Abstract | ||
---|---|---|
We propose an efficient online reconstruction algorithm for the problem of highly undersampled dynamic magnetic resonance imaging (DMRI). Our approach reconstructs the dynamic time series by processing only a small batch of frames at a time. We adapt an online subspace tracking algorithm based on manifold optimization to the DMRI reconstruction setting and propose a novel extension of the algorithm to enable robust subspace tracking based on a local low-rank plus transform sparse model. Our experiments on real and synthetic data show that proposed approach gives comparable results to methods that reconstruct the entire image series at once while requiring only a fraction of the memory and computational demand. The dramatic memory savings allows robust subspace-based methods to be applied to much larger datasets than previously allowed. |
Year | Venue | Keywords |
---|---|---|
2017 | IEEE Global Conference on Signal and Information Processing | Online algorithms,dynamic MRI reconstruction,Grassmannian optimization,robust PCA,low-rank plus sparse |
Field | DocType | ISSN |
Computer vision,Online algorithm,Subspace topology,Sparse model,Computer science,Image Series,Synthetic data,Reconstruction algorithm,Artificial intelligence,Dynamic contrast-enhanced MRI,Manifold | Conference | 2376-4066 |
Citations | PageRank | References |
0 | 0.34 | 10 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Greg Ongie | 1 | 67 | 8.18 |
Saket Dewangan | 2 | 0 | 0.34 |
Jeffrey A. Fessler | 3 | 5 | 1.10 |
Laura Balzano | 4 | 410 | 27.51 |