Session: 07-10-03: Medical Robotics, Rehabilitation, and Surgery III

Paper Number: 167146

AI-Driven Task Optimization for Upper Limb Robot-Assisted Rehabilitation 

Rehabilitation therapy is a cornerstone of recovery for patients with limb impairments, yet traditional exercises often fail to optimally engage targeted muscle groups, resulting in prolonged recovery times and suboptimal outcomes. This study presents a novel approach that integrates artificial intelligence (AI), surface electromyography (sEMG), and inertial measurement units (IMUs) to identify and optimize rehabilitation trajectories that maximize muscle engagement. The central hypothesis is that identifying effective movement trajectories within a reduced workspace will enhance rehabilitation efficiency by promoting optimal muscle activation while minimizing unnecessary limb motion. By analyzing the relationship between joint angles, limb postures, and muscle activation patterns, the study aims to establish a geometrically defined workspace consisting of “effective cloud points”—specific limb positions and orientations that elicit maximal muscle engagement based on threshold-based sEMG activity levels. The study focuses on three critical upper-limb muscles: Biceps Brachii (BB) for flexion, Triceps Lateral (TL) for extension, and Triceps Long (TLo) for pronation and supination of the forearm. To validate the proposed method, an experiment was conducted using five healthy subjects performing standardized upper-arm rehabilitation tasks at varying speeds—slow, medium, and fast—while sEMG and IMU data were recorded. The collected signals were processed using a MATLAB-based algorithm that extracted IMU values corresponding to sEMG activation thresholds, allowing for the identification of optimized rehabilitation trajectories. Preliminary findings from the study revealed significant variations in muscle engagement across different trajectories, with the Triceps Long (TLo) muscle consistently demonstrating lower activation levels for most traditional tasks, highlighting the need for refined movement paths that ensure balanced muscle utilization. A wireless Trigno DELSYS system was used to capture sEMG signals from the BB, TL, and TLo muscles while participants followed predefined rehabilitation tasks, with IMU-tracked limb motions being color-coded to correspond with muscle activation levels. Analysis of trajectory data demonstrated that traditional rehabilitation exercises often fail to fully engage all target muscles, emphasizing the necessity of a trajectory optimization framework to improve therapeutic effectiveness. The study provides fundamental insights into human limb motion and muscle engagement while laying the groundwork for a data-driven approach to rehabilitation that reduces both workspace requirements and patient recovery time. The ability to generate distinct and robot-based reproducible cloud points that define an optimized movement workspace has significant implications for clinical rehabilitation, as it offers a more effective and personalized approach to therapy.  By leveraging multi-modal sensor data and AI-driven analysis, this study presents a step forward in the development of intelligent rehabilitation strategies that enhance motor recovery and improve patient outcomes.

Presenting Author: Yimesker Yihun North Carolina Agricultural and Technical State University (NCAT)

Presenting Author Biography: --

Authors:

Yimesker Yihun North Carolina Agricultural and Technical State University (NCAT)
Hailemicael Yimer North Carolina Agricultural and Technical State University (NCAT)
Safeh Clinton Mawah North Carolina Agricultural and Technical State University
Amanuel Tereda North Carolina Agricultural and Technical State University
Amirhossein Majidirad Purdue University Northwest