Claudia Lölkes (Fraunhofer EMI): Multimodal Computed Tomography – Combining Deep Learning Methods and Iterative Reconstruction

Wed, 6. May 2026, 9-10.00 h in H-C 7326

In this talk, Claudia will present deep learning approaches for multimodal computed tomography reconstruction, which she explores as part of her PhD at the Fraunhofer Ernst-Mach-Institut (EMI).

Computed Tomography (CT) enables occlusion-free visualizations of the inside of the human body or of objects by combining projections from multiple angles. However, different imaging modalities come with individual trade-offs: x-ray CT provides high spatial resolution but poses a health risk due to ionizing radiation, while radar CT, for example, offers sensitive material characterization without harmful radiation but suffers from limited resolution. To combine the advantages of different imaging methods while compensating for their weaknesses, this research aims to develop multimodal reconstruction methods that integrate complementary imaging modalities with x-ray CT. Modelling the inverse problem of tomographic reconstruction from a Bayesian perspective, the key idea is to replace hand-crafted priors with learned conditional priors. The talk will cover the fundamentals of CT, introduce the Bayesian framework for multimodal reconstruction, and present the proposed deep learning approaches.

Claudia Lölkes is a researcher at the Fraunhofer-Institut für Kurzzeitdynamik, Ernst-Mach-Institut (EMI). She received her Master’s degree in Computer Science and Autonomous Systems from TU Darmstadt. She is passionate about combining machine learning methods with classical CT reconstruction to advance multimodal imaging.