At this year's national congress, The Danish Society for Biomedical Engineering awarded students, Aske Bluhme Klok and Joakim Edin, from DTU/KU, 1st prize for the presentation of their research: Automatic Sleep Stage Classification: a new weapon in fighting neurodegenerative diseases.
Aske explains how Parkinson's disease and a range of dementia conditions are often detected very late and how we need new methods to determine these conditions earlier.
The algorithm that he and Joakim have developed helps to detect a biomarker that can detect these diseases approximately 16 years before the classic symptoms occur.
This biomarker is defined by muscle activity during the dream sleep stage. Present methods to determine this biomarker are executed manually by trained experts. As this is expensive, time consuming and not very precise, they are working on introducing their algorithm for automatic staging of sleep stages, making it easier to assess if there is muscle activity during dream sleep.
The algorithm is based on supervised machine learning, designed to mimic the work of the sleep experts, and has an accuracy of 92.3% and a higher performance than state of the art.
Photo: Haslefoto.dk/Henrik Olsen