Project Details

Low-Cost Video Artificial Intelligence for Neonatal Respiratory Distress in Resource-Limited Units

Carl Britto, MBBS, DPhil

Summary

BACKGROUND: Neonatal respiratory distress (nRD) is a leading cause of neonatal mortality in low- and middle-income countries, contributing to up to 35% of the 2.3 million global newborn deaths each year, including 0.2–0.4 million in India. Limited trained staff, infrastructure, and diagnostic capacity lead to delayed or inaccurate assessments. GAP: No objective, scalable, or low-cost tool currently exists for real-time, accurate assessment of nRD in resource-limited settings, and existing clinical frameworks such as IMNCI have low specificity and high misclassification rates. HYPOTHESIS: A low-cost, video-based AI model can generate an objective, continuous nRD severity score that correlates with clinical assessments and predicts the need for respiratory support escalation in LMIC NICUs. METHODS: We will develop and validate a hybrid deep-learning video model using retrospective recordings and synchronized clinical data from 2,184 neonates across 12 Indian NICUs, followed by prospective external validation in a new cohort at the same sites. Participants will include neonates monitored through an established tele-NICU system. RESULTS: Pending. IMPACT: A video-based AI tool for early detection of nRD could enable timely intervention, reduce preventable neonatal deaths, and support frontline providers in resource-limited settings. The work is scalable and foundational for future multicenter clinical trials and global deployment.