Thomas Ernst (UMD)

Title:

Abstract:

Magnetic resonance imaging (MRI) is ideal for assessing the structure, physiology, and function of the human brain in vivo. MRI affords high spatial and temporal resolution and is non-invasive and repeatable. Over the past decades, MRI has seen impressive improvements in image quality. For instance, an early (1980s) brain MRI scan at a resolution of 1x1x5mm may have taken 10 minutes. On a modern scanner, the same contrast can be acquired routinely at sub- mm resolution (0.7x0.7x1mm) within 2-3 minutes. These dramatic improvements are the combined result of multiple innovations, including higher magnetic field strengths, better gradients (which encode spatial information), use of phase-array receiver coils, undersampling of k-space, and improved reconstruction techniques.

However, with ever-improving image resolution, subject motion becomes an increasing problem, and patients are commonly required to lie very still (to <1mm/1°) during MRI scans for optimal image quality. These strict requirements are difficult to meet by many individuals, including children, individuals who are sick or agitated, and patients with movement disorders and dementia. The inability to remain still results in images with motion artifacts that may be non-diagnostic or require repeat scans.

Our presentation will describe some recent techniques to attenuate or eliminate motion artifacts, with a focus on mathematical aspects. MRI data are acquired directly in a complex Fourier (k-)space, but movements cause signal errors that result in artifacts in reconstructed images. We will present some previous and newly developed methods to correct motion- induced signal errors, especially in the context of undersampled k-space trajectories. Finally, we will highlight some unsolved mathematical problems in motion correction.