1, L2…L…). n ), where L1 is the length of the first bone, L2 is the length of the second bone, L n Let be the length of the nth bone.

[0070] For any two rehabilitation training sample patients, the cosine similarity of their skeletal feature sequences can be calculated as the lower limb feature similarity. A hierarchical clustering algorithm can then be used to initially group multiple rehabilitation training sample patients based on their lower limb feature similarity, thus determining multiple first sample groups.

[0071] For any two rehabilitation training patients included in the first sample group, the cosine similarity of their lower limb dysfunction features can be calculated as the lower limb dysfunction feature similarity between the two patients. A hierarchical clustering algorithm can then be used to further group the patients in the first sample group based on the lower limb dysfunction feature similarity between any two patients in the first sample group, thus determining multiple second sample groups within the first sample group.

[0072] Referring to the method for calculating the lower limb feature similarity between two rehabilitation training sample patients, the lower limb feature similarity of the current patient with the lower limb feature of each first sample group is calculated. The first sample group with a lower limb feature similarity greater than the lower limb feature similarity threshold (e.g., 0.7) is designated as the similar first sample group. Referring to the method for calculating the lower limb dysfunction feature similarity between two rehabilitation training sample patients, the lower limb dysfunction feature similarity of the current patient with the cluster center of each second sample group included in each similar first sample group is calculated. The similar second sample group with a lower limb dysfunction feature similarity greater than the first lower limb dysfunction feature similarity threshold (e.g., 0.7) is designated as the similar second sample group. The lower limb feature similarity threshold and the first lower limb dysfunction feature similarity threshold can be determined based on human experience or experimental data.

[0073] Calculate the similarity between the current patient's lower limb dysfunction characteristics and the rehabilitation training sample patients included in each similar second sample group. Select the rehabilitation training sample patient with the highest similarity in lower limb dysfunction characteristics as the similar rehabilitation training sample patient, and use the walking function rehabilitation training parameters of the similar rehabilitation training sample patient as the walking function rehabilitation training parameters of the current patient.

[0074] Preferably, patients with lower limb dysfunction feature similarity greater than the second limb dysfunction feature similarity threshold (e.g., 0.9) can be used as similar rehabilitation training sample patients. The walking function rehabilitation training parameters of the current patient are determined by the parameter determination model based on the walking function rehabilitation training parameters of similar rehabilitation training sample patients. The parameter determination model can be a deep learning model.

[0075] If no similar second sample group exists, the patient's walking function rehabilitation training parameters can be determined through manual experience based on the patient's lower limb dysfunction characteristics and lower limb features.

[0076] The training control module is used to control the assistive robot to assist the patient in lower limb walking function training based on the walking function rehabilitation training parameters.

[0077] Training frequency can represent the number of training sessions per week (e.g., 3 times / week), cycle duration can represent the total number of gait cycles in a single training session (e.g., 20 walks), and gait trajectory can include joint angle sequences (e.g., hip joint: 0°→30°→0°, knee joint: 0°→60°→0°).

[0078] Understandably, grouping patients with similar functional impairments and lower limb characteristics ensures that training parameters (such as assistance intensity and training duration) are highly adapted to patient needs, thereby improving rehabilitation efficiency. By combining individual patient characteristics with grouping results, training parameters (such as gait cycle phase adjustment and joint range of motion restrictions) can be dynamically adjusted to avoid a "one-size-fits-all" approach and reduce the risk of secondary injury.

[0079] The system also includes: The heart rate monitoring module is used to acquire the patient's heart rate data during the process of assisting the patient with lower limb walking function training by the assistive robot; The decision-making module is also used to determine whether to adjust parameters based on the patient's heart rate data. If so, it adjusts the patient's walking function rehabilitation training parameters based on the patient's heart rate data and generates the adjusted walking function rehabilitation training parameters. The training control module is also used to control the assistive robot to assist patients in lower limb walking function training based on the adjusted walking function rehabilitation training parameters.

[0080] Specifically, a patient's heart rate threshold can be determined. If the patient's maximum heart rate during lower limb walking function training exceeds the heart rate threshold, parameter adjustments are required. The heart rate threshold can be determined manually or through experimental data. Parameter adjustments can be made by reducing the cycle duration or walking speed.

[0081] Understandably, dynamically monitoring a patient's physiological state through heart rate data allows for timely identification of excessive fatigue or stress responses (such as a heart rate consistently >120 bpm), preventing risks caused by overtraining. Automatically adjusting training intensity based on heart rate changes (e.g., reducing assist speed, shortening single training sessions) ensures the training load matches the patient's tolerance, forming a closed-loop control system of "heart rate monitoring → parameter adjustment → training execution." This enhances the scientific rigor and individualized fit of training, accelerating the rehabilitation process.

[0082] Figure 4 This is a flowchart illustrating an assistive robot control method for improving lower limb walking function after stroke, based on some embodiments of this specification, such as... Figure 4As shown, an assistive robot control method for improving lower limb walking function after stroke may include the following steps: Multimodal test data were collected from patients, including key head cortex point test data, bilateral ankle coordination test data, and balance disorder test data. Based on the patient's multimodal testing data, the characteristics of the patient's lower limb dysfunction were determined; Collect lower limb characteristics from the patient; Based on the patient's lower limb dysfunction characteristics, lower limb characteristics, and sample database, the patient's walking function rehabilitation training parameters are determined. These parameters include training frequency, cycle duration, walking speed, and gait trajectory. Based on the parameters for gait function rehabilitation training, an assistive robot is controlled to help patients perform lower limb gait function training.

[0083] The assistive robot control method for improving lower limb walking function after stroke can be applied to assistive robot control systems for improving lower limb walking function after stroke, which will not be elaborated here.

[0084] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. An assistive robot control system for improving lower limb walking function after stroke, characterized in that, include: The data acquisition module is used to collect multimodal test data from patients, including key head cortical point test data, bilateral joint coordination test data, and walking balance disorder test data. The functional analysis module is used to determine the characteristics of lower limb dysfunction in patients based on their multimodal test data. The data acquisition module is also used to collect the patient's lower limb characteristics; The decision-making module is used to determine the patient's walking function rehabilitation training parameters based on the patient's lower limb dysfunction characteristics, lower limb characteristics, and sample database. The walking function rehabilitation training parameters include training frequency, cycle duration, walking speed, and gait trajectory. The training control module is used to control the assistive robot to assist the patient in lower limb walking function training based on the walking function rehabilitation training parameters.

2. The assistive robot control system for improving lower limb walking function after stroke according to claim 1, characterized in that, The data acquisition module collects test data from key scalp cortex points of the patient, including: Identify multiple test sites on the scalp cortex; Identify multiple test movements; For each test exercise, EEG signals from multiple test scalp cortex points of multiple test patients were acquired during the test exercise. Among them, the lower limb dysfunction of any two test patients is different. Based on the EEG signals from multiple test scalp cortex points of multiple test patients during the test exercise, the β wave power vector of the test exercise for each test patient was determined. Based on the β wave power vector of the test exercise for each test patient, the β wave power difference value corresponding to the test exercise was calculated. Based on the difference in β-wave power corresponding to each test motion, several key test motions are determined from a variety of test motions; Based on the EEG signals of multiple test cerebrocortical points from multiple test patients during each key test movement, multiple left-side key cerebrocortical points and multiple right-side key cerebrocortical points were identified. Based on multiple key left scalp cortex points and multiple key right scalp cortex points, test data of key scalp cortex points of the patient were collected. The test data of key scalp cortex points of the patient included EEG signals of each key left scalp cortex point and EEG signals of each key right scalp cortex point.

3. The assistive robot control system for improving lower limb walking function after stroke according to claim 2, characterized in that, The data acquisition module collects bilateral joint coordination test data from the patient, including: Based on the patient's key cerebral cortex test data, key adjacent motor segments were identified; Obtain continuous relative phase curves for each pair of key adjacent motor segments on both sides of the patient.

4. The assistive robot control system for improving lower limb walking function after stroke according to claim 3, characterized in that, The data acquisition module collects the patient's walking balance disorder test data, including: Determine multiple key gait cycle percentiles; Collect joint kinematic characteristics of the patient during one gait cycle; Based on multiple key gait cycle percentiles, multiple key joint angles on the left and right sides of the patient are extracted from the joint kinematic features of the patient in one gait cycle, corresponding to each key gait cycle percentile.

5. The assistive robot control system for improving lower limb walking function after stroke according to claim 4, characterized in that, The data acquisition module determines the characteristics of the patient's lower limb dysfunction based on the patient's multimodal test data, including: The lateralization index was calculated based on the EEG signals of each key left scalp cortex point and each key right scalp cortex point. The bilateral joint coordination index is calculated based on the continuous relative phase curves of each pair of key adjacent motor segments on both sides of the patient. The balance index is calculated based on the left and right key joint angles corresponding to the percentile of each key gait cycle of the patient. The patient's lower limb dysfunction characteristics include lateralization index, bilateral joint coordination index and balance index.

6. The assistive robot control system for improving lower limb walking function after stroke according to any one of claims 1-5, characterized in that, The data acquisition module collects the patient's lower limb characteristics, including: Acquire images of the patient's lower limbs; The patient's bone length was determined based on images of the lower limbs.

7. The assistive robot control system for improving lower limb walking function after stroke according to any one of claims 1-5, characterized in that, The sample database includes multiple rehabilitation training samples, which include lower limb dysfunction characteristics, lower limb characteristics, and walking function rehabilitation training parameters of the patients in the rehabilitation training samples. The decision-making module determines the patient's walking function rehabilitation training parameters based on the patient's lower limb dysfunction characteristics, lower limb characteristics, and sample database, including: Based on the lower limb dysfunction characteristics and lower limb features of patients in the rehabilitation training sample, multiple rehabilitation training sample patients were grouped to generate grouping results; Based on the grouping results and the patients' lower limb dysfunction characteristics and lower limb features, the parameters for rehabilitation training of patients' walking function were determined.

8. The assistive robot control system for improving lower limb walking function after stroke according to claim 7, characterized in that, Based on the lower limb dysfunction characteristics and lower limb features of the patients in the rehabilitation training samples, multiple rehabilitation training samples were grouped, and grouping results were generated, including: Calculate the similarity of lower limb features between any two rehabilitation training sample patients; Based on the similarity of lower limb features between any two rehabilitation training sample patients, multiple rehabilitation training sample patients are first grouped to determine multiple first sample groups; For each first sample group, calculate the similarity of lower limb dysfunction characteristics between any two rehabilitation training sample patients included in the first sample group. Based on the similarity of lower limb dysfunction characteristics between any two rehabilitation training sample patients included in the first sample group, perform a second grouping of multiple rehabilitation training sample patients included in the first sample group to determine multiple second sample groups.

9. The assistive robot control system for improving lower limb walking function after stroke according to any one of claims 1-5, characterized in that, Also includes: The heart rate monitoring module is used to acquire the patient's heart rate data during the process of assisting the patient with lower limb walking function training by the assistive robot; The decision-making module is also used to determine whether to adjust the parameters based on the patient's heart rate data. If so, the patient's walking function rehabilitation training parameters are adjusted based on the patient's heart rate data, and the adjusted walking function rehabilitation training parameters are generated. The training control module is also used to control the assistive robot to assist the patient in lower limb walking function training according to the adjusted walking function rehabilitation training parameters.

10. A control method for an assistive robot used to improve lower limb walking function after stroke, characterized in that, The Collect multimodal test data from patients, including key head cortical point test data, bilateral ankle coordination test data, and balance disorder test data; Based on the patient's multimodal testing data, the characteristics of the patient's lower limb dysfunction were determined; Collect lower limb characteristics from the patient; Based on the patient's lower limb dysfunction characteristics, lower limb characteristics, and sample database, the patient's walking function rehabilitation training parameters are determined, wherein the walking function rehabilitation training parameters include training frequency, cycle duration, walking speed, and gait trajectory. Based on the parameters for gait function rehabilitation training, an assistive robot is controlled to help patients perform lower limb gait function training.

Images