Application area Sleep Research and Sleep Medicine
Vision
Sleep medicine is a very diverse medical discipline with extensive and complex examinations. Large amounts of data are collected and manual analysis only scratches the surface of the potential information content. Data integration and data science methods are needed to unlock the full potential.
Data integration makes it possible to assess large amounts of data by harmonizing and digitizing them and to carry out further analyses. Data science can provide support in two ways: (1) by assisting experts in the evaluation process through automation, and (2) by discovering new correlations and information in the data through exploration of new parameters and exploratory analysis. These and other aspects are part of our work in this application area.
Polysomnographic records contain various signals such as electroencephalogram (EEG), electrooculogram (EOG) and electromyogram (EMG). For medical assessment, states and events such as sleep, arousal and apnoea are manually annotated in the signals. We are investigating how this manual annotation can be automated using deep learning models. The image shows the biosignals in the upper channels and the manual annotation and automated classification in the lower channels.
Focus Areas
- Classification of sleep stages from cardiorespiratory signals
- Arousal detection
- Combination of classifiers and detectors for a complete, automated sleep report from the polysomnogram
- Investigation of new parameters for analyzing and evaluating sleep (Delta Power, Hypoxic Burden, ...)
Contact Person
© MF/Stephan Wiegand
Research fellow
NameMs Dr.-Ing. Miriam Goldammer
Send encrypted email via the SecureMail portal (for TUD external users only).
Selected Projects
-
Somnolink — Development of classification and prediction models in the context of sleep apnoea
- Sleep Harmonizer — Harmonization of polysomnograms and their annotations, incl. generation of new, harmonized data sets in EDF+ format