This NCI-funded R01 grant will validate machine-learned algorithms to classify patterns of accelerometer data to better discriminate types of sedentary behavior (SB) and physical activity (PA). Using an annotated data set of SenseCam images in three free-living population subgroups (children, adults, and older adults), we will compare sensitivity, specificity and percent agreement between behavioral classifiers derived from: (a) hip vs. wrist mounted accelerometers (b) the extent to which adding GPS data improves discrimination accuracy over accelerometer only behavior classification and (c) the extent to which adding GIS data improves discrimination accuracy over accelerometer and GPS behavior classification alone. We will assess whether algorithms apply across all population groups or have to be tailored to different ages, genders and BMIs. As part of this project we are developing processes to automatically segment and annotate the SenseCam images.

Principal Investigator: Jacqueline Kerr, PhD
Co-Investigators: Simon Marshall, PhD, Loki Natarajan PhD, Jim Sallis, PhD, Gert Lanckriet PhD, Serge Belongie
Consultants: John Staudenmayer
Project Coordinator: Katie Crist, MPH