Predicting movement and laterality from Deep Brain Local Field Potentials

The use of Deep Brain Local Field Potentials (LFP) in the process of connecting the human brain with artificial devices is one of the most promising fields in neural engineering. Inner mechanisms of our the central nervous system (CNS) can be understood through the study of LFPs. Of special importance are the the LFPs that come from subthalamic nucleus (STN) as they are related to the preparation, execution and imaging of movements. While researchers have focused on decoding movements and its laterality, left or right sided visually cued movements from STN LFPs, there is scope for using the same information for prediction of movements and laterality. In this paper, an algorithm is proposed that can be used to predict movement and laterality using STN LFPs. For this, wavelet packet transform (WPT) is used to generate separated frequency components of the LFPs. Then a selection of time and frequency domain features are used, namely time window based power features, causality features computed using granger causality and cross correlation, and frequency domain features computed using discrete cosine transform (DCT). Utilizing a weighted sequential feature selection process (WSFS), promising results are obtained from a Bayesian classifier along with cross validation procedure.