Decoding Upper-Limb Movements via Inter-Muscle Displacement: From Synthetic to Clinical Data in a Myokinetic Interface

Recent advances in technology and surgery have led to novel interfaces for the seamless control of robotic prostheses. Among them, the myokinetic interface decodes muscle activity by tracking permanent magnets implanted in the muscles. The first clinical implementation of a myokinetic prosthesis demonstrated the potential of this approach, but also highlighted that both the number and placement of the magnets in the muscles are crucial factors for the stability of the control signals. Although multiple magnets per muscle can potentially enhance information content and robustness to limb position effects, they also increase surgical complexity and computational load, suggesting that a trade-off may be beneficial. In this study, we evaluate the feasibility of classifying multiple upper limb movements using a single magnet per muscle (1MM), by measuring inter-muscle displacement during contraction. The classification accuracy from this setup is compared to that of the 2MM approach, which uses two magnets per muscle and leverages both intra- and inter-muscle displacement through all magnet pairs. Experiments are first conducted on a forearm mockup, selecting three muscles and simulating 1MM and 2MM implants and multiple contraction patterns, and then validated using clinical data from a participant with transradial amputation. Results demonstrate that, while a greater number of implanted magnets increases classification accuracy, a single magnet per muscle is sufficient to accurately discriminate multiple grasps, providing information comparable to intra-muscle displacement. These findings are key for optimizing the design of myokinetic interfaces and other technologies that rely on implanted markers.