AI to predict extrauterine growth restriction during transitional nutrition of preterm infants: a retrospective study

Authors: Valentina Bozzetti, Linda Greta Dui, Emanuela Zannin, Silvia Riccò, Paola Coglianese, Valeria Cavalleri, Lucia Iozzi, Maria Luisa Ventura & Simona Ferrante

Keywords

AI, Machine Learning, Preterm infants, Nutrition

Summary

Artificial intelligence as a predictive tool to optimize nutrition in premature newborns. This is the premise behind an innovative study recently published in the Nature portfolio journal Journal of Perinatology, resulting from the work of a group of researchers from the Fondazione IRCCS San Gerardo dei Tintori and the Department of Electronics, Information and Bioengineering (DEIB) at Politecnico di Milano.

The study addresses one of the most delicate stages in the care of preterm infants: the transition from intravenous (parenteral) to oral (enteral) nutrition. This “nutritional transition” phase is extremely critical for neonatal growth and development, since suboptimal management — either excessive or insufficient — may lead to several complications and result in a slowdown of so-called extrauterine growth restriction (EUGR), a condition affecting up to 90% of premature infants and potentially compromising their subsequent neurocognitive development.

To date, there are no international guidelines defining how this transition should be managed. It remains unclear how much of the nutrients administered intravenously and orally is actually absorbed, making it difficult to properly balance the intake of proteins, lipids, and calories. However, through the use of artificial intelligence techniques and the analysis of more than 15 years of clinical and nutritional data from over one thousand premature newborns, the research group developed predictive models capable of identifying the factors that most strongly influence growth during this critical phase. The results show that an adequate intake of proteins and lipids already in the first days of life, together with a good growth rate during the first week, are key elements in predicting EUGR.

Another major contribution of this work was the stratification of patients according to the severity of prematurity. The study demonstrated that nutritional requirements differ across patient groups, paving the way for more personalized care approaches. By integrating neonatal medicine and artificial intelligence, the study provides an initial scientific basis for optimizing nutritional management during this delicate transition phase and for defining more effective and personalized clinical protocols.

For futher details, here is the link for the open access publication.