Identification of phantom movements with an ensemble learning approach.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Machine Learning in Healthcare

Background:

• Phantom limb pain significantly impacts amputees' quality of life and daily activities.
• Phantom limb movements are controllable by most amputees, presenting opportunities for rehabilitation and prosthetic control.
• Existing research on processing surface electromyography (sEMG) signals in amputees is limited, particularly for classifying upper extremity and hand movements.

Purpose of the Study:

• To classify and recognize phantom limb movements across various upper and lower extremity amputation levels.
• To evaluate the efficacy of ensemble learning algorithms for phantom movement recognition.
• To compare the performance of ensemble learning models against Support Vector Machine (SVM), Decision Tree, and k-Nearest Neighbors (kNN) methods.

Main Methods:

• Utilized ensemble learning algorithms for the classification and recognition of phantom movements.
• Collected sEMG signals from 38 amputees and 25 healthy individuals to create a comprehensive dataset.
• Applied advanced signal processing techniques to analyze sEMG data for movement pattern identification.

Main Results:

• Ensemble learning-based models demonstrated superior accuracy in detecting phantom limb movements.
• The proposed ensemble learning approaches outperformed SVM, Decision Tree, and kNN in classification tasks.
• Movement pattern recognition accuracy reached up to 96.33% in healthy individuals and a maximum of 79.16% in amputees.

Conclusions:

• Ensemble learning offers a promising and accurate approach for recognizing phantom limb movements in amputees.
• This technology has significant potential for advancing amputee rehabilitation strategies and improving prosthetic limb control.
• Further research is warranted to enhance recognition accuracy in amputee populations and explore diverse clinical applications.