Advances in Concurrent Motion and Field-Inhomogeneity Correction in Functional MRI.

Head motion and static magnetic field (B0) inhomogeneity are two important sources of image intensity variability in functional MRI (fMRI). Ideally, in MRI, any deviation in B0 homogeneity in an object occurs only by design. However, due to imperfections in the main magnet and gradient coils, and, magnetic susceptibility differences in the object, undesired B0 deviations may occur. This causes geometric distortion in Cartesian EPI images. In addition to spatial shifts and rotations of images, head motion during an fMRI experiment may induce time-varying field-inhomogeneity changes in the brain. As a result, correcting for motion and field-inhomogeneity effects independently of each other with a static field map may be insufficient, especially in the presence of large out-of-plane rotations. Our primary concern is the correction of the combined effects of motion and field-inhomogeneity induced geometric distortion in Cartesian EPI fMRI images. We formulate a concurrent field-inhomogeneity with map-slice-to-volume motion correction, and develop a motion-robust dual-echo bipolar gradient echo static field map estimation method. We also propose and evaluate a penalized weighted least squares approach to dynamic field map estimation using the susceptibility voxel convolution method. This technique accounts for field changes due to out-of-plane rotations, and estimates dynamic field maps from a high resolution static field map without requiring accurate image segmentation, or the use of literature susceptibility values. Experiments with simulated data suggest that the technique is promising, and the method will be applied to real data in future work. In a separate clinical fMRI project, which is independent of the above work, we also formulate a current density weighted index to quantify correspondence between electrocortical stimulation and fMRI maps for brain presurgical planning. The proposed index is formulated with the broader goal of defining safe limits for lesion resection, and is characterized extensively with simulated data. The index is also computed for real human datasets.PhDElectrical Engineering: SystemsUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/60787/1/tbyeo_1.pd