Algorithm and Validation Method for Spike Sorting Based on Wavelet Analysis and a Genetic Algorithm

Spike sorting is a fundamental task in neuroscience that sets the basis for many neurophysiology studies. Although different solutions are available, the problem of classifying spikes generated by an unknown number of neurons is still not fully solved. Furthermore, the relevance of the task makes the quest for new, computationally efficient, spike sorting algorithms highly current. In the present work, we present a spike sorting algorithm that utilizes a wavelet transform to characterize spikes and a genetic algorithm to classify them into an unknown number of clusters. To validate our algorithm, we use microelectrode recordings acquired by stereotactic neurosurgery in rats. We employ spikes obtained from clear single-cell recordings from different animals to build artificial neuronal signals with randomly mixed spikes originated by different neurons. The spike sorting algorithm proposed is computationally efficient and robust. Besides, our solution could be useful for dealing with the separation of signals of an unknown number of sources, different than spikes, a classic signal analysis problem with a myriad of applications in diverse fields.