PhD Scholarship in Design of Interpretable End-to-end Deep Learning Models for Diagnosis of Sleep Disorders and Sleep Quality Evaluation

The Department of Electrical Engineering invites applicants for a 3-year PhD in sleep analytics designing biomedical signal processing and machine learning for nocturnal polysomnography (PSG) signals and other data. Half of the 3 years will be spent at DTU and Rigshospitalet, and half at Stanford University in California, USA. The PhD is part of a longstanding collaboration between these Universities for this topic. Professors with experience in technical science and sleep medicine will supervise closely the student in both locations.

In this position, you can expect to perform data analysis on tens of thousands of laboratory recorded sleep studies, in combination with genetic, proteomic and medical/questionnaire data. A sleep study, or polysomnography (PSG), is comprised of multiple physiological signals (electroencephalography, electrocardiography, electrooculography, chin and leg electromyography, respiratory, oxygen saturation, etc.) recorded simultaneously throughout the night to provide physiological measures of human activity and behavior during sleep.

You will be expected to transform and analyze these digital signals by innovating biomedical signal processing methods as well as innovative machine learning and deep learning techniques (e.g. convolutional and recurrent neural networks), together with other signals and data.

The purpose of the PhD project is to discover fully data-driven methods to estimate sleep quality based on biomedical analysis and automatic modelling of large cohorts of PSG and genetic data. In the process, new data-driven patterns of physiological activity during sleep will be discovered and compared to known patterns (micro-arousals, sleep stages, sleep apnea, periodic leg movements, etc.) defined by medical standards. The data processing pipeline aims at being able to evaluate sleep quality by estimating known measures of sleep quality and daytime sleepiness as well as sleep disorders such as insomnia and restless leg syndrome. The PhD student will work in a highly collaborative environment and develop novel algorithms for processing PSG and genetic data based on advanced interpretable machine learning techniques to describe and evaluate sleep quality. Collaborations across other labs and across departments are encouraged. Experience in advanced biomedical signal processing, programming, advanced mathematics, and a solid background in statistical analysis are needed. Excellent interpersonal skills and the ability to interact effectively with members of the research teams are essential to the success of the individual in this position. The successful candidate must be able to learn and work independently yet collaborate effectively with co-workers.

Candidates should have a two-year master’s degree (120 ECTS points) or a similar degree with an academic level equivalent to a two-year master’s degree.