Deep learning-based MRI analysis reveals Lewy body co-pathology accelerates brain aging in Alzheimer's disease

INTRODUCTION: Lewy body (LB) co-pathology is common in Alzheimer's disease (AD), yet its in vivo impact on neurodegeneration remains unclear. METHODS: We trained a 3D-DenseNet on multi-cohort T1-weighted magnetic resonance imaging (MRI) from cognitively unimpaired individuals to estimate brain age (BA) and applied it to cognitively impaired Alzheimer's Disease Neuroimaging Initiative (ADNI) participants stratified by cerebrospinal fluid (CSF) p-tau₁₈₁/Aβ₄₂ (AD) and α-synuclein seed amplification assay (LB) status into: AD+LB+, AD+LB-, AD-LB+, and AD-LB-. We compared baseline and longitudinal brain-age gap (BAG), identified saliency-derived anatomical contributors, assessed regional atrophy and cognition, and evaluated whether BAG explains LB-related clinical decline within AD. RESULTS: The model robustly captured normative aging. AD+LB+ exhibited the greatest and fastest-increasing BAG, with saliency maps emphasizing regions that also showed steeper longitudinal atrophy and concordant cognitive decline, and BAG mediated a substantial portion of LB-related cognitive impairment within AD+. DISCUSSION: LB co-pathology confers an additional neurodegenerative burden in AD, underscoring the importance of combined biomarker assessment and targeted interventions.