Christopher L. Welsh1, 2, Edward W. Hsu1, Ganesh Adluru2, Jeffrey S. Anderson2, Edward VR DiBella2
1Department of Bioengineering, University of Utah, Salt Lake City, UT, United States; 2UCAIR, Department of Radiology, University of Utah, Salt Lake City, UT, United States
Diffusion Tensor Imaging (DTI) is useful for characterizing tissue microstructure, but suffers from low temporal resolution and the associated low spatial resolution and SNR. A model-based strategy is presented to reconstruct undersampled multi-coil DTI data, accelerating acquisition. The reconstruction is performed by fitting the diffusion tensor directly to the acquired data via minimizing a spatially regularized cost function. The proposed approach, along with traditional compressed sensing, GRAPPA and SENSE, are compared against the fully sampled case. SENSE performs the best in terms of fiber orientation estimation, while the model-based approach performs the best in terms of FA estimation.