Ander Biguri Institute: University of Cambridge Date: November 14, 2024 Title: Combining the mathematics of inverse problems and machine learning for CT reconstruction: the LION toolbox Abstract: Data-Driven methods are being overwhelmingly used in medicine, including for CT image reconstruction. However, blindly reconstructing the images with a purely learned method has drawbacks, as there is no guarantee that the final image matches the measured data. In the past years, the inverse problems community have worked in producing methods that mix traditional CT reconstruction methods with data-driven methods, sometimes achieving mathematical guarantees of convergence, which are desired. This talk introduces the families of methods for CT reconstruction and showcases the LION toolbox, a python library for learned reconstruction research.
