Ordinal Classification with Distance Regularization for Robust Brain Age Prediction

Age is one of the major known risk factors for Alzheimer's Disease (AD).Detecting AD early is crucial for effective treatment and preventingirreversible brain damage. Brain age, a measure derived from brain imagingreflecting structural changes due to aging, may have the potential to identifyAD onset, assess disease risk, and plan targeted interventions. Deeplearning-based regression techniques to predict brain age from magneticresonance imaging (MRI) scans have shown great accuracy recently. However,these methods are subject to an inherent regression to the mean effect, whichcauses a systematic bias resulting in an overestimation of brain age in youngsubjects and underestimation in old subjects. This weakens the reliability ofpredicted brain age as a valid biomarker for downstream clinical applications.Here, we reformulate the brain age prediction task from regression toclassification to address the issue of systematic bias. Recognizing theimportance of preserving ordinal information from ages to understand agingtrajectory and monitor aging longitudinally, we propose a novel ORdinalDistance Encoded Regularization (ORDER) loss that incorporates the order of agelabels, enhancing the model's ability to capture age-related patterns.Extensive experiments and ablation studies demonstrate that this frameworkreduces systematic bias, outperforms state-of-art methods by statisticallysignificant margins, and can better capture subtle differences between clinicalgroups in an independent AD dataset. Our implementation is publicly availableat https://github.com/jaygshah/Robust-Brain-Age-Prediction.