Accelerated cardiac mri by direct inversion of constrained dynamic modelsmethods, optimization, and applications

Resumo

Speed is an important factor in many dynamic MRI applications including cardiac imaging. Respiratory artifacts are commonly avoided by acquiring all data during breath-holding. An important limiting factor in cardiac imaging is thus the breath-hold duration that the subject can sustain. For patients with severe cardiovascular disease, and even for healthy subjects, there are limits to both the breath-hold duration and the number of consecutive breath-hold scans that can be acquired. Faster acquisitions allow the same image to be acquired in a shorter breath-hold, higher-resolution images to be acquired in the same breath-hold time, and complete multislice coverage of anatomy to be completed in fewer breath-holds by scanning multiple slices in one breath-hold. When imaging time pushes the limits of the patient's breath-hold capacity, shorter scan times may actually improve the quality of the images. Reduced breath-hold times may improve patient throughput and result in more successfully completed studies. Dynamic imaging often involves repeated acquisitions of an image plane in which only part of the FOV changes over time. For example, in cardiac MRI the heart moves during the cardiac cycle, but the lungs and shoulders remain in the same location. As a rule, the dynamic parts of the image contain the interesting structures in the image. Sometimes static regions provide an anatomical reference, but often the only reason for including static regions in the FOV is to avoid contamination of relevant dynamic regions by aliasing or wraparound artifacts. Conventional dynamic MRI acquires sufficient data to reconstruct these static regions independently for each image of a time sequence. Such data are intrinsically redundant, since ideally the static region is identical in all images. In this thesis, a novel acceleration strategy called "Noquist" is presented. Using Noquist, a more sparsely sampled dynamic image sequence is reconstructed cor