Dear all,
Is there a Spike Sorting revolution taking place? There is always progress and new ideas on how to sort electrophysiological spikes in order to study single unit activity. I have the feeling that there is something new going on lately, maybe driven by the increase in channel number of silicon probes. Just to refer to two papers:
Yger, P. et al. Fast and accurate spike sorting in vitro and in vivo for up to thousands of electrodes. bioRxiv 067843 (2016).
Pachitariu, M., Steinmetz, N., Kadir, S., Carandini, M. & Harris, K. D. Kilosort: realtime spike-sorting for extracellular electrophysiology with hundreds of channels. bioRxiv 061481 (2016). doi:10.1101/061481
What other platforms am I missing with such multi-channel capabilities?
Assuming I have a standard set of recordings: a couple of hours per recordings, with 100 channels giving me unit data, ...:
Which one is more accurate with less human input? This is, which one would require less human check, in your experience.
Which one is easier to set up with some Python and Matlab knowledge?
Are these platforms easily 'downscalable' to 16 of wired data (single wire or tetrode)? Or not worth the effort?
Thank you in advance.
A few keywords for our friendly search engines: Neuroscience, electrophysiology, single-unit, spike-sorting.
Dear Roger L Redondo
Read this paper may be help you
"Fast and accurate spike sorting of high-channel count
probes with KiloSort"
for Marius Pachitariu
Dear Roger. The two algorithms are sharing similar principles, and I would say would obtain similar performances. One is using GPU and Matlab (KiloSort), the other several CPUs and Python (SpyKING CIRCUS). The choice is up to you then :-)
Indeed, there is some sort of a "revolution" in the spike sorting community, with the design of new algorithms, using template-matching based approaches instead of massive clustering. Speaking about SpyKING CIRCUS, that I obviously know more, it will downscale fine. In fact, this is even working for a single channel, and because the algorithm is able to solve the problem of overlapping spikes, it will be better than naive clustering approaches. We wrote a full and exhaustive documentation there http://spyking-circus.rtfd.org.
Regarding the manual curation, there would still be some required at some point. SpyKING CIRCUS may produce less templates than KiloSort, and thus slightly less manual burden, but still, you'll have to review the results manually at this stage. The good point if that both algorithms are using the same visualization software (phy), such that it is easier for the user to compare/appreciate differences.
Let me know if you have more specific questions,
Best
Pierre
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python