Cargando…

Longitudinal Case Study of Regression-Based Hand Prosthesis Control in Daily Life

Hand prostheses are usually controlled by electromyographic (EMG) signals from the remnant muscles of the residual limb. Most prostheses used today are controlled with very simple techniques using only two EMG electrodes that allow to control a single prosthetic function at a time only. Recently, mo...

Descripción completa

Detalles Bibliográficos
Autores principales: Hahne, Janne M., Wilke, Meike A., Koppe, Mario, Farina, Dario, Schilling, Arndt F.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Frontiers Media S.A. 2020
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7318897/
https://www.ncbi.nlm.nih.gov/pubmed/32636734
http://dx.doi.org/10.3389/fnins.2020.00600
Descripción
Sumario:Hand prostheses are usually controlled by electromyographic (EMG) signals from the remnant muscles of the residual limb. Most prostheses used today are controlled with very simple techniques using only two EMG electrodes that allow to control a single prosthetic function at a time only. Recently, modern prosthesis controllers based on EMG classification, have become clinically available, which allow to directly access more functions, but still in a sequential manner only. We have recently shown in laboratory tests that a regression-based mapping from EMG signals into prosthetic control commands allows for a simultaneous activation of two functions and an independent control of their velocities with high reliability. Here we aimed to study how such regression-based control performs in daily life in a two-month case study. The performance is evaluated in functional tests and with a questionnaire at the beginning and the end of this phase and compared with the participant’s own prosthesis, controlled with a classical approach. Already 1 day after training of the regression model, the participant with transradial amputation outperformed the performance achieved with his own Michelangelo hand in two out of three functional metrics. No retraining of the model was required during the entire study duration. During the use of the system at home, the performance improved further and outperformed the conventional control in all three metrics. This study demonstrates that the high fidelity of linear regression-based prosthesis control is not restricted to a laboratory environment, but can be transferred to daily use.