Assessment of upper limb muscle tone level based on estimated impedance parameters

• Published in: 2016 IEEE EMBS Conference on Biomedical Engineering and Sciences (IECBES) pp. 742 - 747
• Main Authors: Htoon, Zaw Lay, Sidek, Shahrul Na'im, Fatai, Sado, Yunahar, Taufik
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2016
• Subjects: 3-DOF robot-assisted; Artificial Neural Network (ANN); Artificial neural networks; Damping; Estimation; Impedance; Impedance measurement; Limbs; Mathematical models; Medical treatment; Muscles; occupational therapists; post-stroke patients; recursive least square estimator; robot-assisted rehabilitation; Robots; upper-limb impedance parameter
• DOI: 10.1109/IECBES.2016.7843549

Many strategies have been developed by occupational and physical therapists for the assessment of post-stroke patients' upper limb muscle tone and physical recovery progress. Despite, having the appropriate skills, they face serious challenges in quantifying continuously, the patients' recovery progress. Moreover, the therapy has become more costly and time consuming since the patients are required to have a face-to-face contact with the therapist over a long period of time. By deploying robot-assisted rehabilitation therapy, some of these problems have been addressed, however, serious challenges still exist in the aspect of proper estimation and assessment of patients muscle tone level and recovery progress during rehabilitation therapy. This paper proposes an appropriate strategy for prediction and assessment of subjects' muscle tone level and recovery based on the estimation of upper-limb mechanical impedance parameters. The subjects' mechanical impedance parameters are estimated using a recursive least square estimator method and the muscle tone level are predicted by Artificial Neural Network (ANN) which has been trained using the estimated impedance parameters. Preliminary experimental result shows that the upper-limb impedance parameters can be estimated to an accuracy level of 90%, while simulation studies have revealed that the muscle tone level can be reliably predicted at 95.01% accuracy level.