Body placement of inertial measurement units differentially affects physical activity assessment accuracy in drug-naïve Parkinson’s disease

Abstract

Abstract Parkinson’s disease (PD) affects physical activity, and physical activity reduces the burden of PD. Although shown in studies using inertial measurement units (IMU), it remains unclear at which position physical activity change can best be detected in this population. Within the FAIRPARK-II trial, a subgroup of 25 newly diagnosed persons with PD (pwPD) not taking disease-specific medication yet documented their physical activity and, in parallel, wore IMUs on the most affected ankle, wrist and the lower back for two weeks. Participant-reported physical activity was transformed into Metabolic Equivalents of Tasks (METs) in 15-minute intervals using The Compendium of Physical Activities; Euclidean Norm Minus One (ENMO) values were calculated and averaged over the same intervals for the IMU data. Data of at least 3 days with at least four simultaneous 15-minute epochs of both valid IMU and diary data within the time window (9.00 to 18.00) per participant was included, resulting in a total of 8,494 15-minute epochs used for this analysis. Root mean square error (RMSE) values were calculated between scaled normalized IMU-derived ENMO and normalized MET values for each of the nine IMU-MET combinations (three IMU positions × three MET intensity levels). The wrist and lower back IMU showed comparable RMSE values across all MET intensity levels, with both IMU positions showing lower RMSE values than the ankle position. Tremor affected RMSE negatively, whereby the lower back position may be slightly favorable for the assessment of physical activity in those with tremor. This prospective longitudinal dataset from a very rare cohort provides novel insights into the assessment of physical activity during the earliest clinically evident phase of Parkinson’s disease without disease-specific medication, which may inform future clinical trials and observational studies.