lilykoff
commented
1 year ago Possible data sources: Wrist and hip worn sensors AD data from Vadim
Ideas Correlation between PA and steps in NHANES like this paper https://www.mdpi.com/1424-8220/23/20/8542#B15-sensors-23-08542 - here they calculate correlation between wrist-based measures of PA and lumbar-based step counts
Algorithms https://scikit-digital-health.readthedocs.io/en/latest/ - only for lumbar https://cran.r-project.org/web/packages/GENEAclassify/index.html - requires manual labeling https://pubmed.ncbi.nlm.nih.gov/36146441/ - claims very high accuracy in gait detection (not step counts) but not open source
i) accelerometry-assessment of gait has been focusing on detecting/describing bouts of sustained walking (typically, > 60sec), ii) plot something like accuracy vs bout walking length, by each algorithm iii) understand what are characteristics of walking/subjects when most of the algorithms agree and when they disagree. iv) it would be nice to wear fitbit, apple watch, and actigraph and then compare all available algorithms vs the step count from those two most common consumer grade devices. v) Tudor-Locke, Catrine, Sarah M. Camhi, Claudia Leonardi, William D. Johnson, Peter T. Katzmarzyk, Conrad P. Earnest, and Timothy S. Church. "Patterns of adult stepping cadence in the 2005–2006 NHANES." Preventive medicine 53, no. 3 (2011): 178-181. was using some sort of peak counting algorithm built in the older actigraph, or I might be wrong. vi) wear fitbit, apple watch, and actigraph and then compare all available algorithms vs the step count from those two most common consumer grade devices. Literature: Objective assessment of fall risk in Parkinson's disease using a body-fixed sensor worn for 3 days Associations between daily-living physical activity and laboratory-based assessments of motor severity in patients with falls and Parkinson's disease