Open Collections

Abstract

Innovative digital health technologies present an opportunity to create affordable, sustainable, and scalable monitoring technology to expand access to care in low-income and middle-income countries. This study aims to develop computer algorithms that automatically estimate respiratory rate from videos of infants captured using a smartphone. Based on the gaps identified in previous research, this project focuses on using smartphone video data taken in environments containing lighting variations, with handheld videos containing movement artifacts. For this study, 57 videos were captured from 39 infants <3 months of age at Kamuzu Central Hospital in Malawi, East Africa. Five small paper stickers were applied to the infant’s skin to facilitate respiratory rate tracking. Implemented in MATLAB, algorithms were developed to identify stickers on the infants’skin, track sticker movement coming from respiratory rate, and estimate breath count. This study indicated that non-contact respiratory rate monitoring for infants could be an effective method of accurately estimating respiratory rate using smartphone videos with lighting variations and movement artifacts. However, given the high accuracy and reliability required for effective monitoring tools, the current methods developed require further development to be successful. Overall, this study's results align with previous research as lighting variation and movement artifacts presented as major technical challenges to achieving high accuracy of respiratory rate estimation.