Dissertation/Thesis Abstract

Nonlinear epilepsy forewarning by support vector machines
by Ashbee, William S., M.S., University of South Alabama, 2013, 51; 1604736
Abstract (Summary)

Epilepsy affects approximately 1% of the population and can be as benign as causing social awkwardness or as malign as causing death. Using algorithms to predict seizures can lead to event and danger mitigation that could treat a subset of epileptics. The prediction algorithm uses Takens' theorem to generate phase space graphs that can be used to discover EEG anomalies, which would lead to the pre-ictal state. The phase space graphs that patients' EEG generate are compared to stored phase space graphs in order to observe a phase transition. Support Vector Machines are able to achieve high accuracy as the machine learning method chosen to learn the relevant patterns and make the predictions based on observed anomalies.

Indexing (document details)
Advisor: McDonald, Jeffrey T.
Commitee: Hively, Lee M., Johnsten, Tom D., McDonald, Jeffrey T.
School: University of South Alabama
Department: Computer Science
School Location: United States -- Alabama
Source: MAI 55/02M(E), Masters Abstracts International
Subjects: Neurosciences, Artificial intelligence, Computer science
Keywords: Anomaly, Chaos, Epilepsy, Phase-transition, Prediction, Svm
Publication Number: 1604736
ISBN: 978-1-339-28764-5
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy