A novel unsupervised spike sorting implementation with variable number of features

We propose a new fully automatic spike sorting algorithm that is able to match, or even improve, the performance of semiautomatic solutions with supervised intervention from expert users. We achieved this by incorporating: 1) a set of heuristic criteria inspired by the expert actions following the solution from semiautomatic algorithms, and 2) an improved feature selection method that increases the number of units that can be isolated from a single electrode recording. We evaluated the performance of the proposed method with real and simulated data.