Automated Spike Sorting in Microneurography: A Proof-of-Concept through Classification on Ground Truth Data
Troglio Alina, Fiebig Andrea, Kutafina Ekaterina, Barbara Namer · 2024
Microneurography is an electrophysiological method to examine, e.g., nociceptive nerve fibers related to pain and itch signaling.For microneurography data, spike sorting remains a critical challenge.Recent advances in automatic spike (action potential) sorting have limited application in the setup of in-vivo single electrode recordings with low signal-to-noise ratios.Part of the evoked spikes can usually be reliably sorted due to repetitive stimulation and the marking method, which time-locks a subset of spikes.In this work, we are analyzing the potential of using those sorted spikes as labeled train data for the automatic classification of other remaining spikes.To assess the results, we apply a special electrical stimulation protocol, which allows us to time-lock all spikes.This protocol simulates the usual sorting problem but allows quantifying the accuracy of classification by providing labels for the test set.Using support vector machine (SVM) classification, our approach achieves an accuracy of 0.96 in categorizing action potentials.