Time-scale analysis and classification of electroencephalographic signals (EEG): Application on remote--surveillance of epilepsy

Abstract

Electroencephalography (EEG) stands as a cornerstone in non-invasive brain activity monitoring, offering invaluable insights with high temporal resolution. This dissertation focuses on harnessing EEG for the detection of epileptic activity, specifically targeting the precise delineation of epileptic zones responsible for abnormal electrical patterns within the brain. The meticulous mapping of these zones is pivotal for assessing patients with pharmacoresistant epilepsy, paving the way for targeted seizure-free interventions. Thus, this dissertation introduces two complementary subsystems: the Representative EEG Channel Creator (RECC) and the Seizure Affected EEG Channel Detector (SAECD). The RECC contributes to enhanced dimensionality reduction with up to 93.75%, improving the efficiency of epilepsy pattern detection with a sensitivity of 98.46%. Upon a positive response from the RECC indicating epileptic EEG, the SAECD subsystem is activated. Leveraging the Energy-to-Shannon-Entropy ratio and a k-means clustering approach, SAECD precisely localizes the epileptogenic focus and traces the path of seizure diffusion within the brain with a promising average silhouette range of [51.21-88.18]%. In addition to the innovative subsystems introduced, this study places a particular emphasis on the selection of an appropriate mother wavelet. The careful choice of this latter plays a crucial role in enhancing the accuracy and efficiency of the proposed epilepsy detection system. In addition to advanced signal processing techniques, this research incorporates engineered features. These latter are strategically designed to capture nuanced aspects of epileptic activity, contributing to the overall robustness and effectiveness of the proposed methodology. Furthermore, the dissertation embraces the realm of telemedicine with the introduction of Epileptica, an asynchronous application designed for remote access to the results generated by the developed EEG-type-independent system. This telemedical platform enhances accessibility to critical information, supporting neurologists in making informed decisions regarding patient care and the suitability for seizure-free surgery.