Pablo Villanueva-Perez: Physics-Informed AI for X-ray Imaging

Artificial Intelligence, and particularly deep learning, is transforming X-ray imaging by enabling fast, accurate analysis of massive datasets and uncovering complex correlations. At high-brilliance sources such as MAX IV, physics-informed deep learning offers the possibility to address long-standing reconstruction challenges without relying on traditional supervised learning or paired datasets.

In this talk, we present end-to-end frameworks that merge state-of-the-art AI with fundamental physics, i.e., modeling X-ray interactions, propagation, image formation, and dynamic processes. Specifically, we will discuss solutions for (i) phase retrieval under coherent illumination, (ii) 3D and 4D reconstructions from sparse spatiotemporal data, and (iii) efficient data analysis pipelines. These approaches demonstrate how integrating physics and AI enables high-speed, high-fidelity imaging, opening new frontiers in fields such as materials science and biology for academic and industrial applications.

Pablo Villanueva-Perez is an Associate Professor in Synchrotron Radiation Research at Lund University’s Department of Physics. He earned his PhD in theoretical physics in 2013 from the University of Valencia within the BaBar Collaboration at SLAC. Transitioning to X-ray imaging, he completed postdoctoral research at the Paul Scherrer Institute (TOMCAT group) and CFEL at DESY (Coherent X-ray Imaging group). Since joining Lund University in 2019, his work has focused on developing advanced X-ray techniques for 3D and 4D imaging across scales from micro to nano and timescales from seconds to femtoseconds. He frequently uses synchrotron and XFEL facilities for his research.