One Scan, Many Tomorrows: Predicting Longitudinal Brain Development with Implicit Neural Representations

Why seeing a baby brain’s “future” matters right now

Can neural implicit representations predict brain developmment. Image created with DALL-E.

In the last trimester of pregnancy and the first weeks after birth, the human brain grows and folds at breakneck speed. Subtle deviations in this choreography can foreshadow later challenges in cognition, motor function, or learning. Clinicians and scientists therefore want not only average growth charts, but individualized forecasts — credible, subject-specific “tomorrows” derived from the scan of a single fetus or neonate today. A new MICCAI 2025 paper by Maik Dannecker and Daniel Rueckert shows how to do exactly that by turning each brain into a continuous mathematical function that can be steered forward and backward in time. It is an approach that replaces vague generalities with precise, personalized trajectories, and it does so with surprising data efficiency.

How this work raises the bar

Population atlases are wonderful at summarizing what a typical brain looks like at 28 or 40 weeks, but they blur away the idiosyncrasies that make each infant different — and those differences matter most when you are trying to detect risk. Dannecker and Rueckert introduce the first atlas designed to predict individualized…