 |
 |
|
|
Ofer Mazor, Christophe Pouzat, and Gilles Laurent
Automated model-based spike-sorting: a method for classification with confidence intervals
California Institute of Technology
The use of extracellular recordings requires objective and reliable spike-sorting procedures as a first stage of data processing. To apply such techniques on a large scale (hours of recording tens of electrodes), automatic spike-sorting methods are needed that provide an estimate of the quality of the classification they produce. We propose a novel, simple, and expandable procedure for classification and estimation based on a probabilistic model of data generation. A key feature of this approach is that it estimates the statistical properties of the background noise inthe recording. Using this information, it automatically generates a model of the underlying units, sorts spikes (including overlapping spikes), and cancels sampling jitter, all based on objective measures. Finally, it provides the experimenter with quantitative tests to assess the quality of the classification. The routines and methods we propose (or ones derived from them) could form the basis for a standardization of extracellular spike waveform analysis, providing both experimenter and reader with objective means to evaluate the primary data and facilitating the widespread use of these recording techniques.
|
|
 |
| |
The Swartz Foundation is on Twitter: SwartzCompNeuro
|
|
|
 |
print