The new version (in progress) of SpikeOMatic. A free software for spike sorting.

The Page of the New SpikeOMatic

SpikeOMatic is a R package which also uses GGobi to do Spike Sorting.

Some key features of SpikeOMatic which are in fact mainly R and GGobi features are:

All classical clustering algorithms of the statistical literature (eg, kmeans, Gaussian Mixtures with EM, Bagged Clustering, etc) are available thanks to R default and contributed clustering packages . As soon as I will be done with these "classical" methods, I will add the Markov Chain Monte Carlo ones.
Some functions are already parallelized thanks to the snow (Simple Network of Workstations) package (an R feature again).
Powerful dynamic and multi-dimensional visualization with GGobi. Yes, you can do clustering also by hand if you want and much more than that.
It runs on PCs (Linux and Windows) and Mac (and more).
SpikeOMatic is released under the Gnu General Public License except for two functions: clusterEMclust and clusterEMclustN which are parallel versions of mclust EMclust and EMclustN functions and are therefore released under the same license agreement. Many thanks to Chris Fraley for authorizing me to distribute clusterEMclust and clusterEMclustN.
There is a tutorial. Take a look at it to get an idea of what SpikeOMatic is.

I am currently re-writing the software. What you can find here is a snapshot of work in progress:

April 7 2008. A bug found by Keisuke Kawasaki in function getMPP has been fixed. Function getST is now compatible with the new STAR version.
The source code in form of a R vignette (last update April 7 2008).
The Reference Manual in PDF format (last update April 7 2008).
The new Tutorial (also in PDF), that's where you should start (last update March 22 2006).
The Tutorial vignette if you want to regenerate the tutorial (you have to create a sub-folder named "figures" in the folder from which R is running). WARNING: few changes between R-2.2.1 and R-2.3.0 induce a difference in the number of spikes detected using the two versions (6 spikes among more than 6000). That is no big deal but it means you will not get exactly the same figures if you run version R-2.3.0 (version R-2.2.1 was used to write the tutorial) and you won't be able to automatically regenerate the tutorial from the vignette. I will prepare a new vignette for R-2.3.0 as soon as possible. Sorry for the inconvenience.
The Purkinje cell recording used in the tutorial.
The same data in GGobi XML format:

Rsweeps.xml (after spike detection and feature space dimension reduction)
RsweepsSorted.xml (after manual sorting)
RsweepsKM.xml (after automatic sorting with kmeans)

The tutorial explains how to generate all these files.

The main reason to use GGobi is that it allows you to see and interact with high dimensional data as illustrated by this short movie. The tutorial explains how to go from the raw Purkinje cell recording to this point.

Thanks to Robert Bernier for posting instructions on how to make such movies on Linux.

Back to C Pouzat Home Page Retour à la page titre de C Pouzat