Trial Lecture – time and place
See Trial Lecture.
Adjudication committee
- First opponent: Professor Helen R. Murphy, Norwich Medical School, University of East Anglia, UK
- Second opponent: Professor Henning Beck-Nielsen, Institute of Clinical Research, University of Southern Denmark, Denmark
- Third member and chair of the evaluation committee: Associate Professor Aud Høieggen, Faculty of Medicine, University of Oslo
Chair of the Defence
Professor Emeritus Kristian Folkvord Hanssen, Faculty of Medicine, University of Oslo
Principal Supervisor
Professor Ørjan Grøttem Martinsen, Faculty of Mathematics and Natural Sciences, University of Oslo
Summary
Most diabetes patients have good control of their blood glucose. However, episodes of blood glucose below normal levels, hypoglycaemia, is still a considerable problem for some patients, and can potentially be fatal.
The aim of this thesis was to investigate if measurements with non-invasive sensors attached to the skin could detect hypoglycaemia. 20 participants with type 1 diabetes underwent two sessions in random order, one session with normal blood glucose and one session with induced hypoglycaemia. Sensors for sweating, ECG, temperature, bioimpedance and near-infrared spectroscopy were attached. Participants were specially selected based on an impaired ability to sense when they have hypoglycaemia.
It is known that hypoglycaemia induces a stress response that among other things increases sweating and induces ECG changes. We hypothesized that we could measure this (and other changes) even when patients were unable to recognize it, or at least before they could recognize it themselves.
For the sensors aimed at detecting stress responses we could see changes on a group level, but results were more ambiguous on an individual level. Five of the participants had quite conspicuous changes, but they also reported that they had recognized symptoms of hypoglycaemia. The 15 participants with no symptoms, who would have most use of a potentially alarm system, had less obvious changes.
When we introduced measurements from bioimpedance and near infrared spectroscopy and employed more advanced data processing, including multivariate regression and a probabilistic detection model, we were able to identify hypoglycaemia in up to 19/20 participants with acceptable accuracy, depending on the sensors combined. As such, an alarm system that would detect most hypoglycaemic episodes seems feasible.
The main limitation of the study was that participants were relaxed in bed during procedures. Further testing in real-life situations with wearable sensors is needed to verify the findings.
Additional information
Contact the research support staff.