Title: Electroencephalogram Signals Based on Machine Learning and Cross-Spectrum Features
Authors: José Fiel, Thaynara Ribeiro, Eline Melo, Raphael Navegantes, Francinaldo Gomes, Antonio Pereira
Abstract: We used machine learning tools to discriminate resting-state brain electrical activity measured with electroencephalography (EEG) of patients with refractory epilepsy (RE) from healthy controls (HC). We propose a cross-spectral density-based measure as a signal feature to distinguish between healthy and epileptic subjects using machine-learning algorithms linear discriminant analysis (LDA) and support vector machines (SVM). The resting-state EEG of epileptic patients were obtained from interictal periods without any epileptiform activity. We recorded from 11 epilepsy patients and 7 healthy age-matched controls. Both algorithms obtained 100 % accuracy. Our results show that a distinction between the two groups is possible with high accuracy when a 190-dimensional feature vector is used as input.
Key-words: Cross-spectrum density, Debiased weighted phase-lag index, Electroencephalography, Epilepsy, Machine learning
DOI code: 10.21528/CBIC2019-11
Paper in PDF: cCBIC2019-11.pdf
BibTeX file: CBIC2019-11.bib