Various techniques have been considered in the past to identify distinct spike shapes from mulitunit extracellular recording. These techniques involve adaptive filtering techniques or template matching techniques or hierarchical clustering techniques. In this investigation, we have used Principal Component Analysis followed by various clustering techniques to identify distinct spike shapes. The amplitude filter is used to separate spikes from background neuronal activity. The correlation matrix of the spike data is used to compute principal component wave forms. Each spike is thus represented by the coefficients of principal components. Then, We have used agglomorative hierarchical clustering algorithm to perform the initial clustering of the data set. The clustering results are then refined by the application of the Estimation Maximization Algorithm. The Bayesian Information Criteria(BIC) is used to find out best fit of the model to the data set.
If you have any questions please contact the ETD Team, libetd@njit.edu.