Back

A Large Harmonized Upper and Lower Limb Accelerometry Dataset: A Resource for Rehabilitation Scientists

Miller, A.; Lohse, K.; Bland, M. D.; Konrad, J. D.; Hoyt, C. R.; Lenze, E. J.; Lang, C. E.

2024-08-16 rehabilitation medicine and physical therapy
10.1101/2024.08.15.24312066 medRxiv
Show abstract

Wearable sensors can measure movement in daily life, an outcome that is salient to patients, and have been critical to accelerating progress in rehabilitation research and practice. However, collecting and processing sensor data is burdensome, leaving many scientists with limited access to such data. To address these challenges, we present a harmonized, wearable sensor dataset that combines 2,885 recording days of sensor data from the upper and lower limbs from eight studies. The dataset includes 790 individuals ages 0 - 90, nearly equal sex proportions (53% male, 47% female), and representation from a range of demographic backgrounds (69.4% White, 24.9% Black, 1.8% Asian) and clinical conditions (46% neurotypical, 31% stroke, 7% Parkinsons disease, 6% orthopedic conditions, and others). The dataset is publicly available and accompanied by open source code and an app that allows for interaction with the data. This dataset will facilitate the use of sensor data to advance rehabilitation research and practice, improve the reproducibility and replicability of wearable sensor studies, and minimize costs and duplicated scientific efforts.

Published in Data in Brief · training set

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.