Characterisation of Spatial Correlation of High Density Intracortical Neural Recordings

Abstract

High-density electrophysiology recordings such as Neuropixels provide rich spatio-temporal neural data but impose stringent data-rate constraints. These multichannel recordings exhibit pronounced spatial structure whose contribution to compressibility remains poorly quantified. In this work, we present a general analysis of spatial information in in vivo data. Using spatial singular value decomposition and rate-distortion analysis, we characterise the low-rank structure of multichannel signals under different preprocessing and masking regimes. Comparisons with delta encoding schemes reveal that temporal correlations maintain low-rank behaviour, whereas spatial differencing primarily acts as a decorrelating (whitening) operation. These findings can inform future developments of compression algorithms for neural data.