Bayesian Learning Maps Alzheimer's Disease Progression Continuously

Alzheimer's disease (AD) is understood as a continuous biological process, yet many existing AI methods for neuroimaging-based diagnosis are limited to discrete classifications or cross-sectional predictions. This study addresses this gap by proposing Disease Continuum Positioning (DCP), a novel longitudinal Bayesian Learning framework. DCP is designed to continuously estimate the severity of AD progression using longitudinal diffusion tensor imaging (DTI) data. The framework models disease severity as a low-dimensional probabilistic latent variable, integrating both longitudinal observations and weak clinical supervision. From this model, a Disease Continuum Score (DCS) is derived, which quantifies an individual's precise position along the AD continuum, complete with associated uncertainty estimates. Extensive evaluations on the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort demonstrate that DCP consistently surpasses the performance of other representative disease progression methods. Crucially, validation analyses confirm that DCS accurately reflects disease severity, shows strong clinical relevance, preserves the natural longitudinal evolution of the disease, and effectively predicts future disease conversion. These findings suggest that DCS provides a valuable quantitative, imaging-derived representation for assessing AD progression beyond traditional diagnostic labels and clinical scores.