Objective: We aim to increase the number of neural spikes that can be detected in a single channel extracellular neural recording.
Approach: We propose a pre-emphasis method facilitating stochastic resonance (SR), where we introduce the band-pass-filtered noisy extracellular recording to an overdamped Brownian particle in a monostable well. The x-position of the Brownian particle is the output of the proposed pre-emphasis method. Threshold is applied on the output for spike detection. To characterize the dynamics and the solution of the system, we use a synthetic dataset generated by adding Gaussian white noise at different intensities to an intracellular recording. Then, we evaluate and compare the spike detection performance of the proposed method on a public synthetic extracellular dataset.
Main results: The proposed SR-based spike detection improves the signal-to-noise ratio of the intracellular-based synthetic dataset as much as 7.35 dB and outperforms the state-of-the-art pre-emphasis methods in false positive and false negative rates in 15 of the 16 synthetic extracellular datasets, with 100% sensitivity and positive predictivity values in seven of the recordings.
Significance: The method has the potential of significantly increasing the number of neurons that can be monitored from a single-channel extracellular recording.