eduzhai > Applied Sciences > Engineering >

Unpaired Deep Learning for Accelerated MRI using Optimal Transport Driven CycleGAN

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Recently, deep learning approaches for accelerated MRI have been extensivelystudied thanks to their high performance reconstruction in spite ofsignificantly reduced runtime complexity. These neural networks are usuallytrained in a supervised manner, so matched pairs of subsampled and fullysampled k-space data are required. Unfortunately, it is often difficult toacquire matched fully sampled k-space data, since the acquisition of fullysampled k-space data requires long scan time and often leads to the change ofthe acquisition protocol. Therefore, unpaired deep learning without matchedlabel data has become a very important research topic. In this paper, wepropose an unpaired deep learning approach using a optimal transport drivencycle-consistent generative adversarial network (OT-cycleGAN) that employs asingle pair of generator and discriminator. The proposed OT-cycleGANarchitecture is rigorously derived from a dual formulation of the optimaltransport formulation using a specially designed penalized least squares cost.The experimental results show that our method can reconstruct high resolutionMR images from accelerated k- space data from both single and multiple coilacquisition, without requiring matched reference data.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×