Spike source identification using artificial intelligence techniques

We present a methodology for the automatic detection of target regions in the brain for ablation, stimulation and restorative surgery for Parkinson's disease and other neurological disorders. The methodology includes wavelets for the correct characterization of the non-stationarity of the spike train and hidden Markov models as a suitable tool for describing dynamic behavior of the signal across time. Similarity measure and Kullback-Leibler distance were used for discriminant evaluation of HMM. We also compare HMM with other artificial intelligence techniques for the classification task. Results show classification performance up to 97% with the proposed methodology.