eduzhai > Applied Sciences > Engineering >

Bidirectional Mapping Generative Adversarial Networks for Brain MR to PET Synthesis

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Fusing multi-modality medical images, such as MR and PET, can provide variousanatomical or functional information about human body. But PET data is alwaysunavailable due to different reasons such as cost, radiation, or otherlimitations. In this paper, we propose a 3D end-to-end synthesis network,called Bidirectional Mapping Generative Adversarial Networks (BMGAN), whereimage contexts and latent vector are effectively used and jointly optimized forbrain MR-to-PET synthesis. Concretely, a bidirectional mapping mechanism isdesigned to embed the semantic information of PET images into the highdimensional latent space. And the 3D DenseU-Net generator architecture and theextensive objective functions are further utilized to improve the visualquality of synthetic results. The most appealing part is that the proposedmethod can synthesize the perceptually realistic PET images while preservingthe diverse brain structures of different subjects. Experimental resultsdemonstrate that the performance of the proposed method outperforms othercompetitive cross-modality synthesis methods in terms of quantitative measures,qualitative displays, and classification evaluation.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×