A pilot study assessing ipsilateral vs. contralateral feedback in EMG-force models of the wrist for upper-limb prosthesis control

Many advanced EMG-based upper-limb prosthesis control methods require model training in which subjects produce supervised forces/movements. Since unilateral limb-absent subjects cannot produce forces/movements on their affected side, one technique (mirrored bi-lateral training) relates forces/motions produced on the sound side to EMG on the affected side. However, the efforts made by the phantom limb may not fully reflect those of the sound limb. To understand this issue, three able-bodied subjects produced mirrored bi-lateral forces during constant-posture contraction at the wrist. EMG-force models were formed for 1- and 2-degree of freedom tasks and results compared to previous trials in which ipsilateral training had been conducted. We found that contralateral training generally, but not always, produced errors (in percent maximum voluntary contraction) that were 6-56% larger than those found from ipsilateral training. Our results suggest that a substantial portion-but not all-of the errors found in mirrored tasks may be due to contralateral tracking errors. Further study with a larger population is indicated.