Advances in Denoising Spikes Waveforms for Electrophysiological Recordings

The processing of neural signals faces significant challenges, particularly within the framework of electrophysiological recordings involving electrical stimulation. Extensive efforts have been dedicated to automating this analysis, but several aspects, such as artifact removal, still require further improvements. To address some of these challenges, we have devised a novel strategy based on waveform clustering, spike detection, artifact template matching, and advanced filtering techniques. This approach facilitates an automated workflow for neural denoising and sorting, thereby diminishing human bias and enabling automatic processing within a reasonable timeframe. To assess the efficacy of our algorithms, we conducted tests on electrophysiological recordings acquired during experiments involving electrical stimulation of the human visual cortex. The results demonstrate a strong overall performance, as evidenced by robust F1 scores. These newly developed tools are available as open source on Github.