Multi-Phase Seizure Classification Using 2D Spectrograms and Convolutional Neural Networks

Andrew Thomas, Anif Zarus (2025) Multi-Phase Seizure Classification Using 2D Spectrograms and Convolutional Neural Networks. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Epilepsy affects millions of people around the world. It is marked by recurrent seizures that have distinct phases: pre-seizure, seizure, and post-seizure. Classifying these phases accurately using EEG signals can improve patient care. This study looks at how convolutional neural networks, specifically a custom CNN and pretrained models like VGG16 and ResNet50, can classify seizure phases from EEG spectrograms in the CHB-MIT dataset. The study also focuses on how preprocessing steps like channel selection, filtering, and epoching impact model performance. Results indicate that both the custom CNN and VGG16 reached about 90% accuracy. In contrast, ResNet50 performed poorly, especially in detecting post-seizure phases. The study also points out limitations related to dataset size and computational issues. Future research should look into using Continuous Wavelet Transform (CWT) for better feature extraction. It should also involve larger and more diverse datasets and explore better model architectures to improve the accuracy of seizure phase classification and its clinical use.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Subjects:	R Medicine > Healthcare Industry R Medicine > RC Internal medicine > RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry > Neurology. Diseases of the Nervous System.
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	24 Aug 2026 15:34
Last Modified:	24 Aug 2026 15:34
URI:	https://norma.ncirl.ie/id/eprint/9616

Actions (login required)

View Item