Decoding Movements from Human Deep Brain Local Field Potentials Using Radial Basis Function Neural Network

Abstract

Research on neural process is fundamental to understand neurodegenerative disorders and develop its interventions. This also enhances the development of brain machine interfaces to assist neurologically impaired human and rehabilitation. This study aimed to decode deep brain local field potentials (LFPs) related to voluntary movement activities and its forthcoming laterality, left or right sided visually cued movements. The frequency related components of local field potentials from the sub thalamic nucleus (STN) were decomposed by time scale domain using wavelet packet transform (WPT). In each frequency component, event related instantaneous power was considered as features for decoding. Decoding of movement (Event vs. Rest) and its sequential laterality (Left vs. Right) were performed using radial basis function neural network (RBFNN). The average classification accuracy achieved 85.93% for distinguishing movement from the rest, while laterality discrimination, the accuracy achieved 70.81% with 10 fold cross validation. The RBFNN classifier successfully managed to achieve decoding accuracy better than the chance level during movement and its laterality for all subjects.