Multiresolution bayesian detection of multiunit extracellular spike waveforms in multichannel neuronal recordings.

We describe multiresolution Bayesian tests for spike detection in multielectrode recordings. We derive results for single channel and multi electrode data, and show that the use of the array model substantially improves the detection performance. The effect of signal and noise spatial correlation characteristics is also discussed. Our approach focuses on blind signal detection without any assumptions about the underlying signal parameters. It is therefore suitable for chronic recordings with electrode arrays where typically temporal and spatial nonstationarity of the extracellular waveforms complicates the estimation of the true action potential.