Abstract

The development of the computer technology and cell recording techniques have made possible continuous recording and sorting of extracellulary-recorded action potentials. At the same time, a reliable sorting of action potentials became a significant focus of interest. In this study, we describe and summarize a method of sorting of extracellulary-recorded spikes based on the Principal Component Analysis(PCA). In this method, a strategic number of principal components are chosen and each cell is represented in the feature space formed by the chosen components. The number of data cluster(i.e., cells) in the space is determined by the Maximum Likelihood Estimator(MLE) and the center of each cluster is determined by the Learning Vector Quantization(MLQ) method, an unsupervised learning algorithm. The distances between every cell pair and the center of each cluster are calculated. According to the Euclidean distance from the center, the data are sorted into each duster. Removing the outliers of each cluster based on the distribution of the distance completes the sorting process. A computer program, 'WAVESORTER' was written using the Matlab(The Mathworks Inc.) to realize all the phases of the sorting processes. In this paper, the logic and routines of the 'WAVESORTER' is described.