eduzhai > Applied Sciences > Engineering >

Compressive MR Fingerprinting reconstruction with Neural Proximal Gradient iterations

  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: Consistency of the predictions with respect to the physical forward model ispivotal for reliably solving inverse problems. This consistency is mostlyun-controlled in the current end-to-end deep learning methodologies proposedfor the Magnetic Resonance Fingerprinting (MRF) problem. To address this, wepropose ProxNet, a learned proximal gradient descent framework that directlyincorporates the forward acquisition and Bloch dynamic models within arecurrent learning mechanism. The ProxNet adopts a compact neural proximalmodel for de-aliasing and quantitative inference, that can be flexibly trainedon scarce MRF training datasets. Our numerical experiments show that theProxNet can achieve a superior quantitative inference accuracy, much smallerstorage requirement, and a comparable runtime to the recent deep learning MRFbaselines, while being much faster than the dictionary matching schemes. Codehas been released at this https URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×