Spike sorting is the process of detecting and clustering action potential waveforms from extracellular voltage recordings to identify spikes of putative single neurons. Typically, spike detection is done using a fixed voltage threshold and shadow period, but this approach can lead to missed spikes during high firing rate epochs or noisy conditions. We developed a novel spike detection method utilizing a computationally simple form of template matching that efficiently detects spikes from candidate single units and is tolerant of high firing rates and electrical noise without a whitening filter. Template matching was based on a sliding cosine similarity between mean spike waveforms of candidate single units and the extracellular voltage signal. Performance was tested in whisker somatosensory cortex (S1) of anesthetized mice in vivo. The method consistently detected whisker-evoked spikes that were missed by a standard fixed voltage threshold. Detection was most improved for spikes evoked by strong stimuli (40-70% increase), less improved for weaker stimuli, and unchanged for spontaneous spiking. This reflected the failure of standard detection during spatiotemporally dense spiking. Template-based detection revealed higher signal-to-noise ratio for sensory responses and sharper sensory tuning. Thus, this template matching method (and other model-based spike detection methods) critically improve the quantification of single-unit spiking activity.
DNN‐aided read‐voltage threshold optimization for MLC flash memory with finite block length