PASER for automated analysis of neural signals recorded in pulsating magnetic fields

Abstract

Thanks to the advancements in multichannel intracranial neural recordings, magnetic neuroimaging and magnetic neurostimulation techniques (including magnetogenetics), it is now possible to perform large-scale high-throughput neural recordings while imaging or controlling neural activity in a magnetic field. Analysis of neural recordings performed in a switching magnetic field, however, is not a trivial task as gradient and pulse artefacts interfere with the unit isolation. Here we introduce a toolbox called PASER, Processing and Analysis Schemes for Extracellular Recordings, that performs automated denoising, artefact removal, quality control of electrical recordings, unit classification and visualization. PASER is written in MATLAB and modular by design. The current version integrates with third party applications to provide additional functionality, including data import, spike sorting and the analysis of local field potentials. After the description of the toolbox, we evaluate 9 different spike sorting algorithms based on computational cost, unit yield, unit quality and clustering reliability across varying conditions including self-blurring and noise-reversal. Implementation of the best performing spike sorting algorithm (KiloSort) in the default version of the PASER provides the end user with a fully automated pipeline for quantitative analysis of broadband extracellular signals. PASER can be integrated with any established pipeline that sample neural activity with intracranial electrodes. Unlike the existing algorithmic solutions, PASER provides an end-to-end solution for neural recordings made in switching magnetic fields independent from the number of electrodes and the duration of recordings, thus enables high-throughput analysis of neural activity in a wide range of electro-magnetic recording conditions.