QMCVNCFeb 10, 2016

Comparison of feature extraction and dimensionality reduction methods for single channel extracellular spike sorting

arXiv:1602.03379v1
AI Analysis

This work addresses incremental improvements in spike sorting accuracy for neuroscience researchers, focusing on optimizing existing methods.

The study compared various feature extraction and dimensionality reduction methods for spike sorting in extracellular recordings, finding that PCA with 46-55 features from 64-sample waveforms at 24 kHz performed best (p < 0.001).

Spikes in the membrane electrical potentials of neurons play a major role in the functioning of nervous systems of animals. Obtaining the spikes from different neurons has been a challenging problem for decades. Several schemes have been proposed for spike sorting to isolate the spikes of individual neurons from electrical recordings in extracellular media. However, there is much scope for improvement in the accuracies obtained using the prevailing methods of spike sorting. To determine more effective spike sorting strategies using well known methods, we compared different types of signal features and techniques for dimensionality reduction in feature space. We tried to determine an optimum or near optimum feature extraction and dimensionality reduction methods and an optimum or near optimum number of features for spike sorting. We assessed relative performance of well known methods on simulated recordings specially designed for development and benchmarking of spike sorting schemes, with varying number of spike classes and the well established method of $k$-means clustering of selected features. We found that almost all well known methods performed quite well. Nevertheless, from spike waveforms of 64 samples, sampled at 24 kHz, using principal component analysis (PCA) to select around 46 to 55 features led to the better spike sorting performance than most other methods (Wilcoxon signed rank sum test, $p < 0.001$).

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes