An Uncertainty-Aware, Shareable and Transparent Neural Network Architecture for Brain-Age Modeling

TL;DR

This paper presents an uncertainty-aware, shareable neural network for brain-age prediction that improves robustness and validation, providing reliable uncertainty estimates and preventing spurious associations in neuroimaging data.

Contribution

It introduces MCCQR, a novel neural network architecture that incorporates uncertainty quantification and is designed for sharing and transparency in brain-age modeling.

Findings

Lower error rates across multiple datasets.

Robust uncertainty estimates improve model reliability.

Enhanced power to detect brain-aging acceleration.

Abstract

The deviation between chronological age and age predicted from neuroimaging data has been identified as a sensitive risk-marker of cross-disorder brain changes, growing into a cornerstone of biological age-research. However, Machine Learning models underlying the field do not consider uncertainty, thereby confounding results with training data density and variability. Also, existing models are commonly based on homogeneous training sets, often not independently validated, and cannot be shared due to data protection issues. Here, we introduce an uncertainty-aware, shareable, and transparent Monte-Carlo Dropout Composite-Quantile-Regression (MCCQR) Neural Network trained on N=10,691 datasets from the German National Cohort. The MCCQR model provides robust, distribution-free uncertainty quantification in high-dimensional neuroimaging data, achieving lower error rates compared to existing models across ten recruitment centers and in three independent validation samples (N=4,004). In two examples, we demonstrate that it prevents spurious associations and increases power to detect accelerated brain-aging. We make the pre-trained model publicly available.