A wavelet approach for on-line spike sorting in tetrode recordings.

A new method for spike sorting of tetrode recordings during data acquisition is introduced. For each tetrode channel, putative spikes are detected by means of a threshold, and then convolved with a cascade of wavelet filters. These transformed putative spikes are averaged and this average is used as a matched filter to find portions of signals that are likely to contain a spike. A collection of vectors containing the correlation coefficients between putative spikes and the matched filters is then clustered using K-Means. Centroids of the resulting clusters contain enough information to sort spikes recorded by all tetrode channels simultaneously. On-line sorting is achieved by measuring euclidean distance between putative new spikes and the cluster centroids.