Using Johnson's SU distribution for modeling the background activity in extracellular neural recordings

Abstract

Extracellular neural recordings obtained from awake behaving subjects through chronically implanted microelectrode arrays provide information about the functioning of the brain with sub-millisecond temporal resolution at the level of individual neurons. After bandpass filtering in a frequency range suitable for spike detection, these recordings consist of spikes and background activity. Methods exist to segment the background activity automatically using truncation thresholds and Otsu-based methods. In previous work, truncation thresholds have been computed using the truncated Normal distribution. Here, we use the truncated Johnson's SU distribution instead to examine whether it segments the background activity better. We also find that the truncated Johnson's SU distribution explains the background activity segmented by Otsu-based thresholds. These results are useful for developing invasive brain-computer-interfaces that automatically extract information from extracellular neural recordings in real time.