Akademik Veri Yönetim Sistemi
12th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2021 held in conjunction with MICCAI 2021, Strasbourg, Fransa, 27 Eylül 2021, cilt.13131 LNCS, ss.199-207
© 2022, Springer Nature Switzerland AG.Artefacts constitute a paramount issue in medical imaging where the prevalence of artefacts may severely impact the clinical diagnosis accuracy. Specifically, the mistriggering and motion family of artefacts during the cardiac MR image acquisition would eventually damage the visibility of certain tissues, such as the left ventricular, right ventricular, and myocardium. This would cause the ejection fraction to be incorrectly estimated and the patient’s heart condition to be incorrectly evaluated. Despite much research on medical image reconstruction, relatively little work has been done for cardiac MRI artefact correction. In this work, inspired by the image reconstruction literature, we propose to use an auto-encoder-guided mistriggering artefact correction method, which not only corrects the artefacts in the image domain but also in the k-space domain with the introduction of an enhanced structure. We conduct a variety of experiments on photos and medical images to compare the performances of different network architectures under mistriggering artefacts and gaussian noise. We demonstrate the superiority of the cross-domain network in the case of k-space-related artefacts.