EMBC 2024 - Leveraging motor unit spatial activation patterns for channel selection in finger force regression
16 Jul 2024The aim of this project is to test the hypothesis that we don’t need a large number of surface EMG channels to accurately decode individual finger forces when recording muscle activity with dense electrode arrays (high-density EMG), even for such a fine control task like individual finger activations. Because these EMG channels are placed very close together, they pick up similar activation patterns in the muscle. In other words, neighboring channels are highly correlated.
Given this observation, we wanted to explore computationally efficient ways of selecting a subset of channels with the highest information content about the task. One of the easiest approaches can be to just select a priori some columns and/or rows from this dense array and use those channels as input to a linear regression model. Piece of cake, right?
Now, which rows or columns to pick? And would this even generalize across subjects?
For the first question, we explored a few strategies guided by our understanding of motor unit activation patterns in the forearm muscle and some basic anatomy of muscle fiber orientation. We hypothesized that channels lying on the proximal-distal contain more information as they cover the full length of the muscle.
For the second question, we tested individual finger force regression performance across 20 healthy subjects when they were asked to peform a random isometric contraction for different channel selection strategies.
Design philosophy
Same as always, I wanted to keep text to a minimum and only have figures and illustrations with captions. All illustration are made with BioRender and the poster is designed using Latex. Unlike in the ISEK conference, I think I correctly respected the poster dimensions guidelines this time.
I must say this is one of the posters I mostly proud up to date :)
Paper Abstract
Individual finger forces can be predicted by regression of high-density surface electromyography (sEMG) signals. This is promising for applications in human-machine interfaces, specifically prosthesis control, although the large number of electrodes imposes high computational requirements. In this study, we present strategies for a-priori channel selection guided by motor unit spatial activation patterns to reduce computational costs without compromising decoding accuracy. In contrast to subject-specific data-driven selection, we test the hypothesis that pre-selecting sEMG channels for finger-specific force estimation can still generalize across subjects. We show that a subset of 32 channels, out of a total of 256, achieves an RMSE of 6.32 ± 2.34 % of the Maximum Voluntary Contraction (MVC) on the HYSER RANDOM dataset, competitive with the state-of-the-art baseline model, using all channels, which attains an RMSE of 5.57 ± 1.94 % MVC. These results highlight the potential of simple, a-priori channel selection strategies in decoding finger forces from sEMG, which would be particularly suited for applications with limited computational resources.