Evaluation of Spike Sorting Algorithms: Application to Human Subthalamic Nucleus Recordings and Simulations

Abstract

An important prerequisite for the analysis of spike synchrony in extracellular recordings is the extraction of single unit activity from the recorded multi unit signal. To identify single units (SUs), potential spikes are first extracted and then separated with respect to their potential neuronal origins ('spike sorting'). However, different sorting algorithms yield inconsistent unit assignments which seriously influences the subsequent analyses of the spiking activity.To evaluate the quality of spike sortings performed with different algorithms (K-Means, Valley-Seeking, Expectation Maximization) offered by a standard commercial spike sorter ('Plexon Offline Sorter') we first apply these algorithms to experimental data (ED), namely recordings in the Subthalamic Nucleus of patients with Parkinson's disease, obtained during Deep Brain Stimulation surgery. Since this procedure leaves us unsure about the best sorting method for the ED we then apply all methods again to artificial data (AD) with known ground truth (GT). AD consists of pairs of SUs and perturbation signals embedded in the background noise adapted to mimic the ED. We generate four AD sets which differ in the similarity of embedded SU shapes.The spike sorting evaluation is performed in terms of the influence of the respective methods on the SU assignments (e.g., number of units) and its effect on the resulting firing characteristics (e.g., firing rates). We find a high variability in the sorting results obtained by different algorithms that increases with SU shape similarity. dardized validation procedures for spike sorting based on GT data. The recipe suggested here is simple enough to become a standard procedure. It provides a good basis for the evaluation of spike sorting results in order to ensure reliability and reproducibility.

Diversify your workflow! - An inconvenient suggestion to analyze spike data from intracranial recordings

“…However, seemingly each research group determines the set of parameters needed by its match with the purpose of their study depending on their historical experience. Thus, results for the same data set show human error and inconsistencies [14 – 15]. As a result, a method that has been shown to be successful in one lab may yield less successful results in another study [14 – 15].…”

“…Thus, results for the same data set show human error and inconsistencies [14 – 15]. As a result, a method that has been shown to be successful in one lab may yield less successful results in another study [14 – 15]. Therefore, validating the reliability of spike analyzing methods is bound by ambiguities and prevents the establishment of a Gold-standard method minimizing ambiguity and enabling consistency and reliability.…”

An Accurate and Robust Method for Spike Sorting Based on Convolutional Neural Networks

“…That is, it is necessary to determine which spike corresponds to which neuron, because each electrode records the extracellular field in its vicinity and can detect the spikes emitted by several neurons adjacent to the electrode site [6][7][8][9]. This identification is universally termed spike sorting [10][11][12], which is conducted on the basis of the similarity of spike waveforms. The reason lies in the fact that each neuron fires spikes of a particular shape, depending on the morphology of its dendritic tree, its distance and orientation relative to the recording site, etc.…”

An Adaptive Detection for Automatic Spike Sorting Based on Mixture of Skew-t distributions

“…Spike sorting algorithms usually consist of three main steps 1 . The first step is activity detection, where spikes are extracted from the bandpass filtered signal.…”

“…The first step is activity detection, where spikes are extracted from the bandpass filtered signal. A common way to detect spikes is to determine activities that cross a threshold 1 . The threshold is usually calculated based on the estimated noise variance, without considering dynamical changes of recorded signal, which causes spike loss or false alarms.…”