Physical Disturbance Rejection Methods in Myokinetic Control of Prosthetic Limbs.

Abstract

The myokinetic control interface for limb protheses involves implanting small permanent magnets in the residual muscles of the stump, retrieving their displacement induced by muscle contraction using external magnetic sensors, and mapping such displacements to motor commands. We have previously shown the feasibility of tracking several magnets implanted in an anatomically relevant workspace. However, to clinically translate the interface, strategies to address different kinds of external disturbances compromising its functioning are mandatory. Among these, physical disturbances, viz. relative displacement between the sensors and the magnets not due to voluntary contraction, could significantly hinder the interface usability. Here we propose three rejection methods sought to mitigate this problem: one aims to realign the sensors and the magnets when a disturbance is detected, through numerical approximation methods; the others exploit differential measurements. We applied these concepts to upper limb prostheses, by mimicking the presence of physical disturbances in simulations and through a physical setup, reproducing four target muscles of a human forearm to be potentially implanted. Finally, we evaluated the rejection ability of one of those methods during the first-in-human implementation of the myokinetic interface. All methods proved capable of rejecting the disturbances, showing median localization errors below 10% the displacement undergone during contraction. Results suggest that the optimal rejection method is application-specific, and provide hints for assessing different factors influencing the best choice. Other than being crucial for the myokinetic interface development, these outcomes also provide interesting insights for many biomedical applications exploiting remote magnetic tracking.