Abstract
Neural spike detection and classification, or spike sorting, is the first and a critical step prior to any single unit based neuroscientific studies and applications. A good spike sorter is usually characterized by high detection and classification accuracy, robust to changes in signal-to-noise ratio, objectivity in detection results or less user dependency, and real-time applicability. Here we present an automatic and robust spike detection and classification system, the M-Sorter, based on the multiple correlation of wavelet coefficients (MCWC) detection algorithm in conjunction with template matching for classification. Unlike many existing spike sorters that make use of a series of complex spike classifiers to deal with the challenges resulted from a low performance spike detector, the M-Sorter relies on a high performance yet computationally efficient detection algorithm and thus a simple classifier suffices to generate high quality spike sorting results. In this paper we provide step by step implementation procedures of the M-Sorter, and compare its performance with other popular sorters.
Original language | English (US) |
---|---|
Pages (from-to) | 281-290 |
Number of pages | 10 |
Journal | Journal of Neuroscience Methods |
Volume | 210 |
Issue number | 2 |
DOIs | |
State | Published - Sep 30 2012 |
Keywords
- Classification
- Detection
- K-Means
- MCWC
- Neural waveform
- Sorting
- Spike
- Template Matching
ASJC Scopus subject areas
- General Neuroscience