journals.iucr.org
High-speed quantitative X-ray multi-contrast imaging with deep learning based modulated pattern analysis
This work presents a neural-network framework that accelerates X-ray multi-contrast imaging—simultaneously delivering absorption, phase, and dark-field, while maintaining high spatial resolution, and outperforms correlation-based methods in speed, achieving a favorable balance between resolution and throughput. The method is agnostic to the modulation source (sandpaper, coded masks, gratings), enabling flexible deployment across setups and supporting real-time 2D/3D quantitative imaging for high-speed, in situ studies in materials science and biomedical applications.