Charting Brain Growth and Aging at High Spatial Precision

Abstract

Defining reference models for population variation, and the ability to study individual deviations is essential for understanding inter-individual variability and its relation to the onset and progression of medical conditions. In this work, we assembled a reference cohort of neuroimaging data from 82 sites (N=58,836; ages 2-100) and use normative modeling to characterize lifespan trajectories of cortical thickness and subcortical volume. Models are validated against a manually quality checked subset (N=24,354) and we provide an interface for transferring to new data sources. We showcase the clinical value by applying the models to a transdiagnostic psychiatric sample (N=1,985), showing they can be used to quantify variability underlying multiple disorders whilst also refining case-control inferences. These models will be augmented with additional samples and imaging modalities as they become available. This provides a common reference platform to bind results from different studies and ultimately paves the way for personalized clinical decision making.

Non-Gaussian Normative Modelling With Hierarchical Bayesian Regression

“…Furthermore, the different UK-biobank datasets [16] were merged, because the datasets are known to be wellcontrolled for site effects. Counts per site and sex can be found in section Appendix E. For full details about the data, we refer to Rutherford et al [15]. Most data in this dataset are publicly available, but for several of the clinical samples, we do not have explicit permission for data sharing.…”

“…To allow a meaningful comparison with the current methods, we adapt and use the data from Rutherford et al [18]. This is a large neuroimaging dataset collected from 82 different scanners, containing 58834 control subjects and 2925 patient subjects 5 .…”

The Normative Modeling Framework for Computational Psychiatry

“…Validity is arguably more challenging to assess but should be established by means of out of sample model fit. In other recent work, normative models were fit using a lifespan (age 3-100) big data sample (N=58,836) and carefully tested out-of-sample (variance explained, skewness, kurtosis, and standardized mean squared error) showing excellent model fit (12-68% variance explained) in an independent test set from a sample (and site) that was not included in the training set 85 . This work suggests validity, but this is an on-going evaluation and out of sample model fit must always be considered and reported.…”

Using normative models pre-trained on cross-sectional data to evaluate longitudinal changes in neuroimaging data

“…Normative modelling is a relatively new area of research and thus, despite its potential, longitudinal normative models have not been systematically explored [6, 13]. Indeed, virtually all large-scale normative models released to date are estimated on cross-sectional data [6, 14] and a recent report [13] has provided empirical data to suggest that such cross-sectional models may underestimate the variance in longitudinal data [13]. However, from a theoretical perspective, it is very important to recognise that cross-sectional models describe group-level population variation across the lifespan, where such group level centiles are interpolated smoothly across time.…”

“…Training these models requires significant amounts of data and computational resources, limiting their use for smaller research groups. However, the availability of pre-trained models has made them more accessible to researchers from a wider range of backgrounds, as reported by Rutherford et al [14].…”