Sleep arousals are defined as temporary increases in the EEG frequency.  As the number of arousals increase, it interrupts the continuity of sleep and leads to sleep fragmentation.

In our group, we investigate real-time sleep arousal detection methods. Systems such as wearables for brain stimulation require timely detection of sleep arousals. Available methods do not address the real-time processing of the data, hence induce delays for real time systems. In addition, as arousals occur rarely in sleep (~80 arousal per night for a healthy subject with the age 18-30, typical arousal length 3-15s), the dataset is quite imbalanced for the deep learning algorithms.

In this project, you will work with EEG data and develop a deep learning algorithm for real-time systems. You will analyze the delay induced by the algorithm. With this algorithm, we aim to minimize or eliminate the delay of the real-time arousal detections.

For further information please contact Tugce Canbaz by e-mail.