Computer Vision News - February 2021

14 Cone-beam (CB) Computed Tomography (CT) refers to modern CT systems which use flat-panel detectors acquiring two-dimensional X-ray images instead of a single line of values. They are commonly used e.g. in image-guided surgery as so called C-arm systems, in radiation therapy or in industrial computed tomography for inspecting wares. The current challenge with CBCT is that the added flexibility comes at the price of reduced data quality compared to standard diagnostic CT. This reduced quality leads to artifacts in the reconstructed images since the different X-ray images are inconsistent with each other. In my PhD work, I used this inconsistency between X-ray images to develop novel algorithms to compensate for artifacts in the reconstruction solely based on the raw measurement data itself. I addressed two different classes of artifacts: 1) Artifacts which are caused by the polychromatic spectrum of conventional X-ray tubes called beam-hardening artifacts; and 2) Artifacts which arise due to motion of the object or inaccurate system calibration which can be subsumed as geometry artifacts. Tobias Würfl (@wuerflts) recently completed his PhD at Friedrich-Alexander University Erlangen-Nürnberg , where he developed raw- data-based artifact compensation algorithms for Computed Tomography (CT). He currently works as an application developer forMagnetic Resonance Imaging on deep-learning-based reconstruction at Siemens Healthineers . Our readers might remember Tobias telling us about the workshop on Machine Learning for Medical Image Reconstruction (MLMIR) at the latest MICCAI . Congrats Tobias and best of luck from all of us! Congrats, Doctor!