WO2022146271A1

WO2022146271A1 - A participation assessment system based on multimodal evaluation of user responses for upper limb rehabilitation

Info

Publication number: WO2022146271A1
Application number: PCT/TR2020/051505
Authority: WO
Inventors: Erkan ODEMIS; Cabbar Veysel BAYSAL
Original assignee: Cukurova Universitesi Rektorlugu
Priority date: 2020-12-31
Filing date: 2020-12-31
Publication date: 2022-07-07
Also published as: TR202022809A2

Abstract

The invention relates to a system that including a therapy exercise patient participation assessment method, which evaluates the participation and performance of patients during the upper extremity rehabilitation exercises based on a multimodal sensor fusion formed by the trajectory tracking error signal and the patient's physiological responses (heart rate and skin conductance) and adjusts difficulty levels of therapy exercises according to the patient's participation.

Description

A PARTICIPATION ASSESSMENT SYSTEM BASED ON MULTIMODAL EVALUATION OF USER RESPONSES FOR UPPER LIMB REHABILITATION

TECHNICAL AREA

The invention relates to a system including a therapy exercise patient participation assessment method, which evaluates the participation and performance of patients during the upper extremity rehabilitation exercises based on a multimodal sensor fusion formed by the trajectory tracking error signal and the patient's physiological responses (heart rate and skin conductance) and adjusts difficulty levels of therapy exercises according to the patient's participation.

BACKGROUND

Stroke and Spinal Cord Injury (SCI) are some of the most common causes of neuromuscular disorders, which have a severe impact on the ability of patients to achieve activities of daily living (ADL). Physiotherapy and rehabilitation are widely used methods for treating patients with neuromuscular disorders. In conventional therapy, rehabilitation exercises are performed with therapists. Still, this method has some disadvantages, such as not enough time being spent with each patient or performance reduction of therapists in exercises due to overloading. On the other hand, performing exercises with robotic devices have appeared as a new approach to overcoming the disadvantages of conventional therapy.

In conventional and robotic rehabilitation, the patient's active participation in the exercises is crucial in obtaining maximum functional outputs and improving neural plasticity. In the rehabilitation exercises with robotic devices, the Assist-as-Needed (AAN) paradigm has emerged to ensure patients' active and voluntary participation [6], In the AAN approach, the robotic assistance provided to patients is determined based on the patient's performance. Also, therapy tasks and their difficulty levels are adjusted according to the patient's performance to ensure the exercises are challenging enough for the patient. In the AAN strategies, robotic assistance is applied when the patients diverge from reference trajectories or could not complete the desired therapy tasks. Thereby, it is aimed to enable patients to participate more actively in therapy exercises and to prevent passive training therapies that do not give much functional output. The core of the AAN strategies is the patient performance evaluation method, which is the base for determining the robotic assistance provided to the patient and adjusting the therapy tasks. In the literature, there are many different patient performance evaluation methods. However, the existing methods in the literature are based on either some specific device designs, or some certain therapy tasks. Also, some performance evaluation methods ignore the patient's changing capabilities during the exercises. Krebs et al. first implemented a performance-based progressive robotic therapy, where the patient’s performance was estimated using the patient’s active force and motion accuracy. The stiffness of the system has been determined by the patient’s performance of the last reaching movement. Papaleo et al. proposed a patient-tailored adaptive therapy for an upper limb robotic rehabilitation device, where a section for the evaluation of the patient’s biomechanical performance and a section for the modulation of the robotic assistance were included. The parameters of these methods are defined in discrete values and updated after a therapy task is completed. That restricts the adaptability of the system and ignores the changing capabilities of patients during the exercises. Some studies have used specific indices based on the performance or experimental measurements of the patient to evaluate their therapy engagement. The drawback of these approaches is that the index calculation methods are only suitable for certain therapy tasks.

Leconte and Ronsse proposed a performance-based assistive strategy that measured the movement performance based on three features (smoothness, velocity, and amplitude) of the patient's motion by using an adaptive oscillator on rhythmic circular arm exercises. Evaluation of movement performance with this approach is suitable only on rhythmic circular therapy tasks. Pehlivan et al. introduced a minimal assist-as-needed (mAAN) strategy, which relied on patients' sensorless force estimation by using a Kalman filter. In another approach, Chen et al. simulated patient interaction forces with a dynamic human arm model and real-time measurement of patient-exerted torques. A significant shortcoming of both approaches was that the estimation of the patient’s capabilities had been based on the dynamic model of the robotic devices. As the robotic device's mechanical structure becomes complex, it is challenging to estimate patient performance precisely.

Carmichael and Liu proposed a model-based AAN structure that estimated the muscular capability of the patient by using a musculoskeletal model. In this approach, a Task model (TM) calculates the strength required to perform desired upper limb tasks, and a Strength Model (SM) estimates the patient’s strength capability on a muscular level by using a musculoskeletal model. Patient capabilities depend on other factors such as tiredness, joint stability, comfort, and the ability to coordinate movements. In this approach, these other factors have been ignored. Also, musculoskeletal model parameters have not been adjusted according to the patient.

Physiological responses (such as skin conductance, heart rate, respiration, and skin temperature) are good references for evaluating the patients' physical activity and emotional states during the rehabilitation exercises. In the existing literature, stress level detection or emotion recognition based on physiological responses were implemented by researchers. In another approach, the research studies using physiological signals assessed patients' mental status like valance and arousal based on these signals. In the studies mentioned, the researchers adjusted the therapy difficulties according to patients' these psychological states. However, to the best of our knowledge, none of the previous researches has been used physiological responses to evaluate the patient's performance and therapy engagement. Also, in the related studies, new systems and/or methods have been proposed for motion analysis during therapy exercises. Nevertheless, the patient's therapy performance was not evaluated in any of these motion analysis systems.

BRIEF DESCRIPTION OF THE INVENTION

The invention relates to a system that including a therapy exercise patient participation assessment method, which evaluates the participation and performance of patients during the upper extremity rehabilitation exercises based on a multimodal sensor fusion formed by the trajectory tracking error signal and the patient's physiological responses (heart rate and skin conductance) and adjusts difficulty levels of therapy exercises according to the patient's participation. The method evaluates the therapy performances of patients independently from any therapy tasks or device designs. The developed system also evaluates the patient's tiredness and slacking, which are substantial factors affecting therapy performance. This method presents a low-cost system and can be applied in any kind of rehabilitation exercises. Since the system evaluates patient performance independently from any device design, it can also be used for performance evaluation in conventional therapy and sports exercises. LIST OF FIGURES

Figure 1. Overview of The Designed Participation Assessment System

Figure2. Patient Response Estimator Subsystem

Figures. Membership Functions of HR

Figured Membership Functions of GSR

Figures. Membership Functions of Tracking Error

Figure 6. Membership Functions of HR Change, GSR Change, and Tracking Error Change

Figure?. Membership Functions of Performance, Tiredness, and Slacking

Figure 8. Implemented algorithms for performance assessment

Figure9. Implemented algorithms for tiredness assessment

Figure 10. Implemented algorithms for slacking assessment

Figure 11 . Subject 1 Performance Assessment

Figure 12. Subject 2 Performance Assessment

Figure 13. Subject s Performance Assessment

Figure 14. Subject 4 Performance Assessment

Figure 15. Subject s Performance Assessment

Figure 16. Subject s Performance Assessment

Figure 17. Subject ? Performance Assessment

Figure 18. Tasks level changes for all subjects during the experiments

Figure 19. HR signals of all subjects during experiments

Figure 20. GSR signals of all subjects during experiments

Figure 21 . Trajectory tracking error signals of all subjects during experiments

Figure 22. Tiredness assessment of the 7. subject

Figure 23. Tasks level changes for all subjects during the slacking assessment experiments

Figure 24. HR Signals of the subjects during the slacking assessment experiments

Figure 25. GSR Signals of the subjects during the slacking assessment experiments

Figure 26. Tracking Error Signals of the subjects during the slacking assessment experiments

Figure 27. Slacking Evaluation Experiments - Subject 1 Performance

Assessment Figure 28. Slacking Evaluation Experiments - Subject 2 Performance Assessment

Figure 29. Slacking Evaluation Experiments - Subject 1 Slacking Assessment

Figure 30. Slacking Evaluation Experiments - Subject 2 Slacking Assessment

Correspondence of the letters given in the figures:

A. Multimodal Sensory Subsystem

B. Upper Limb Kinematic Module

C. Patient Response Estimator Subsystem

D. Therapy Task Management Module

E. Graphical User Interface

L. Low

M. Medium

H. High

DETAILED DESCRIPTION OF THE INVENTION

The system of the invention contains software that applies a therapy exercise patient participation assessment method which evaluates the participation and performance of patients during upper extremity rehabilitation exercises based on a multimodal sensor fusion formed by the trajectory tracking error signal and the patient's physiological responses (heart rate and skin conductance). The method evaluates the therapy performances of patients independently from any therapy tasks or device designs. The developed system also evaluates the patient's tiredness and slacking, which are substantial factors affecting therapy performance.

Physiological responses are good indicators for evaluating the patients' physical activities during the rehabilitation exercises. Especially the researches in the last decade showed that a combination of the HR and skin conductance signals would be a robust method for estimation of the physical and mental workload of a person during virtual rehabilitation tasks. These studies have revealed that the HR and skin conductance signals increase with high physical activity and decrease with the fall of physical effort during the therapy exercises. Therefore, skin conductance, heart rate signals, and alteration of these signals (increase or decrease) are employed in the proposed method for evaluating the physical effort and voluntary participation of the patient in the therapy exercises. Also, the desired trajectory tracking error signal, which expresses the difference between the desired therapy task and the patient's position during rehabilitation exercises, is a driving signal for motor learning. Therefore, we combined the trajectory tracking error signal and physiological responses for estimating the patient's performance during therapy to improve motor learning.

The proposed method can be considered a complex system comprising five main parts; multimodal sensory subsystem (A), upper limb kinematic module (B), Patient Response Estimator Subsystem (PRES) (C), therapy task management module (D), and Graphical User Interface (GUI) (E).

During the rehabilitation exercises, patient physiological responses are measured via HR and skin conductance sensors. The upper extremity joints angles are measured via inertial measurement unit (IMU) sensors. For determining the trajectory tracking error signal, the desired therapy task is compared to the end-effector position of the patient obtained using the upper limb kinematic module (B). The PRES (C) evaluates the patient's therapy performance, tiredness, and slacking using physiological responses and trajectory tracking error signal. Therapy tasks and difficulty levels are adjusted by the therapy task management module (D) based on the PRES's outputs. For rehabilitation exercises and interaction with the patient, a Matlab Graphical User Interface (GUI) (E) is operated. An overview of the designed participation assessment system is given in Figure 1 .

The multimodal sensor subsystem (A) contains two sensory data; inertial measurement unit (IMU) sensors and physiological signals sensors. All the sensor data is transferred to the target PC via serial communication with a 10 Hz sampling rate and processed in a MATLAB® / Simulink model. For measuring the upper arm joints' angles and evaluating the desired trajectory tracking performance of the patient during therapy, two IMU sensors are used. These sensors are mounted on the patient's right arm using cuffs, one for the upper arm and one for the forearm during rehabilitation exercises.

The quaternion outputs of IMU sensors were first transformed into upper extremity joint angles. Then, using the upper extremity kinematic module (B), the patient's arm movements during the exercises were estimated. The trajectory tracking error signal is achieved by comparing the desired therapy task, determined by the therapy task management module, and the patient's arm movement obtained using the upper limb kinematic module. Heart rate indicates the number of heartbeats in one minute and increases depending on the sympathetic nervous system's activity. The heart rate is measured by using the Heart Rate sensor, which uses the photoplethysmography (PPG) technique. PPG is a noninvasive and low-cost optical technique applied to detect blood volume changes in tissues affected by heartbeats. The heart rate sensor is located on the subject's left-hand index finger.

Skin conductivity (or galvanic skin response (GSR)) describes the electrical conductivity of the skin, which varies depending on the secretions of the sweat glands and represents the activity of the sympathetic nervous system. The selected sensor for measuring the skin conductance is the GSR sensor. Electrodes of the sensor are placed on the middle finger and thumb of the subject's left hand. GSR sensor measures the microvolts between these fingers. For obtaining skin conductance, first, human skin resistance (in ohm) has been calculated using Equation 1. Then the human skin resistance has been converted to skin conductance. During the exercises, the subjects were asked not to move their left hands in which the HR and GSR sensors were attached for not affecting the measurements.

Human Resistance = ((1024 + 2xSensor_Reading)xl0000)/(512 — Sens or -Reading)

Equation (1 )

The patient's therapy performance, tiredness, and slacking during the rehabilitation exercises are assessed by PRES, using a Fuzzy Inference System (FIS). The FIS is the process of mapping from a given input to an output using the fuzzy set theory and fuzzy logic. The FIS has been applied to a wide variety of problems such as robotic, control, biomechanics and is also used for medical applications successfully. The FIS usually are implemented in Mamdani and Sugeno methods. In this research, the Mamdani method was selected due to its easy-to-apply intuitive structure and compatibility with human inputs. Also, the Mamdani method can be implemented with observations and is suitable for the guidance of the physiotherapists.

The FIS contains three central units; fuzzification, decision-making, and defuzzification. Fuzzification is a mathematical operation for converting an element in the universe of discourse into a fuzzy set membership value. The fuzzification process receives the elements x,y e X and produces the membership degrees |i (x), |i (y), |i_B(x) and |i_B(y). The Decision-making unit achieves mapping of a given input to an output using membership functions, logical operations, and predefined IF- THEN rules. This process maps the fuzzified inputs to the rule base and produces a fuzzified output for each rule. Defuzzification is a mathematical process used to convert a fuzzy set to a real number. The defuzzification unit transforms the fuzzy results of the interface into output variables. The defuzzification process used in this work is the centroid of the area (CoA) method.

The designed FIS takes the HR, GSR, the desired trajectory tracking error, and alterations of these signals during the therapy exercises as inputs. Since physiological signals can vary according to environmental factors and from person to person, changes in these signals during the exercises (increase or decrease) are considered in the patient's performance and tiredness evaluation. Alteration in the tracking error signal is used for tiredness evaluation only. The proposed FIS has three outputs; Performance, Tiredness, and Slacking, as shown in Figure 2. In the figure, Ae(t), AHR and AGSR represent the changes in trajectory tracking error, HR, and GSR signals, respectively.

All input and output variables of the FIS are defined by membership (MS) functions |i (x) that can take values from “Low” to “High” and ranged from 0 to 1. Membership functions of the FIS have been determined experimentally and characterized based on the researches in literature. For deciding the alterations, HR, GSR, and the tracking error signals are averaged every five seconds of simulation time and compared to their last mean values. If these signals increase or decrease, they take the "High" or "Low" value, respectively. The medium value is used for unchanged signals. The membership functions of all the inputs and outputs of the FIS are given in Figure 3-7.

After determining the membership functions, a fuzzy rule-based module with a total of 58 IF-THEN rules has been defined. All the rules have been designated experimentally. Some examples of the rules are given in Table 1 .

IF-THEN Statements

IF (HR Change is High) and (GSR Change is High) and THEN (Performance is

(Error is High) Low)

IF (HR Change is Mid) and (GSR Change is High) and THEN (Performance is

(Error is Mid) MidLow) IF (HR Change is Mid) and (GSR Change is Mid) and THEN (Performance is (Error is Mid) Mid)

IF (HR Change is Mid) and (GSR Change is Mid) and THEN (Performance is (Error is Low) MidHigh)

IF (HR Change is Low) and (GSR Change is Low) and THEN (Performance is

(Error is Low) High)

IF (HR is Low) and (GSR is Low) and (Error is Low) THEN (Tiredness is Low) and (Slacking is Low)

IF (HR is Mid) and (GSR is Mid) and (Error is Mid) THEN (Tiredness is MidLow) and (Slacking is Mid)

IF (HR is High) and (GSR is High) and (Error is High) THEN (Tiredness is High) and (Slacking is Low)

IF (HR is Low) and (GSR is Low) and (Error is High) THEN (Tiredness is Low) and (Slacking is High)

Table 1. Some Examples of the FIS IF-THEN Rules

As mentioned earlier, HR and GSR signals are good references for evaluating the patient's physical activity during the virtual rehabilitation tasks. Thence in the proposed method, performance evaluation is carried out using the tracking error signal with alterations in the HR and GSR responses. For instance, if the HR and GSR signals are decreasing and the tracking error value is "Low", this indicates that the patient has successfully performed the exercise; the exercise is not enough challenging for the patient. The performance of the patient is considered "High" for these conditions, Suppose the HR and GSR signals are increasing, and the tracking error is also high.

In that case, this condition is considered the patient could not perform the exercise despite his/her effort and performance will take "Low" value. Patients' performance may change momentarily during exercises. These instant performance variations should not be neglected. However, these variations can cause therapy task difficulty levels to change frequently when therapy tasks are determined by patient performance. Therefore, to not neglect the instantaneous performance variations of the patients and prevent the therapy task difficulty levels from changing too often due to these variations, the performance assessment was implemented by taking the mean of the PRES's performance output values every twenty seconds of simulation time. If this mean performance value is more than 0.7, the therapy task's difficulty level is increased; if it is less than 0.55, it is decreased. Upper and lower boundaries of the performance assessment, in which the default values are 0.7 and 0.55, have been specified experimentally. These boundaries could be changed by using the sliders on the GUI screen to make the exercises easier or more difficult based on the patient's impairment level and performance.

Patients may get tired during rehabilitation exercises, and this affects their therapy performance. When a patient fails to perform therapy exercises successfully, it is critical to determine whether this is due to tiredness or lack of motor skills. For this reason, in this study, fatigue due to continuous physical exertion has also been evaluated in addition to patient performance assessment. HR and GSR signals are accurate indicators for predicting a person's tiredness. Therefore, tiredness evaluation is achieved based on HR, GSR, and the tracking error signal. For instance, if the patient's HR and GSR signals are high, and the tracking error signal is medium, the patient's tiredness is evaluated as "MidHigh", if the tracking error signal is high, then tiredness is considered as "High". For improving the tiredness estimation, the tracking error signal's alteration was also used in the assessment. If the patient's HR, GSR, and the trajectory tracking error signal increase, tiredness is considered "MidHigh". Tiredness assessment is performed by averaging the tiredness output values of the PRES for every ten seconds of simulation time. By the averaging operation, momentary changes in HR and GSR signals are prevented from affecting tiredness assessment. If the patient's mean tiredness value is more than 0.7, the therapy difficulty level is reduced, regardless of performance evaluation. The boundary of the tiredness assessment has been specified experimentally. This boundary could be changed by using the sliders on the GUI screen to make the exercises easier or more difficult based on the patient's impairment level and performance.

Especially in rehabilitation exercises performed with robotic devices, patients may exhibit a slacking motor control behavior that will allow robotic devices to take control of the tasks. For detecting this kind of motor behavior, the slacking assessment is also considered in this work. The slacking evaluation is achieved in a reverse manner to the tiredness evaluation. If the patient's HR and GSR signals are low, and the tracking error signal is high, the patient's slacking is evaluated as "High". If the slacking value is estimated at more than 0.7, a warning message for the patient appears on the GUI screen. This value has been specified experimentally and could be changed by using the sliders on the GUI screen. The implemented algorithms for performance, tiredness, and slacking assessment is given in Figure 8-10.

In this method, five different well-known therapy tasks have been selected as the rehabilitation exercises; Elbow Flexion, Shoulder Scaption, Shoulder Horizontal Adduction, Forward Elevation, and Diagonal Shoulder Raise. Each of the selected therapy tasks has three different speed levels for the trajectory target. Thence, three different difficulty levels have been defined for each task. The transition between therapy tasks and difficulty levels of these tasks has been carried out by the therapy task management module (D) according to the PRES's performance and tiredness outputs.

On the rehabilitation exercise interface, there are illustration figures about the exercise patterns. During the therapy exercises, the patients have been asked to catch the trajectory target on the desired trajectory path. On the GUI screen, the desired trajectory path has been given on a parallel to the patient’s frontal plane. Giving feedback to the patient about exercise performance enhances motor learning and influences motivation positively. Therefore on the GUI screen, the patient was provided with motivational feedback on his/her performance, tiredness, and slacking status.

The proposed method offers a low-cost, easy to apply, and noninvasive solution for the performance assessments of the patients during the rehabilitation exercises, which is crucial to increase the functional outputs received from therapy. This system is suitable for use in-home, clinical treatment environments, and telerehabilitation applications. Since the system evaluates patient performance independently from any device design, it can also be used for performance evaluation in conventional therapy and sports exercises.

The efficacy of the proposed method was tested on seven healthy subjects experimentally. The subjects carried out the exercises for ten to twelve minutes, depending on their performance. During the experiments, when the participants completed the three levels of difficulty set for each exercise, they moved on to the next therapy task. The performance assessment of all the subjects is given in Figure 11-17. In these figures, vertical lines indicate the 20-second intervals that the performance of subjects was evaluated. The variations in task difficulty levels for all the subjects are given in Figure 18. In this figure, the exercises are indicated by numbers. When a subject completes the three difficulty levels of a therapy task, they move on to the next task. The transition between therapy tasks and difficulty levels of these tasks has been carried out by the therapy task management module according to the PRES's performance and tiredness outputs. The three subjects completed the experiments at the last task and exercise level, two subjects have reached the final task level but not the final exercise level, and two subjects finished the experiment at the fourth task level. All subjects' HR, GSR, and trajectory tracking error signals during the experiments are given in Figure 19-21. The physiological responses of all the subjects show that these signals increase due to physical workload as therapy tasks are getting difficult. The increase in these signals indicates that using HR and GSR signals for evaluating patients' participation and physical effort in therapy exercises is a suitable method. However, as a result of subjects getting used to therapy tasks over time and performance increases, these signals decrease until a new therapy exercise. Therefore, using alterations of the physiological responses instead of their levels in evaluating patient performance occurs as a much more convenient method. The instant high trajectory tracking error signals of the subjects during the therapy usually arise from the adaptation period to the new therapy task or exercise level. During therapy, patients' performance can change momentarily, as seen in performance assessments of all the subjects given in Figure 11 -17. Hence, carrying out patient performance evaluation during a certain exercise period provides a more accurate assessment. These experimental results demonstrated that the proposed method successfully evaluates the subjects’ performances.

During the exercises of the seventh subject, the system detected tiredness. Due to tiredness detection, the exercise level was reduced twice during this subject's experiment. Tiredness evaluation of this subject is also given in Figure 22. In this figure, vertical lines indicate the 10-second intervals that the tiredness of the subject was evaluated. When the seventh subject's physiological signals are analyzed, it is seen that the GSR and especially the HR signal increase excessively during the exercises. The main reason for these increases is the excessive effort exerted by the subject to perform the tasks; therefore, a tiredness detection has occurred. The fifth subject's GSR signal exceeded the medium level (20 pS) during the experiment. However, this subject's HR signal remained stable throughout the experiment. Therefore, no tiredness determination has been performed for this subject. These results indicate that the proposed method successfully evaluates the tiredness of the subjects, as well as their performances. For evaluating the proposed method's slacking assessment, two subjects have performed the therapy exercises with healthy behavior until reaching the fourth therapy task level (Forward Elevation therapy task). Then, the subjects have completed the therapy tasks with simulated post-stroke behavior. For simulating the post-stroke behavior, subjects intentionally failed to perform the tasks by going in the wrong direction or repeatedly stopping during the exercises. The variations in task difficulty levels for these subjects during the slacking assessment experiments are given in Figure 23. The subjects' HR, GSR, and trajectory tracking error signals during the slacking evaluation experiments are given in Figure 24-26. In these figures, the points at which patients begin to perform the exercises with simulated post-stroke behavior are indicated as points A and B on the physiological responses of subjects 1 and 2, respectively. When the physiological signals of the subjects during the slacking assessment experiments were analyzed, it is seen that these signals decrease from the beginning of the simulated post-stroke behavior. These alterations show that physiological signals are successful indicators for evaluating the patient's physical activity and therapy participation. The performance and slacking assessments of the subjects during these experiments are given in Fig. 27-30. The trajectory tracking error signals of both subjects increased during the simulated post-stroke behavior. Along with these increases, with the decreases of physiological responses, the subjects' performances have been evaluated as low, and the developed system has decreased the exercise levels. Due to these changes in HR, GSR, and trajectory tracking error signals, the simulated post-stroke behavior of the subjects has been considered as slacking by the PRES. These experimental results have shown that the proposed method could successfully detect the slacking behavior of the subjects. To summarize, all experimental results demonstrated that the proposed method successfully estimates the subject's changing capabilities and therapy participation regardless of any therapy device or therapy tasks and adjusts the therapy tasks and task difficulty levels based on the subjects’ performance and tiredness.

In this method, the patient's pulse and skin conductance physiological responses are measured using HR and GSR sensors. Upper extremity joint angles are measured using inertial measurement unit (IMU) sensors. . All the sensor data is transferred to the target PC via serial communication with a 10 Hz sampling rate and processed in a MATLAB® I Simulink model. The upper limb kinematic module (B) estimates the patient's arm movements during rehabilitation exercises using measured joint angles. The trajectory tracking error signal is achieved by comparing the desired therapy task, determined by the therapy task management module (D), and the patient's arm movement. The PRES (C) evaluates the patient's therapy performance, tiredness, and slacking using physiological responses and trajectory tracking error signal based on a Fuzzy Inference System. Therapy tasks and difficulty levels are adjusted by the therapy task management module (D) based on the patient’s performance and tiredness. Rehabilitation exercises are performed using a Graphical User Interface (GUI) (E) screen.

Claims

CLAIMS A therapy participation assessment system for upper extremity rehabilitation based on multimodal evaluation of user responses by using the Inertial Measurement Unit (IMU) sensors, physiological signals sensors, and a Fuzzy Inference System characterized by comprising a multimodal sensory subsystem, an upper limb kinematic module, a patient response estimator subsystem, a therapy task management module and a graphical user interface.