4th Annual Mountain West
Biomedical Engineering Conference

IntraFascicular Multi-electrode Stimulation (IFMS) has been suggested as a means to restore fatigue-resistant, controllable motor function in spinal cord injured patients. IFMS using multi-electrode arrays can selectively target small motor unit groups within a single muscle. Interleaved stimulation of different motor units via multiple, independent electrodes can provide fatigue-resistant, tremor-free muscular force. Initial physiological experimentation with classical closed-loop control strategies demonstrated reasonable tracking of time-varying force trajectories for single-electrode stimulation. However, to achieve time-varying desired force trajectories with minimal tremor for IFMS, the time-varying stimulation amplitudes and phases of currents delivered via each electrode must be optimized to obtain accurate feedforward estimates. This was achieved using a linear model of isometric force production based on data from feline gastrocnemius muscle: single-pulse stimulation from each of six electrodes evokes a unique muscle twitch, and these summate linearly. The model uses a gradient-descent adaptive filtering method to optimize the amplitude and phase of stimuli for each electrode to achieve a minimal tremor, target force trajectory. The integral of the squared error between desired force and estimated force was used as the cost function in this optimization. Step, sine, and triangle trajectories, as well as composite periodic functions, were set as the desired force profiles. Utilizing a six-electrode, 36-Hz initial composite stimulation frequency interleaved scenario, simulations showed that desired steps in force can be achieved with rapid rise times (<0.5 s), low overshoot (<6%), low steady state tremor (<5%), and rapid steady state convergence of stimulation parameters (<0.8 s). Simulations also showed that the adaptive algorithm can converge on an optimal solution for periodic functions within a very limited number of cycles (<5 cycles for a desired 2-Hz sine wave trajectory). The optimized solution for periodic functions showed low amplitude error (<10%) and low phase delay (<10% of cycle period). This optimized multi-electrode stimulation method is being experimentally verified and will be incorporated into a closed-loop, feedback control method.