An adaptive strategy of classification for detecting hypoglycemia using only two EEG channels

Abstract

Hypoglycemia is the most common but highly feared side effect of the insulin therapy for patients with Type 1 Diabetes Mellitus (T1DM). Severe episodes of hypoglycemia can lead to unconsciousness, coma, and even death. The variety of hypoglycemic symptoms arises from the activation of the autonomous central nervous system and from reduced cerebral glucose consumption. In this study, electroencephalography (EEG) signals from five T1DM patients during an overnight clamp study were measured and analyzed. By applying a method of feature extraction using Fast Fourier Transform (FFT) and classification using neural networks, we establish that hypoglycemia can be detected non-invasively using EEG signals from only two channels. This paper demonstrates that a significant advantage can be achieved by implementing adaptive training. By adapting the classifier to a previously unseen person, the classification results can be improved from 60% sensitivity and 54% specificity to 75% sensitivity and 67% specificity.