Identifying number of neurons in extracellular recording

One of the most difficult aspects of spike sorting is choosing the number of neurons in extracellular recording. The paper proposes a methodology for estimating the number of neurons based on the Gaussian mixture model. The following criteria have been examined: Bayesian selection method, Akaikes information criteria, minimum description length, minimum message length, fuzzy hyper volume, evidence density and partition coefficient. In order to validate the procedure, an experimental comparative study was carried out, comparing the proposed methodology with three spike sorting algorithms. The proposed methodology has an advantage of setting the minimum number of parameters and is very robust to background noise. We conclude that only fuzzy hyper volume and evidence density criteria are able to identify the correct number of neurons across different noise levels.