Real-Time Sleep Microstructure Detection on a Wearable Device

Ulm University

Processes that occur during sleep have been shown to interact with numerous physical and mental health issues such as Parkinson's, Hypertension, ADHD and Depression. In all of these areas, we observe some form of deviation from the sleep patterns that we observe in a healthy individual. Our recent studies suggest that by using a mobile device to perform targeted interventions during sleep we can modulate the observed sleep patterns to adjust them closer towards what would be considered healthy. One of the biggest challenges in this task is to apply an appropriate intervention at the optimal timing to achieve the desired goal. A central component in achieving good interventions are the detection of various characteristics of sleep microstructure such as sleep stages, K-complexes and spindles that can only be seen in sleep EEG. Recent supervised machine learning techniques have shown the ability to perform such detection at close to expert level but mostly in an offline setting. To perform sleep modulation we need the detection available in real-time and on the embedded hardware.

This is where your project comes in. Based on your literature review you will implement and evaluate  existing approaches and innovate a solution that is suitable for the purpose. You will have the opportunity to implement the algorithms on existing wearable hardware  and test your approach.

If you are excited about using machine learning and mobile electronics to tackle a real-world healthcare problem and contribute to an active research topic, we would be delighted to hear from you. While some experience with machine learning, programming in python or a similar language, and interest in electronics would help you to excel in the project, we are also happy to discuss the project details with you and tailor it to what you are excited about.

For further information please contact Luzius Brogli by e-mail.