A Probabilistic Neural Network Approach for Prediction of Movement and Its Laterality from Deep Brain Local Field Potential

Abstract

Prediction of neural activity relating to movement is essential to understanding and treatment of neurodegenerative diseases and cybernetic interfaces. Here we had shown that it is possible to decode deep brain local field potentials (LFPs) related to movements and its laterality, left or right sided visually cued movements using Probabilistic Neural Network (PNN) classifier. The frequency related components of LFPs were extracted using the wavelet packet transform (WPT). Then the signal features were computed as the instantaneous power of each band using the Hilbert Transform (HT) with defined windows for motor response. Based on the extracted feature, PNN classifier was designed and evaluated using 10-fold cross validation method to identify the robustness for predicting movements. The Classification accuracy 82.72 ± 7.2 % achieved for distinguishing movement condition from the rest. While for subsequent discrimination of left and right movement, the accuracy reached up to 74.96 ± 10.5 %. Considering the classification performance (accuracy, sensitivity, specificity and the area under the Receiver Operating Characteristic (AUC) curve), PNN classifier successfully achieved better than chance level. The proposed modality and computational process may promisingly effective and powerful method to open up several possibilities for improving BMI applications, diagnosis of chronic neurological disorders and robust monitoring system with propitious result.