Artificially-Intelligent Fascicle-Selective Bidirectional Peripheral Nerve Interfaces.

This paper presents a comprehensive overview of our latest work on two design strategies for minimally invasive, battery-free wireless neural interfaces for adaptive neuromodulation therapy in the peripheral nervous system (PNS). These active PNS neural interfaces provide alternative treatment options for patients with motor, sensory and other neurological deficits, aiming to improve their quality of life. Each of the two presented designs includes a nerve-scale integrated circuit (IC) performing fascicle-selective extraneural recording and stimulation. For fascicle-selective neural recording, the IC interacts wirelessly with a wearable interrogator which performs machine-learning inference on the recorded neural signals. For fascicle-selective neuro-stimulation, two techniques - subthreshold current pulses and temporal interference stimulation methods are adopted. Two oversampling neural ADC architecture choices are also compared: an energy-efficient passive 2nd-order delta-sigma ADC and a high-resolution noise-shaping SAR ADC. Inductive and ultrasound energy harvesting for wireless powering and data reception are also discussed. In vivo results from rodent studies are included to support the validity of the discussed design strategies.