Assessing Physical Activities Occurring on Sidewalks and Streets: Protocol for a Cross-Sectional Study.

Area of Science:

• Public Health
• Kinesiology
• Computer Science

Background:

• Outdoor physical activity (PA) measurement on sidewalks and streets is crucial.
• The traditional Block Walk Method (BWM) is labor-intensive and unchanged since 2006.
• Advancements in PA measurement are needed to better understand behavior.

Purpose of the Study:

• Develop and test a novel BWM using a wearable video device (WVD).
• Integrate machine learning and computer vision for automated PA data extraction.
• Improve accuracy and reduce data collection burden for street-based PA.

Main Methods:

• Trained observers used a WVD and traditional BWM along predetermined routes.
• Video data captured by WVD was analyzed by investigators and deep convolutional neural networks (CNNs).
• Bland Altman methods and intraclass correlation coefficients (ICCs) assessed agreement; moderator analyses evaluated error sources.

Main Results:

• Preliminary studies show BWM is reliable for PA mode (Cramer V=.89), location (Cohen kappa=.85), and participant count (ICC=.85).
• PA counts correlated with sidewalk quality (r=.39) and neighborhood aesthetics (r=.49).
• CNNs successfully detected pedestrians, vehicles, and park facility users.

Conclusions:

• The WVD-enhanced BWM is expected to improve measurement accuracy and efficiency.
• Future expansions will include caloric expenditure and environmental condition analysis.
• This technology offers a significant advancement in understanding PA behavior in urban environments.