Statistical Description of Electrophysiological Features of Neurons across Layers of Human Cortex

Abstract

Modem deep learning technologies have helped scientists and engineers to solve many difficult problems in pattern recognition and decision making. Existing deep learning architectures face an important issue of high power dissipation. As an alternative solution spiking neural networks (SNN) are being recently developed by researchers with promise to overcome the bottleneck of power loss. An SNN is primarily inspired from a biological neuronal network consisting of real neurons. However, state of the art SNN models consist of layers with identical neurons which is not true in human cortex.In this study, a statistical approach is presented to show that the layers of a human cortex differ in terms of their member neurons. Neurons of different types are not identically distributed across the layers of a human cortex. Neurons are classified on the basis of key electrophysiological parameters such as spiking frequency, inter spike intervals, height and width of action potentials and energy dissipated per spike.The rich database provided by Allen Institute for Brain Science is used for the present study. It has been shown that spiking phenomenon is sparse across all the layers of a human cortex in temporal domain.