Investigating well potential parameters on neural spike enhancement in a stochastic-resonance pre-emphasis algorithm

Area of Science:

• Neuroscience
• Signal Processing
• Biophysics

Background:

• Background noise in extracellular neural recordings impedes reliable spike detection.
• Limited spike detectability hinders the advancement of neuroscientific research and technologies.
• Stochastic resonance (SR) offers a potential method to enhance weak signal identification in noisy conditions.

Purpose of the Study:

• To investigate the application of stochastic resonance (SR) for improving neural spike detectability.
• To analyze the impact of potential well shape and damping status on signal-to-noise ratio (SNR) improvement.
• To evaluate spike detection performance using SR-based methods compared to existing algorithms.

Main Methods:

• Modeled an SR-based pre-emphasis algorithm using a particle in a 1D potential well.
• Compared SNR enhancement across different well shapes (shallow/steep monostable, bistable) and damping statuses (overdamped/underdamped).
• Assessed spike detection performance via thresholding on optimized SR configurations using synthetic datasets.

Main Results:

• Optimal SR configuration depends on input noise levels.
• The underdamped shallow-wall monostable well significantly improved SNR (over four orders of magnitude) in low-noise conditions.
• This configuration surpassed state-of-the-art methods in spike detection sensitivity and positive predictivity on a public dataset.
• For higher noise intensities, the overdamped steep-wall monostable well demonstrated superior spike enhancement.

Conclusions:

• SR, particularly with specific well-shape and damping parameters, can substantially enhance neural spike detectability.
• The noise-dependent SNR improvement suggests potential for designing adaptive detectors sensitive to electrode proximity.
• This approach holds promise for improving downstream analyses like spike sorting in neural recordings.