Feature extraction and classification of electrocardiogram (ECG) signals related to hypoglycaemia

Abstract

Nocturnal hypoglycaemia has been implicated in the sudden deaths of young people with diabetes. Experimental hypoglycaemia has been found to prolong the ventricular repolarisation and to affect the T wave morphology. It is postulated that abnormally low blood glucose could in certain circumstances, be responsible for the development of a fatal cardiac arrhythmia. We have used automatic extraction of both time-interval and morphological features, from the electrocardiogram (ECG) to classify ECGs into normal and arrhythmic. Classification was implemented by artificial neural networks (ANN) and linear discriminant analysis (LDA). The ANN gave more accurate results. Average training accuracy of the ANN was 85.07% compared with 70.15% on unseen data. This study may lead towards the demonstration of the possible relationship between cardiac function and abnormally low blood glucose.