Master of Science Marc Weitz will tuesday June 17th, 2025, at 12:15 hold his Thesis Defense for the PhD degree in Science. The title of the thesis is:
« Device-based measurement of lifestyle-related variables from the hip
Methods for 24 h wear protocols under free-living conditions »
Physical activity epidemiology is the study of patterns, risks, and effects of physical activity within populations. It plays a crucial role in understanding how physical activity influences health outcomes, such as reducing the risk of chronic diseases like cardiovascular disease, diabetes, and obesity. An important prerequisite is the accurate measurement of physical activity, which can be assessed in various ways. A popular method due to its cheap and unobtrusive application is accelerometry. In accelerometry, small sensors are attached to the body to measure raw acceleration from which various lifestyle-related variables such as physical activity, sedentary time and sleep are estimated. However, raw acceleration patterns are highly sensitive to where the sensor is placed on the body. In this thesis, we focus on acceleration data collected from the hip. Hip placement has previously only been used during daytime with the accelerometer being removed over night. While methods for detecting sleep from other device placements have been developed, such methods are missing for hip accelerometry. Here, we introduce the first algorithm to estimate time in bed specifically from raw hip acceleration to extend the methodology to 24 h recording. The algorithm achieves an accuracy of 94 % on the training data and 92 % on unseen data from the same population. Together with other recently developed and hip-specific algorithms we provide this algorithm in a comprehensive and publicly available software package. Our subsequent research focuses on identifying and investigating potential limitations of hip-based accelerometry in general, but also with emphasis on the 24 h methodology. We show that posthoc autocalibration leads to up to 14 % less observed moderate-to-vigorous physical activity. Autocalibration utilizes quasi-static data, e.g., from during night time, to estimate and correct the device’s calibration. Finally, we explore the impact of algorithm selection on the estimation of sedentary time. To estimate sedentary time in 24 h recordings, non-wear time and sleep have to be identified and removed. We show that estimated sedentary time varies between 3 h and 13.5 h on the same dataset between different algorithm combinations, thus severely affecting the interpretation of sedentary behavior especially between studies. Collectively, these contributions enhance and extend the hip-based accelerometry methodology for 24 h recordings. The proposed time in bed algorithm enables the analysis of sedentary time and time in bed itself on a population level. This is of particular interest for studies currently using a 24 h hip-based protocol like the Tromsø Study or the German National Cohort Study. Additionally, this work facilitates retrospective analysis in studies which have initially collected data from hip, but are now using, for example, wrist accelerometry. However, it remains a challenge to overcome the huge variations observed in estimates resulting from different methods and method combinations.
1st Opponent: Associate Profesor Alex Rowlands, Department of Population Health Sciences, The University of Leicester, University Road, Leicester, United Kingdom
2nd Opponent: Profesor Elin Ekblom Bak, Department of Physical Activity and Health, The Swedish School of Sport and Health Sciences, GIH, Lidingövägen 1, Stockholm
Internal member and leader of the committee: Associate Professor Helge Fredriksen, IFI, campus Bodø, NT-fak, UIT
The defence and trial lecture will be streamed from these following links at Panopto:
Defence (12:15 - 16:00)
Trial Lecture (10:15 - 11:15)
The thesis is available at Munin Here.