Available masters and doctoral projects

An overview of opportunities for students to work with scientists at Reykjavik University on research projects. Please contact the person listed within each project for further information.

Undergraduate level projects are on the Icelandic portion of RU's website: Available undergraduate level projects

  • Arousal Detection in EEG

Arousal detection in EEG

MSc level project, School of Science and Engineering

Abstract

Develop an algorithm to automatically score cortical arousal in EEG signals measured during sleep. According to the scoring rules set forward by the American Academy of Sleep Medicine, arousals are short events, lasting at least 3 seconds occurring on average 10-20 times per hour of sleep in healthy people, but may occur hundreds of times per hour in people suffering sleep disorders. EEG arousals are identified by abrupt shifts in EEG frequency. The method will be developed using a large data set of manually scored EEG data.

Description

Sleep disorders are diagnosed by analyzing data collected during a sleep study where physiological signals are measured during sleep. Sleep studies have a varying level of detail and complexity related to them, with the most detailed and complex studies collecting detailed signals relating to sleep quality, breathing quality, and limb movements. These studies, polysomnographs (PSG), require a technician to set up an electroencephalogram (EEG), an electrooculogram (EOG), and an electromyogram (EMG). These are 11 signals, time series, which are used to determine the patients sleep stage and sleep stability. A 2 minute example of the signals is shown in the figure. Each 30 second period, an epoch, of these signals, during the night, is analyzed to determine the patients sleep stage. A person can be in one of 3 non-REM sleep stages, REM sleep, or be awake. This is normally done automatically by a computer and a trained technician then looks at each epoch and makes sure it is correctly scored. 

Every so often a sleeping person may have a cortical arousal. A cortical arousal is marked by a shift in EEG frequencies and changes in the EMG signal. An arousal may be spontaneous or it may be due to a disturbance from the environment or caused by a sleep disorder such as obstructive sleep apnea or periodic limb movements. An arousal is a rare event and may only happen around 10 times per hour of sleep. Furthermore, arousals are short events lasting from 3 seconds to several seconds. This rarity and short duration of arousals pose a certain challenge to apply machine learning to detecting them accurately.

Goal

The goal of the project is to develop an automatic method of scoring arousals from conventional EEG, EOG, and EMG signals recorded during a PSG study.

Funded position

No

Other

This project is a collaboration between Reykjavik University's Language and Voice Lab and Nox Medical.

Contact

Jón Guðnason, jg@ru.is