2019/02/26 by Veronica Corona, Corona, Veronica, Angelica I. Avilés-Rivero +11
Medicine · #Medical Imaging Techniques and Applications #Advanced MRI Techniques and Applications #Radiomics and Machine Learning in Medical Imaging
paper · pdf · doi:10.48550/arxiv.1902.10025
This work addresses a central topic in Magnetic Resonance Imaging (MRI) which\nis the motion-correction problem in a joint reconstruction and registration\nframework. From a set of multiple MR acquisitions corrupted by motion, we aim\nat - jointly - reconstructing a single motion-free corrected image and\nretrieving the physiological dynamics through the deformation maps. To this\npurpose, we propose a novel variational model. First, we introduce an L2\nfidelity term, which intertwines reconstruction and registration along with the\nweighted total variation. Second, we introduce an additional regulariser which\nis based on the hyperelasticity principles to allow large and smooth\ndeformations. We demonstrate through numerical results that this combination\ncreates synergies in our complex variational approach resulting in higher\nquality reconstructions and a good estimate of the breathing dynamics. We also\nshow that our joint model outperforms in terms of contrast, detail and blurring\nartefacts, a sequential approach.\n