A validated deep learning workflow for 3D quantification of amyloid plaques in cleared mouse brain

Abstract Accurate quantification of amyloid plaque pathology is essential for understanding Alzheimer's disease (AD) progression and evaluating therapeutic interventions. Yet, conventional histology relies on sampling thin tissue sections and therefore incompletely captures three‐dimensional (3D) plaque distributions, introducing bias and limiting scalability. Here, we present a validated and scalable 3D analysis pipeline for whole‐brain quantification of amyloid plaques in cleared mouse brain tissue imaged by light‐sheet microscopy. This pipeline also integrates deep learning‐based segmentation with atlas registration to enable automated, region‐specific quantification across intact brain volumes. Quantitative measurements obtained using this approach showed strong agreement with gold‐standard stereology, while enabling higher throughput and comprehensive volumetric assessment of amyloid plaque burden. As a demonstration of its utility in a preclinical setting, we applied the pipeline to assess the neuroprotective effects of chronic saffron supplementation in APP/PS1 mice, identifying a significant reduction in hippocampal plaque load ( p < 0.01). Together, these findings establish an accessible and validated framework for scalable 3D quantification of amyloid pathology, providing a platform for neuropathological phenotyping and therapeutic evaluation in preclinical AD research.