Validation studies of the FLASH-TV system to passively measure children's TV viewing

Authors: KUMAR VADATHYA, ANIL - Rice University; GARZA, TATYANA - Children'S Nutrition Research Center (CNRC); ALAM, UZAIR - Children'S Nutrition Research Center (CNRC); HO, ALEX - Rice University; MUSAAD, SALMA - Children'S Nutrition Research Center (CNRC); BELTRAN, ALICIA - Children'S Nutrition Research Center (CNRC); MORENO, JENNETTE - Children'S Nutrition Research Center (CNRC); BARANOWSKI, TOM - Children'S Nutrition Research Center (CNRC); HAIDAR, NIMAH - Rice University; HUGHES, SHERYL - Children'S Nutrition Research Center (CNRC); MENDOZA, JASON - Fred Hutchinson Cancer Research Center; VEERARAGHAVAN, ASHOK - Rice University; YOUNG, JOSEPH - Rice University; SANO, AKANE - Rice University; O'CONNOR, TERESIA - Children'S Nutrition Research Center (CNRC)

Submitted to: Scientific Reports
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/25/2024
Publication Date: 11/30/2024
Citation: Kumar Vadathya, A., Garza, T., Alam, U., Ho, A., Musaad, S.M., Beltran, A., Moreno, J.P., Baranowski, T., Haidar, N., Hughes, S.O., Mendoza, J.A., Veeraraghavan, A., Young, J., Sano, A., O'Connor, T.M. 2024. Validation studies of the FLASH-TV system to passively measure children's TV viewing. Scientific Reports. 14. Article 29805. https://doi.org/10.1038/s41598-024-81136-0.
DOI: https://doi.org/10.1038/s41598-024-81136-0

Interpretive Summary: Children’s screen use, including TV, mobile devices, computers and gaming systems, have been associated with several health, developmental and psychosocial outcomes. However, to date almost all studies have relied on child- or parent-report of general estimates of the time children engaged with such screens. Passive, objective assessment tools are needed to more accurately measure children’s exposure to different digital screens, including TVs and large gaming systems. Many children still use such screens, despite an increase in use of mobile devices as well. We have previously published on the development of the FLASH-TV 1.0 system that allows for passive assessment of children’s TV viewing or videogame playing on TV screens. This paper described the enhancements of the system to FLASH-TV 3.0 and three independent validation studies to assess FLASH-TV 3.0 ability to accurately capture children’s TV viewing time compared to gold standard (human labeled video data for the child’s gaze on the TV). FLASH-TV relies on video images captured in front of a TV and processes the images in three sequential steps: face detection, face recognition, and gaze detection, using machine learning algorithms. The video data is then deleted and only the processed output retained. The three validation studies included a pilot study in our observation lab with 10 families, a validation study in our observation lab with 30 families, and validation study in children’s homes with 20 families. For each family a target child between the ages of 5-12 years old was identified. Sixty children completed the studies and 59 had FLASH-TV data that could be analyzed. When combining the data from these 59 children, FLASH-TV 3.0 performed well compared to the gold standard, with a mean accuracy of 85%, mean sensitivity of 80%, mean specificity of 86%. The mean intra-class correlation coefficient (ICC) of child's gaze on TV (TV viewing) durations of FLASH-TV 3.0 to gold-standard was good at 0.86. We deliberately recruited an ethnically and racially diverse sample of children for the studies, to test if FLASH-TV performed as well across diverse children. Overall, FLASH-TV 3.0 correlated well with the gold standard across a diverse sample of children, but with higher variability among Black children than others. These studies showed that FLASH-TV provides an important new tool to estimate children's TV viewing and increase the precision of research regarding the effect of TV viewing's on children's health.

Technical Abstract: TV viewing is associated with health risks, but existing measures of TV viewing are imprecise due to relying on self-report. We developed the Family Level Assessment of Screen use in the Home (FLASH)-TV, a machine learning pipeline with state-of-the-art computer vision methods to measure children's TV viewing. In three studies, lab pilot (n=10), lab validation (n=30), and home validation (n=20), we tested the validity of FLASH-TV 3.0 in task-based protocols which included video observations of children for 60 min. To establish a gold-standard to compare FLASH-TV output, the videos were labeled by trained staff at 5-second epochs for whenever the child watched TV. For the combined sample with valid data (n=59), FLASH-TV 3.0 provided a mean 85% (SD 8%) accuracy, 80% (SD 17%) sensitivity, 86% (SD 8%) specificity, and 0.71 (SD 0.15) kappa, compared to gold-standard. The mean intra-class correlation (ICC) of child's TV viewing durations of FLASH-TV 3.0 to gold-standard was 0.86. Overall, FLASH-TV 3.0 correlated well with the gold standard across a diverse sample of children, but with higher variability among Black children than others. FLASH-TV provides a tool to estimate children's TV viewing and increase the precision of research on TV viewing's impact on children's health.