Automated monitoring of inflammation intensity (Completed Project)

By analysing large amounts of data from the vagus nerve during experimental inflammation, the project aimed to identify signal patterns that represent inflammation intensity. The goal was to create a lexicon of signal patterns to extract detailed information about the inflammation. The researchers used data available in the literature and their own data to develop machine learning methods, based on auto-encoders and clustering, for identifying nerve signals that can be linked to different cytokines. Pro-inflammatory cytokines are produced by the immune system and secreted in response to injury, stress or inflammation, among other things. The results show that the new techniques are effective for identifying relevant signal types, especially for TNF, but also that there is considerable variation between different recordings and that it is very important to be able to conduct controlled experiments with improved signal quality in the future.

The programme was co-directed by Peder Olofsson, Medical Doctor and Professor in Bioelectronic Bedicine at Karolinska Institutet, and Henrik Hult, Professor in Mathematical Statistics at KTH.

The project was conducted within the programme Bioelectronic Medicine.