Spike sorting with Gaussian mixture models

Abstract

The shape of extracellularly recorded action potentials is a product of several variables, such as the biophysical and anatomical properties of the neuron and the relative position of the electrode. This allows isolating spikes of different neurons recorded in the same channel into clusters based on waveform features. However, correctly classifying spike waveforms into their underlying neuronal sources remains a challenge. This process, called spike sorting, typically consists of two steps: (1) extracting relevant waveform features (e.g., height, width), and (2) clustering them into non-overlapping groups believed to correspond to different neurons. In this study, we explored the performance of Gaussian mixture models (GMMs) in these two steps. We extracted relevant features using a combination of common techniques (e.g., principal components, wavelets) and GMM fitting parameters (e.g., Gaussian distances). Then, we developed an approach to perform unsupervised clustering using GMMs, estimating cluster properties in a data-driven way. We found the proposed GMM-based framework outperforms previously established methods in simulated and real extracellular recordings. We also discuss potentially better techniques for feature extraction than the widely used principal components. Finally, we provide a friendly graphical user interface to run our algorithm, which allows manual adjustments.

Beta2 Oscillations in Hippocampal-Cortical Circuits During Novelty Detection

“…Then, waveforms were detected using a threshold of eight times the median absolute deviation as in Quiroga et al ( 2004 ) and aligned by their interpolated peak. We used the wavelet and weighted-PCA approach described in Souza et al ( 2019 ) to automatically sort the waveforms of each channel. Although we could not separate spiking activity into single units, the different MUA clusters found in the same channel presented unique activity patterns, and we, therefore, analyzed their activity separately.…”

Compatibility Evaluation of Clustering Algorithms for Contemporary Extracellular Neural Spike Sorting

“…With such a wide range of algorithms reported, the selection of a particular algorithm is usually subjective Kaufman and Rousseeuw ( 2008b ). Some of the algorithms widely adopted for the purpose of spike sorting are K-means (Pachitariu et al, 2016 ; Caro-Martín et al, 2018 ), GMM (Souza et al, 2019 ), SPC (Blatt et al, 1996 ; Rey et al, 2015 ; Niediek et al, 2016 ), Klustakwik (Rossant et al, 2016 ), and methods based on statistical aggregations such as t-distributions or chi 2 distribution (Harris et al, 2000 ; Shan et al, 2017 ). We reviewed 58 algorithms out of which 27 are presented in the report.…”

Low-cost and versatile electrodes for extracellular chronic recordings in rodents

“…We filtered the signal between 300Hz and 6000Hz and used the threshold of 8σn ( Quiroga et al., 2004 ). The spike sorting was performed using Gaussian mixture models in two steps, extracting relevant features (principal components, wavelets) and GMM fitting parameters (e.g., Gaussian distances - Souza et al., 2019 ). We use the graphical user interface provided by Souza et al.…”