Advanced computational tools for real-time MR imaging

Abstract

Real-time Magnetic Resonance Imaging (MRI) has the potential of successfully guiding interventional applications. Overall, the requirements of real-time MRI can be categorized as: (i) fast data acquisition, (ii) fast image reconstruction, and (iii) good image quality. Fast data acquisition is provided by optimized real-time sequences, by parallel MRI (pMRI) techniques, or by non-Cartesian acquisition schemes (e.g. spiral and radial trajectories). However, fast image reconstruction is non-trivial, especially when computations demanding pMRI methods or non-Cartesian trajectories are involved. Even though signal-to-noise ratio (SNR) can be relatively high during real-time imaging, spatial resolution is limited. Thus, improved visual feedback during real-time MRI guided interventions is a must. This thesis defined three specific aims to improve real-time imaging: (i) real-time image reconstruction for pMRI, (ii) real-time image reconstruction for non-Cartesian trajectories, and (iii) fast MRI post-processing for improved visual feedback during interventions. Thesis contributions include: (i) real-time hybrid domain TGRAPPA based pMRI reconstruction algorithm (currently the fastest TGRAPPA based algorithm), (ii) first real-time implementation of GRAPPA Operator Gridding algorithm for radial acquisitions, (iii) multi-phase 3D angiography roadmaps for MRI guided interventions, (iv) improved active device visualization during real-time MRI guided interventions, (v) integration of a real-time active device localizer algorithm.|Keywords: Real-time MRI, pMRI, Non-cartesian trajectories, Parallel processing, MRI guided interventions.