Distance Metric Approach for Nearest Neighbour Recall of Neonatal EEG

Abstract

Clinical neurophysiologists often find it difficult to recall rare EEG patterns despite the fact that this information could be diagnostic and help with treatment intervention. Having the neurophysiologist physically searching through previous neonatal EEG recordings is a cumbersome and time consuming task. This paper examines the performance of a brute force distance metric approach to locate similar neonatal EEG patterns. This preliminary work is to set a baseline for neonatal EEG nearest neighbour pattern recall. A fixed point distance metric and an elastic distance metric are evaluated in this paper on the time series data and on the features extracted from data. The system was tested on six different neonatal EEG pattern types with 430 events in total and the results are presented.