Title: Classification of Pain Induced by Electrical Stimulation Using EEG-based features
Abstract: The overview of my work focuses on applying features extracted from pain-related EEG signals to classify pain perception levels and pain types related to the activation of nociceptive fibers. Even though studies reported using EEG for evaluating pain, neither high classification accuracy for multiple pain perception levels nor the differentiation between pain types related to the activation of nociceptive fibers has been achieved. A major reason lies in the lack of effective features to reflect the nonlinear and interaction between multiple processes of pain-related information processing. In this work, aiming to investigate the effective features extracted from pain-related EEG evoked by electrical stimulation for pain-classification, several feature extraction methods were proposed, including nonlinear analysis: Higuchi’s fractal dimension (HFD), Grassberger-Procaccia (GP) correlation dimension for evaluating pain perception levels, and brain connectivity analysis: Granger-causality (GC) analysis for classifying nociceptive fiber activations. Through the investigation, it was made clear that: 1) by application of combined nonlinear features based on Fisher’s criterion, classification accuracy for multiple pain perception levels is better than any existed features; 2) classification performances with GC features measures between EEG components on each frequency band are better than other those of GC features measures based on channel distribution. As a result, the classification systems based on the proposed features could realize high accuracy for pain perception levels and pain types. The findings could be used further to improve the performances of EEG-based classification systems for pain.
|Date||April 22, 2021 (Thu) 14:00 - 15:00|