Spike sorting in real time

logo_spykingTogether with Olivier Marre, we developed a code to address the problem of Spike Sorting for large Multi Electrode Arrays, inspired from the following publication, and name "SpyKING CIRCUS". By using a greedy template-matching approach, and state-of-the art clustering techniques, the code is able, using multi CPU and GPUs, to run in realtime on a small number of nodes. For example, 252 channels with a total of 500 templates (so putative cells) extracted from recordings are sorted in realtime with 5 nodes, each of them with a GPU. Scaling is straightforward, linear, and not bounded by any memory issue. Templates are extracted once for all at the beginning of the sorting, from the whole dataset, and then fitted on-the-fly to the data. If you are interested, please contact us



To know more about it, see the Following Section