Prediction of canine epilepsy

Seizure prediction is a problem in biomedical science which now is possible to solve with machine learning methods. A seizure prediction system has the power to assist those affected by epilepsy in better managing their medication, daily activities and improving the quality of life. Usage of machine learning algorithms and the availability of long term Intracranial Electroencephalographic (iEEG) recordings have tremendously reduced the complications involved in the challenging seizure prediction problem. Data, in the form of iEEG was collected from canines with naturally occurring epilepsy for the analysis and a seizure prediction system consisting of a machine learning based pipeline was implemented to generate seizure warnings when potential preictal activity is observed in the iEEG recording. A comparison between the different extracted features, dimensionality reduction techniques, and machine learning techniques was performed to investigate the relative effectiveness of the different techniques in the application of seizure prediction. The machine learning protocol performed significantly better than a chance prediction algorithm in all the analyzed subjects. Moreover, the analysis revealed subject-specific neurophysiological changes in the extracted features prior to lead seizures suggesting the existence of a distinct, identifiable preictal state.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2016

Author dc:creator

• Varatharajah, Yogatheesan

Contributors dc:contributor

• Iyer, Ravishankar K.
• Kalbarczyk, Zbigniew T.

• Seizure prediction
• Canine epilepsy
• Preictal state
• Prediction pipeline
• Machine learning
• Dimensionality Reduction

Statement dc:rights

• Copyright 2015 Yogatheesan Varatharajah

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/89052

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/89052