Julio Villalón-Reina has visited ML4PNP

We were delighted to welcome Julio E. Villalón-Reina from the University of Southern California for a research visit and presentation at the ML4PNP Lab.

Julio presented his recent work on an important practical question in normative modelling: how much data do we need to build reliable and well-calibrated normative models?

The presentation followed up on our recent collaborative work, “Lifespan normative modeling of brain microstructure”, published in Nature Communications. In this study, he developed lifespan normative models of white-matter microstructure using diffusion MRI data from more than 54,000 individuals across 19 international datasets, spanning ages 4–91 years.

Building on this large-scale resource, Julio’s latest work examines how training sample size affects the distributional and centile calibration of normative models. Using hierarchical Bayesian regression and progressively larger training samples, the study investigates when normative estimates become sufficiently stable and reliable—a key question for designing future normative modelling studies and translating these methods to new datasets and clinical applications.

The work is described in the recent preprint “Distributional and Centile Calibration of Diffusion Tensor Imaging Normative Models as Training Sample Sizes Increase to 40,000 Subjects”.

The visit also provided a great opportunity to discuss ongoing developments in normative modelling, model calibration, and the methodological requirements for building reliable population reference models.

Thank you, Julio, for visiting us and for the stimulating presentation and discussions!