Charlee Katzka, “Comparison of Wearable Camera First Person Still Imagery and Indirect Calorimetry to Assess Physical Activity Intensity”
Mentor: Scott Strath, Kinesiology
Poster #109

Purpose: To examine if wearable camera (WC) still imagery can accurately and precisely measure physical activity intensity. Methods: 4 healthy participants were observed over four 1-hour visits. During these visits participants wore a Brinno TLC 130 WC and were outfitted with a COSMED K5 portable metabolic measurement system. Images from the WC were annotated with behavior codes derived from the Compendium of Physical Activities. Behavior codes were connected with Compendium MET values, which represent units of physical activity energy expenditure. Breath by breath oxygen consumption data in ml/kg/min from the COSMED K5 were averaged to the minute and then converted into METs by dividing ml/kg/min by 3.5. Image annotations and COSMED K5 data were then matched minute-by-minute, where MET values were grouped into the following physical activity intensities: sedentary (1.0-1.5 METs), light intensity (1.5-3.0 METs), and moderate to vigorous physical activity (MVPA: 3.0 METs+). For statistical analyses, behaviors that occurred for at least 5 minutes were examined. Overall percent agreement comparing COSMED K5 measured MET values to WC still-images MET estimates was computed. Confusion matrices were used to determine overall agreement by physical activity intensity categories. Results: 1260 images were included for analysis. The overall percentage agreement is between measured COSMED K5 METs and WC MET estimates was 70.9%. The agreement for sedentary, light and MVPA was 100%, 86%, and 49% respectively, where images classified 49% of MVPA as light intensity. Conclusion: WC images are accurate for assessing sedentary physical activity behaviors. Future work is warranted to examine what behaviors are being incorrectly classified as light intensity using a higher sample size.