2025/01/01 by Fabio Franzese, Pasan Hettiarachchi, Andreas Holtermann +2 · 1 voice · 2 citations
Health Professions · Medicine · #Balance, Gait, and Falls Prevention #Cardiovascular and exercise physiology #Physical Activity and Health
paper · doi:10.1123/jmpb.2024-0043
openalex publication_date 2025/01/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/02
Background : Compared with other placements, thigh-worn accelerometers offer the advantage to accurately capture postures. However, the lack of standardized sensor placement on the thigh raises concerns about the comparability and harmonization of data across studies. This study aimed to investigate the impact of sensor placement on thigh-worn accelerometer measurements. Methods : Thirty-six participants wore two sensors simultaneously at center- and upper-thigh position on the same thigh during daily activities for several days. Metrics of vector magnitude (Euclidian Norm Minus One), postures, and activities generated by GGIR and ActiPASS were analyzed with Bland–Altman plots and linear regressions. Results : The findings revealed a significant dependence of vector magnitude on sensor placement: 1 cm higher distance between the center- and upper-thigh position was correlated with almost 1 mg higher difference in Euclidian Norm Minus One between the two positions ( b = 0.94; 95% confidence interval [0.26, 1.62]). For time spent sedentary ( b = 0.03; 95% confidence interval [−2.48, 2.55]) and intense physical activities ( b = 0.22; 95% confidence interval [−0.10, 0.54]), the sensor position had no significant effect on the results. High concordance was also observed for other activities, such as walking, running, and sleeping. Conclusions : The analyses suggest that, despite variations in sensor placement, thigh-worn accelerometry data can be compared and harmonized for most postures and activities when analyzed by ActiPASS. The robust metrics by ActiPASS to accelerometer placements is likely because it utilizes multiple features in the acceleration signal (e.g., inclination, rotation) to classify activities and postures, rather than relying on a single feature (vector magnitude).