Meeting Banner
Abstract #4199

Compressed Sensing with Prior Information for Time-Resolved TurboSPI

James A. Rioux1, 2, Steven D. Beyea2, 3, Chris V. Bowen2, 3

1Department of Physics, Dalhousie University, Halifax, Nova Scotia, Canada; 2Institute for Biodiagnostics (Atlantic), National Research Council, Halifax, Nova Scotia, Canada; 3Departments of Physics, Radiology and Biomedical Engineering, Dalhousie University, Halifax, Nova Scotia, Canada

TurboSPI can be used to acquire a time series of images suitable for high temporal resolution relaxometry, but must be significantly accelerated to permit applications such as cellular imaging in vivo. The acquisition of a matched Fast Spin Echo image provides prior information to further sparsify the reconstructed image. We compare two approaches to incorporating this prior information, and show that a modified-CS reconstruction using the known support of the FSE template allows acceleration factors of up to 30 while retaining high image quality throughout the time series.

Abstract PDF E-Poster Video

Keywords

abdomen accelerated accelerating acquisition acquisitions adjacent allow allows appears applications applied approaches assist assuming attempt biomedical blurring boxes cell closely coil common components compressed computing console constrain contrast course dashed degradation denotes density depends described details domain dynamic earlier either equivalent even eventually example excitation exploring fine fixed fully gradient guide guides heavily held highly horizontal hundred identical impact importance improvements in vivo incorporating increasingly influence influencing initial intended interference known leading library likely linear location longitudinal loss magnitude matches matrix minutes mitigate modified normalized nova offers optimal overall patterns peak permit physics preceding prescribe previously prior probability pulse quality ranging rapidly readout reasonable reconstruct reconstructed reconstructing reconstruction reconstructions reduced reflects remain residual resolution resolved retains root sampled sampling saturation schemes sensing series significantly since slab slightly solid sorted space sparse sparsity spin submitted subsequent suitable suited support system target template temporal terms though throughout trans transform variable varying volume volumes water wavelet