Closed-Loop System for the Selective Stimulation of Retinal Ganglion Cells
This research elucidates novel methodologies in retinal stimulation, emphasizing high-density circuit designs, innovative electrical stimulation strategies, and adaptive, closed-loop evaluation techniques for retinal implants.
Exploring high-density electrical stimulation, sinusoidal electric currents in the kilohertz range emerge as a promising candidate for targeted stimulation of retinal ganglion cells (RGCs). The introduced Global-Cluster circuit topology, designed to optimize chip area utilization, facilitates a high-density channel array proficient in delivering sinusoidal waveforms. Within this project, two prototypes of the stimulation System on Chip (SoC) are designed and fabricated. The first one, spanning an area of 3.92 mm2, incorporates 42 output channels. In a second development stage, an optimized SoC with 1,224 output channels over an area of 10 mm2 is realized. The resulting pixel pitch is 60 μm. The stimulation strategy empowers modulation between activation and suppression for ON- and OFF-cells on individual electrodes. Notably, sinusoidal stimulation in this frequency and amplitude range elicits prolonged effects on retinal tissue, significantly influenced by glycine receptors, invoking network-mediated responses predominantly via amacrine cells.
In adaptive retinal stimulation, the research adopts a closed-loop paradigm anchored in deep learning. A fully connected neural network (FCNN), trained on labeled data from multi-spike and single-spike recordings, serves as the foundation for the cell-type dependent classification of retinal recordings. The approach bifurcates into two classification strategies: a Basic method aimed at broader RGC superclasses and a Specific method delineating distinct RGC subtypes. The application of this classifier allows for the implementation of a correlator between stimulation parameters and recorded neural responses.
Collectively, the advancements presented here fortify the foundation for future retinal prosthetics. By harmonizing advanced circuit designs, refined stimulation strategies, and adaptive classification techniques, the research heralds enhanced outcomes for retinal implants, underscoring prospects for heightened and more natural perceptions for patients.