Project Proposals for William Flynn Scholarship: IIndex page

Project Number: 9
Project Title: Detection of Inter-ictal Spikes in EEG Data Using Hybrid Neural Networks
Project Supervisor: Dr. Girijesh Prasad, Professor Martin Mc Ginnity

Aims and Objectives
Between seizures, the electroencephalogram (EEG) of a subject with epilepsy is usually characterised by occasional epileptiform transients or spikes and sharp waves complexes (inter-ictal activity). These are notoriously difficult to detect reliably. The detection of epileptiform transients by visual inspection of the EEG recording is a highly skilled and time consuming process. It is therefore aimed to develop a neurocomputing-based spike classifier to assist the electroencephalographer in the detection of inter-ictal spikes in the EEG data of epileptic subjects. In order to ensure higher reliability and accuracy, this project proposes to develop a classifier using a hybrid neural network architecture.

The Development of the Proposed Classifier
It is proposed to develop a novel neural network classifier system that classifies EEG patterns as either being normal or containing a spike-like waveform. In recent years, application of neural networks in the automated detection of spikes has been investigated by several researchers with limited success. In order to attain significantly higher classification accuracy, the proposed system would incorporate the following special features.

1. The neural classifier would make use of both the time-domain and the frequency domain parameters for neural network training.
2. The training data set would include EEG data recorded during both sleep and wakefulness conditions.
3. A novel self-organising fuzzy neural network (SOFNN) architecture would be used for network design. This hybrid architecture would facilitate the incorporation of fuzzy reasoning capability (in a similar approach to a human EEG expert) in addition to the learning capability of the neural networks.

The development of neural classifier would require EEG data recorded under specified experimental conditions. A formal arrangement is in place to obtain EEG data from a local hospital.

Benefits
Ambulatory monitoring of patients with known or suspected epilepsy has become widely used and this may involve one or more days of continuous EEG recording. Due to wide variety of morphologies of epileptiform transients and their similarities to waves which are part of the background activity and to artefacts (e.g. extracerebral potentials from muscles, eyes, heart, electrodes etc.) the detection of inter-itical activity is far from straightforward. Inter-ictal spikes are always a relatively rare occurrence and are occasionally missed by the clinicians who review the paper records retrospectively. The proposed classifier will help diagnose epilepsy more accurately and at a faster speed. In addition, the automation of the detection process will increase objectivity and uniformity and enable further research studies.

If you are interested in being considered for a studentship please contact
the Group Director, Professor T.M. McGinnity by email:
tm.mcginnity@ulst.ac.uk

or telephone: +44-(0)28-71375417.

See the current research section of this website for details on research projects pursued by existing PhD students