Development of a functional recovery training robot with a quantitative evaluation system for post-stroke patients with upper-limb impairment

Abstract

In the past few decades, clinical treatments utilizing rehabilitation robotics have shown sustainable improvement in sensory motor recovery for post-stroke patients. In addition, analyzing patients exercise progression using robotics systems enable clinicians to objectify the assessment, and improve measurement reliability and precision. In this research, an upper-limb therapeutic robot for stroke patients with upper limb impairments, that supports patients to practice independently, while recording and evaluating their exercise performance was developed. Also, a unique quantitative assessment system to evaluate upper-limb motor function using Mahalanobis-Taguchi system, MTS, was investigated and proposed. First, a calibration experiment was conducted to generate an equation, to compute the force applied by the subject on the robot arm of the upper limb therapeutic robot. A fixed set of weights were attached to the central rotation shaft to apply constant torque to the rotation shaft. The rotation shaft was rotated in a non-load condition and load condition with weights added, and the angular signal of the potentiometer at that time was measured. From the experiment, a relational formula between the mean of central rotation shaft twist angle, and applied torque to the central rotation shaft was successfully generated. In addition, to increase the potentiometer angular signal precision and reliability, all control unit components were grounded. As results, the coefficient of variation reached a maximum of 8.7%, and the noise affecting the signal was suppressed. Next, to establish a quantitative assessment system to evaluate upper-limb motor function, an evaluation experiment with thirty able-body subjects, under six replicated joint of motion restraint conditions using the upper-limb therapeutic robot was conducted. MTA method, T(1), and variation pressure method’s applicability in distinguishing between able-body subjects, in arm-end force controllability between shoulder, elbow, and wrist joints under replicated joint restraint was investigated. In addition, these methods were compared with each other for their distinguishing quality. The subject’s task was to continuously apply a fine constant resistance force, in the opposite tangential direction of the robot arm rotation. No-restraint condition data were set as unit-space dataset and signal dataset. The other six joint restraint condition data were computed as targeted dataset since these were quasi-role data of pre-training condition of a paralyzed limb. From coefficient of variation, CV, results there was a similarity between MTA method and T(1) method in regards of distinguishing between subjects’ arm-end force controllability performance between single-joint restraint conditions and combined joint restraint conditions. For Single-joint restraint conditions, wrist restraint CV result for T(1) is 1.921, and MTA is 0.575. For Combined joint restraint conditions, elbow and wrist restraint CV result for T(1) is 1.955, and MTA is 0.385. As results, the small coefficient of variation with light shoulder joint restraint reflects the influence on force controllability performance, shows that MTA method and T(1) method are effective in detecting differences between joints. Finally, a measurement system of the force applied by the subject on the robot arm of the upper limb therapeutic robot was developed. Also, a unique quantitative assessment system using MTA method and T(1) method was investigated and proposed as promising analytic systems to evaluate upper-limb motor function.