Advancing Neonatal Health: AI-enabled Ultrasound for Enhanced Growth and Nutritional Assessment - Project Summary/Abstract Preterm infants are at high risk for suboptimal growth and neurodevelopmental delays, yet current tools for monitoring nutrition, such as weight, length, and BMI, lack sensitivity to changes in body composition. Advanced methods like air displacement plethysmography provide accurate measures but are costly, impractical for routine use, and infeasible in critically ill infants. This critical gap leaves clinicians without effective tools for monitoring lean and fat tissue accretion during the neonatal intensive care unit (NICU) stay, when interventions can have the greatest impact. Preliminary data from our group demonstrate that ultrasound imaging, when combined with artificial intelligence (AI), can predict body composition with reasonable accuracy. Building on this foundation, the overall objective of this proposal is to improve growth assessment in preterm infants by developing and validating the first AI-enabled ultrasound system designed for longitudinal, point-of-care nutritional monitoring in preterm infants. This system will be low-cost, portable, and fully automated, providing standardized, real-time, non-invasive assessment of lean and fat tissue accretion. Using a prospective, longitudinal design, we will enroll 160 preterm infants across two sites, collecting ultrasound scans, gold-standard body composition measures during hospitalization, and maternal and clinical data, with follow-up through 24 months post-discharge. Our approach will focus on three goals: (1) validating and standardizing ultrasound protocols with real-time feedback mechanisms to ensure high-quality scans; (2) developing and refining AI models that integrate imaging, maternal, and clinical data to predict lean and fat mass with high accuracy; and (3) linking early ultrasound- derived nutritional trajectories to neurodevelopmental outcomes to identify at-risk phenotypes and inform targeted interventions. The expected outcomes of this study are validated scanning protocols, an integrated AI- driven framework for body composition assessment, and new insights into growth trajectories that predict developmental outcomes. This project is innovative in combining real-time ultrasound imaging with advanced AI analytics to create a scalable, clinically feasible tool that has not yet been applied in preterm infants for longitudinal body composition monitoring. The results will improve early detection of growth failure, guide individualized nutritional interventions, and ultimately reduce the burden of adverse health outcomes in this vulnerable population.
