Preterm infants, AI guides personalized nutrition
Artificial intelligence as a predictive tool to support the definition of nutrition in preterm infants. This is the core of the study published in the Journal of Perinatology (Nature portfolio), carried out by researchers from the Fondazione IRCCS San Gerardo dei Tintori (FSGT) and the Department of Electronics, Information and Bioengineering (DEIB) of Politecnico di Milano.
The study analyzes the delicate transition from intravenous (parenteral) to oral (enteral) feeding in extremely preterm infants, a crucial phase for growth and development, still lacking standardized approaches and associated with possible risks related to inadequate nutritional intake.
By analyzing over one thousand electronic medical records of very preterm infants followed in a single center, the team developed predictive models capable of identifying the key factors associated with the risk of extrauterine growth restriction EUGR (Extrauterine Growth Restriction). The results highlight how adequate protein and lipid intake in the first days of life, together with growth rate during the first week, represent determining factors. Furthermore, dividing patients into different prematurity profiles revealed differing nutritional needs, paving the way for increasingly personalized strategies.
Artificial intelligence makes it possible to integrate large volumes of heterogeneous clinical data and transform them into useful tools for research and, progressively, to support clinical decision-making. Value arises from the meeting of expertise: on the one hand methodological rigor and the ability to interpret data complexity, on the other clinical knowledge that gives meaning to numbers and guides the right questions. This is how models become not only accurate, but also interpretable and potentially transferable into practice.
Simona Ferrante, Department of Electronics, Information and Bioengineering