Meeting Banner
Abstract #3348

Reconstruction of TOF Images from Undersampled K-Data Using SENSE, GRAPPA, CS, CS-SENSE, SPIRiT, and L1-SPIRiT

Jerome Yerly1, 2, Michel Louis Lauzon, 23, Richard Frayne, 23

1Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada; 2Seaman Family MR Research Centre, Foothills Medical Centre, Calgary, AB, Canada; 3Departments of Radiology and Clinical Neurosciences, University of Calgary, Calgary, AB, Canada

Recent advances in image reconstruction from sparsely sampled k-space data, such as CS and parallel MR imaging, provide potential solutions to enable visualization of small cerebral vessels without increasing the total acquisition time. In this study, we investigated and compared SENSE, GRAPPA, CS, SPIRiT, CS-SENSE, and L1-SPIRiT techniques to accelerate time-of-flight 3-T MR imaging. The reconstructions involving an L1-norm regularization procedure (i.e., CS, CS-SENSE, and L1-SPIRiT) resulted in lower aliasing interference, but also less conspicuous small cerebral vessels due to blurring. The auto-calibrating techniques (i.e., GRAPPA, SPIRiT, and L1-SPIRiT) exhibited less sensitivity artifacts and most reliably depicted small vessels.

Abstract PDF E-Poster Video

Keywords

absolute accelerated accurate acquisition acquisitions advances aliased aliasing approaches arises arrow arteries artery artifacts assessed assessment best better blurring calibrating central cerebral challenging clinical close coarsest coefficients coil comparing composite compressed conjugate coronal coupled depends depict depicted described detail detecting diagnosis diseases domain domains edges electric enable encode energy enlarged error errors family feature field final fitting flight fold forcing frequencies fully function general graph head healthy help highly homogeneous indicate interfaces interference locations lowest maps measuring methodology noise normalized optimum overall paid parallel polynomial poorly potentially preserve probability processing qualitatively quantitatively radiology randomly recent reconstructed reconstruction reconstructions reduced regularization reliably rendered requirements requires resolution root rotated sampled sampling satisfy scale scanner scanners seaman selected sense sensing sensitivity sequentially simulated slices software source space sparsity spirit squares stacks star stroke suffered susceptibilities tailoring tends terms trans transferred transform unfold variation vascular vessel vessels view views visible visualizing volumes wavelet whole yielded