Individualised prediction of infant and toddler brain growth using longitudinal normative models

Motivation: Normative modelling based growth charts see frequent use for assessment of childhood development, with a recent spate of work focusing on mapping MRI derived neuroanatomical changes over early life. We propose adding longitudinal information to traditionally cross-sectional models to enhance predictive ability and allow creation of individual developmental trajectories. Goal(s): Investigate if longitudinal models of regional brain volume growth allow more accurate prediction of early life developmental trajectories. Approach: Construct cross-sectional and longitudinal models for multiple brain regions in high-/low-field cohorts. Assess cross-sectional and longitudinal model accuracy. Results: Longitudinal models outperform cross-sectional models with lower error in all regions except the globus pallidus. Impact: The addition of longitudinal information into regional volume growth models allows prediction of an individualised developmental trajectories. Comparisons can be made between an individual's actual regional volume and predicted developmental and rather than position within a traditional population distribution.