Abstract. Background and Aim: Monitoring physiological signals during sleep can have substantial impact on detecting temporary intrusion of wakefulness, referred to as sleep arousals, in order to improve the quality of sleep. To overcome the problems associated with the cumbersome visual inspection of these events by sleep experts, automated sleep arousal recognition algorithms have been proposed.
Method: As part of the Physionet/Computing in Cardiology Challenge 2018, this study proposes a deep ensemble neural network architecture for automatic arousal recognition from multi-modal sensor signals. Separate branches of the neural network extract features from electro-encephalography, electrooculography, electromyogram, breathing patterns and oxygen saturation level; and a final fully-connected neural network combines features computed from the signal sources to estimate the probability of arousal in each region of interest. We investigate the use of shared-parameter Siamese architectures for effective feature calibration. Namely, at each forward and backward pass through the network we concatenate to the input a user-specific template signal that is processed by an identical copy of the network.
Result: The proposed architecture obtains a preliminary AUPR score of 0:426 on the hidden test set of the official phase of Physionet/CinC challenge 2018. A similar score of 0:45 is obtained by means of 10-fold cross-validation on the training set provided.