Towards the application of one-dimensional sonomyography for powered upper-limb prosthetic control using machine...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Signal Processing

Background:

• Surface electromyography (sEMG) has limitations for multi-degree of freedom prosthetic control.
• Muscle thickness measured by ultrasound can predict wrist angle, offering an alternative prosthetic control signal.
• Existing ultrasound machines are bulky and expensive, hindering practical application.

Purpose of the Study:

• To assess the feasibility of using portable A-mode ultrasound for one-dimensional sonomyography.
• To evaluate machine learning models (SVM, RBF ANN, BP ANN) for predicting wrist angle from sonomyography signals.
• To explore a cost-effective and portable alternative for prosthetic control.

Main Methods:

• A feasibility study was conducted with nine healthy subjects.
• Subjects performed wrist extensions at varying speeds (15, 22.5, 30 cycles/minute).
• Data from 22.5 cycles/minute trials trained models; other trials were used for cross-validation. Prediction accuracy was measured by RMSE and CC.

Main Results:

• Support vector machine (SVM) demonstrated excellent prediction accuracy (RMSE = 13%, CC = 0.975).
• SVM outperformed radial basis function artificial neural network (RBF ANN) and back-propagation artificial neural network (BP ANN).
• One-dimensional sonomyography shows high potential for accurate wrist angle prediction.

Conclusions:

• One-dimensional sonomyography using portable A-mode ultrasound is a feasible alternative signal for prosthetic control.
• This approach may overcome limitations of surface electromyography in advanced prosthetic devices.
• This technology could significantly advance powered prosthesis design and improve outcomes for amputees.