Ultrafast Reconstruction of UnderSampled (URUS) 3D Late Gadolinium Enhancement Imaging using Deep Complex Network
Hossam El-Rewaidy1, Warren J Manning1,2, and Reza Nezafat1
1Department of Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, 2Department of Radiology, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States
A new framework based on deep complex network is introduced for fast reconstruction of undersampled 3D CMR data. In this work, the complex nature of the acquired data was maintained throughout the network to achieve optimal performance. High resolution 3D LGE dataset of 217 subjects was acquired at prospective acceleration rates ranging from 3 to 5 and used to train and validate (7:3) the proposed methods. Further retrospective undersampling was performed to test the capacity of the proposed methods to perform at higher acceleration rates. LGE scar quantification showed high correlation with the ground-truth.
This abstract and the presentation materials are available to members only;
a login is required.