Early Prediction of Perinatal Brain Injury in the EEG using Wavelets and Artificial Neural Networks Event as iCalendar

(Engineering Science)

03 February 2015

11am - 12pm

Venue: Room 439.201

Location: Level 2, Uniservices House, 70 Symonds St

Host: Dr Richard Clarke

Contact email: rj.clarke@auckland.ac.nz


A Department of Engineering Science research seminar by Hamid Abassi, PhD Candidate, Department of Engineering Science.
 

hamid-abbasi-2014-200x250

Abstract:

Oxygen deprivation either before or during labor is a major cause of death among newborn infants. Therefore, understanding the cause of brain injury due to the lack of oxygen, Hypoxic Ischemia (HI) or Hypoxia, is important.

It has been shown in the fetal sheep model of hypoxic-ischemia that the electroencephalogram (EEG) exhibits a 6-8 hours post insult period, known as the ‘Latent phase’ or “window of opportunity”, after which epileptiform activity of high amplitude appears. Particular micro-scale high frequency transients in the forms of spikes, sharps and complex micro-seizures occur in this period that can be predictive of outcome. Thus, an automated recognition scheme of such embedded transients in the latent phase may prove beneficial in the identification of hypoxia-ischemia and helps for early prediction of the brain injury in newborn babies.

In this talk I will show that how advanced signal processing methods, such as Wavelets and Type-2-Fuzzy Logic systems (Type-2 FLS), can be beneficial for the identification of high frequency micro-scale hypoxic ischemic transient activities along EEG signals after a HI insult in a fetal sheep model. In this talk, I will describe the significant advantage that can be obtained in the detection of sharp-waves using a combination of wavelets and Type-2-FLS method on recorded hypoxic-ischemic EEG signals at high sampling rates 1024 Hz signal as opposed to the conventional 64Hz sampling rates used in clinical study.

I will also demonstrate how the suggested methods identify spike and sharp activities over noise and other similar epileptiform activities with the same frequency characteristics, accurately. This has significantly increased the chance of a reliable detection of the evolutionary brain injury during its primitive levels which helps clinicians to treat an at risk baby and suggest that there should be a movement toward recording high frequency EEG for analysis of hypoxic ischemic micro-scale transients that does not occur at present.