PhD Dissertation Defense: Madeline Hays

Title: Experimental Evaluation of High-Density Neural Recording with Compressive Readout for Bidirectional Implants

Abstract: High-density neural interfaces for restoration and augmentation of sensory or motor functions require simultaneous recording from thousands of neurons with single-cell resolution. However, safety requirements limit power density, making high-throughput data transmission a major obstacle for bidirectional neural implants. The wired-OR compressive readout scheme addresses this problem by exploiting sparsity of neural activity to achieve ultra-low-power compression, but its ability to preserve physiologically relevant neural signals has not been experimentally established. Here we evaluated wired-OR compression for neural waveform preservation, single-cell resolvability, and cell-type identification in ex vivo retinal recordings. Spontaneous activity from macaque retina was recorded using a 1024-channel, 32x32 microelectrode array with 36 μm pitch, reaching compression ratios up to 150x. Spike waveforms and rates obtained from wired-OR compressed recordings were well matched to those obtained from full bandwidth (20 kHz sampling) recordings. Using simple spike sorting methods, individual neurons were resolved from the wired-OR recordings, producing spatiotemporal maps that included somatic, dendritic, and axonal compartments. These compressed spatiotemporal maps and the individual neurons’ spike autocorrelation functions were comparable to those obtained under full bandwidth conditions. Together with other key physiological features, such as inter-spike interval and axon conduction velocity, neuronal populations of distinct cell types were identified. These results demonstrate for the first time that the highly compressive wired-OR approach preserves neural signals for effective spike sorting and downstream cell-type classification in physiological recordings. This approach could prove broadly useful for bidirectional neural implants seeking single-cell resolution and control at the scale of thousands of electrodes.

Please contact Sofia Rakicevic-More for the Zoom link.