ExoMechHand prototype development and testing with EMG signals for hand rehabilitation
Ajdar Ullah1, Asim Waris1, Uzma Shafiq1
1National University of Sciences and Technology, Islamabad 44000, Pakistan.
Medical Engineering & Physics
|February 28, 2024
Summary
ExoMechHand, an affordable robotic device, shows promise in improving hand function for patients with nerve damage. Machine learning accurately predicts optimal resistance levels, aiding rehabilitation without constant therapist oversight.
Area of Science:
- Biomedical Engineering
- Rehabilitation Technology
- Neuroscience
Background:
- Rehabilitation is crucial for patients with disabilities, yet traditional therapy can be costly and inaccessible.
- Advancements in robotics and cost-effective devices are enabling unsupervised use of rehabilitative tools.
- Median and ulnar neuropathies significantly impair hand function, necessitating effective therapeutic interventions.
Purpose of the Study:
- To evaluate the efficacy of the ExoMechHand device for hand rehabilitation in patients with median or ulnar neuropathies.
- To develop and validate a machine learning model for automatically determining optimal resistance levels for rehabilitation.
- To assess the potential of surface-electromyography (sEMG) signals for staging hand conditions.
Main Methods:
- Ten subjects with median/ulnar neuropathies used ExoMechHand with varying resistive plates for 20 days.
- Hand function was assessed via grip strength, wrist range of motion, and metacarpophalangeal joint range of motion.
- Surface-electromyography (sEMG) signals from flexor-carpi-ulnaris and flexor-carpi-radialis muscles were recorded from 15 subjects.
- Machine learning algorithms, including extra-trees, random forest, and gradient boosting, were trained to classify optimal resistive plates based on sEMG features.
Main Results:
- Three out of ten subjects demonstrated improvements in hand grip, wrist range of motion, or metacarpophalangeal joint range of motion.
- The extra-trees classifier achieved 98% accuracy in predicting the appropriate resistive plate, outperforming random forest (95%) and gradient boosting (93%).
- sEMG signals from specific forearm muscles effectively correlated with hand condition staging in subjects with neuropathies.
Conclusions:
- ExoMechHand is a cost-effective and robust rehabilitation tool for improving hand function in patients with median and ulnar neuropathies.
- Machine learning models, particularly extra-trees, can accurately guide rehabilitation by selecting optimal resistive settings based on sEMG data.
- sEMG analysis offers a viable method for objective staging of hand conditions in neurological rehabilitation.
