2019/09/19 by Jan-Nico Zaech, Zaech, Jan-Nico, Cong Gao +11
Engineering · Medicine · #Advanced X-ray and CT Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Medical Imaging Techniques and Applications #Radiation Dose and Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1909.08868
openalex publication_date 2019/09/19 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Metal artifacts in computed tomography (CT) arise from a mismatch between\nphysics of image formation and idealized assumptions during tomographic\nreconstruction. These artifacts are particularly strong around metal implants,\ninhibiting widespread adoption of 3D cone-beam CT (CBCT) despite clear\nopportunity for intra-operative verification of implant positioning, e.g. in\nspinal fusion surgery. On synthetic and real data, we demonstrate that much of\nthe artifact can be avoided by acquiring better data for reconstruction in a\ntask-aware and patient-specific manner, and describe the first step towards the\nenvisioned task-aware CBCT protocol. The traditional short-scan CBCT trajectory\nis planar, with little room for scene-specific adjustment. We extend this\ntrajectory by autonomously adjusting out-of-plane angulation. This enables\nC-arm source trajectories that are scene-specific in that they avoid acquiring\n"poor images", characterized by beam hardening, photon starvation, and noise.\nThe recommendation of ideal out-of-plane angulation is performed on-the-fly\nusing a deep convolutional neural network that regresses a detectability-rank\nderived from imaging physics.\n