Using Consumer Devices to Monitor Acute Medical Conditions for Infants

Abstract

Acute medical conditions need immediate attention, but early detection can require professional experience and specialized equipment that are unavailable at home. Consequently, babies with such conditions risk suffering damage from late interventions. We can leverage the world’s increasingly ubiquitous devices to improve the accessibility of health care outside the hospital through machine learning and integrating a human-centered approach at every step of the process. This dissertation examines this approach through three projects: a smartphone-based system to screen newborns for dangerous levels of jaundice, an exploration on how machine learning can help an existing system better monitor infants with single ventricle heart disease, and a reflection on the methods and insights from working in this space to inform future work.