Unsupervised spike sorting with ICA and its evaluation using GENESIS simulations

Data acquisition for multisite neuron recordings still requires two main problems to be solved-the reliable detection of spikes and the sorting of these spikes by their originating neurons. Approaches and solutions for both problems are difficult to evaluate quantitatively, due to a lack of knowledge about the “truth” behind the experimental data. Biologically realistic simulations allow us to overcome this fundamental problem and to control all the processes which lead to the measured data. Within this framework, the quantitative evaluation of the performance of data analysis methods becomes possible. In this paper, the potential of independent component analysis (ICA) for spike sorting and detection is studied. A biologically realistic simulation of hippocampal CA3 is used to obtain a measure of the quality and usability of ICA for solving the neural cocktail party problem. The results are promising.