Detection of sleep apnoea in infants using ECG and oximetry signals

We present a study into the usage of combined nighttime electrocardiogram (ECG) and pulse oximetry recordings to automatically detect sleep apnoea in infants. The study draws upon the polysomnogram recordings found inside the National Collaborative Home Infant Monitoring Evaluation (CHIME) database. Viable ECG data, pulse oximetry data and scored respiratory information was extracted for 288 subjects from this dataset and time-aligned to 30s epochs. Features were extracted from both the ECG and the pulse-oximetry data and were then used alongside the scored respiratory information to train a classification model based on linear discriminants. Performance of the classifier was evaluated using a leave-one-out cross-validation scheme and an accuracy of 82.6% was achieved, with a specificity of 82. 6% and a sensitivity of 58.0%.