Improving temporal resolution in fMRI using a 3D spiral acquisition and low rank plus sparse (L+S) reconstruction

摘要: Abstract Rapid whole-brain dynamic Magnetic Resonance Imaging (MRI) is of particular interest in Blood Oxygen Level Dependent (BOLD) functional MRI (fMRI). Faster acquisitions with higher temporal sampling the BOLD time-course provide several advantages including increased sensitivity detecting activation, possibility filtering out physiological noise for improving SNR, and freezing head motion. Generally, faster require undersampling data which results aliasing artifacts object domain. A recently developed low-rank (L) plus sparse (S) matrix decomposition model (L+S) one methods that has been introduced to reconstruct images from undersampled data. The L+S approach assumes data, represented as a space-time M, linear superposition L S components, where represents highly spatially temporally correlated elements, such image background, while captures information an appropriate transform This suggests might be suited task or slow event-related fMRI because periodic nature signal Fourier domain slowly varying brain background signals, drift, will predominantly low-rank. In this work, proof concept, we exploit method accelerating block-design using 3D stack spirals (SoS) acquisition performed k z − t We examined feasibility accurately separate component capturing signals component. present acquired control human volunteers at 3 T both retrospective prospectively visual activation task. show SoS acceleration four reconstruction can achieve coverage 40 slices 2 mm isotropic resolution 64 x size every 500 ms.