eduzhai > Applied Sciences > Engineering >

MAGIC Manifold and Graph Integrative Convolutional Network for Low-Dose CT Reconstruction

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 17 pages

Abstract: Low-dose computed tomography (LDCT) scans, which can effectively alleviatethe radiation problem, will degrade the imaging quality. In this paper, wepropose a novel LDCT reconstruction network that unrolls the iterative schemeand performs in both image and manifold spaces. Because patch manifolds ofmedical images have low-dimensional structures, we can build graphs from themanifolds. Then, we simultaneously leverage the spatial convolution to extractthe local pixel-level features from the images and incorporate the graphconvolution to analyze the nonlocal topological features in manifold space. Theexperiments show that our proposed method outperforms both the quantitative andqualitative aspects of state-of-the-art methods. In addition, aided by aprojection loss component, our proposed method also demonstrates superiorperformance for semi-supervised learning. The network can remove most noisewhile maintaining the details of only 10 (40 slices) of the training datalabeled.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×