Inhaltsbereich

This research area includes the development and application of advanced statistical methods in the area of biomedical information engineering. The focus of the biomedical work is on “Neural Computation” methods (Neural Networks, Gaussian Mixture Models, Hidden Markov Models, Support Vector Machines, ...) to analyze medical data, for example, to recognize pathological patterns. One main focus of our research is the intelligent analysis of bio-electrical signals like electroencephalography, electrooculography or electromyography. We have been involved in developing automatic intelligent systems for analysis of polysomnographic recordings in sleep medicine.

 

The research group offers state-of-the-art methods of pattern recognition to clinicians in a variety of areas. Some examples of successful applications are

• Early recognition of sepsis and other responses to infections from laboratory parameters (Ratzinger et al. 2015a, 2015b, 2014)
• A prediction model for mortality in cardiac arrest patients (Aschauer et al. 2014)
• The detection of patterns in time series of suicide frequencies (Vyssoki et al. 2014, Winkler et al. 2016 )

 

Past work included

• Detection of coronary artery disease in thallium scintigrams (Porenta et al. 1994)
• Control of rotary blood pumps (Stöckelmayer et al. 1995)
• Automated analysis of cardiotocograms (Ulbricht et al. 1998)
• Survival analysis for a study on implanted Corinthian stents (Dorffner et al. 2003)
• Estimation of lens position after cataract surgery (Findl et al. 2004)
• Removal of artefacts in EEG (Flexer et al. 2005)

 

Our researchers have been involved in the development and validation of Somnolyzer 24x7, a leading solution for automated sleep staging based on EEG, EOG and EMG (Punjabi et al. 2015, Anderer et al. 2010, Anderer et al. 2005). Another focus is the development of continuous models of sleep and wakefulness that promise to be of great diagnostic value in sleep medicine (Sykacek et al. 2002, Gruber el al. 2002, Flexer et al. 2005). The group has developed a novel type of probabilistic sleep model, revealing a fine-grained microstructure of sleep that better correlates to a sleeper’s subjective assessment of their own sleep quality (Lewandowski et al. 2012, Rosipal et al. 2013). Such models also identify sleep EEG as an individual "fingerprint” of a subject (Lewandowski et al. 2013). The age dependency of sleep parameters (Dorffner et al. 2015) and reliable sleep scoring in pharmaceutical trials (Hoever et al. 2012a, 2012b) have been additional research foci.