Spike Detection Performance of a Noise-Enhanced Nonlinear Filter Across Varied Neuron Distances

The objective of this study is to assess the spike detection performance of a nonlinear filter (NF) using stochastic resonance and subsequent hard thresholding. The evaluation is based on the distance of neural spikes from the electrode. The F1 score variation of the algorithm is examined using two synthetic datasets generated with MEArec and Neurocube extracellular neural recording synthesis tools. In the MEArec recording, the algorithm achieved a 75% F1 score for spikes up to 70 μm, and in the Neurocube recording, it reached 95 μm. This contrasts with three popular unsupervised intracortical neural spike enhancement algorithms, which achieve the same F1 performance only within a 50 μm range. These results suggest that the NF spike enhancement method has the potential to increase the detectable spikes by a single electrode in resource-limited implantable intracortical neural recording systems, where hardware complexity is a critical design consideration.