We proposed a dual-domain self-supervised motion artifacts disentanglement network (DSMAD-Net) for the liver's gadoxetic acid-enhanced arterial phase images. The motion correction is converted to the image-to-image translation problem by assuming that motion-free images and motion-corrupted images belong to different domains. Specifically, image-to-image translation within the same domain is designed to constrain auto-encoders to learn the feature representation by utilizing the input images as supervision information. Moreover, the cross-domain translation explores the cycle consistency in the absence of paired motion-free and motion-corrupted images. Experimental results demonstrate that our method remarkably removes artifacts in the gadoxetic acid-enhanced arterial phase images.
This abstract and the presentation materials are available to members only; a login is required.