Fifty-seventh annual meeting of the American association of physicists in medicine
SU-D-303-04: A Survey of MRI Distortion Measurements
Magnetic Resonance Imaging (MRI) is hampered by geometrical distortions that arise from inhomogeneity in the B0 field, gradient non-linearity, and patient-induced susceptibility artifacts. We provide a survey of precise and accurate distortion measurements made over eight years on a number of different scanner/protocol arrangements.
A phantom was originally developed for the Alzheimer's Disease Neuroimaging Initiative (ADNI Phantom) to provide detailed mapping of MRI image distortion. The phantom contains an array of polycarbonate spheres each filled with a copper sulfate solution to provide image contrast. Included are 158 1.0 cm diameter and 2 1.5 cm diameter fiducial spheres enclosed in a 20 cm diameter spherical urethane shell. The center of gravity of the spheres can be determined to sub-voxel precision through segmentation of the phantom images. A software analysis determines the actual locations of the spheres independent of the position and orientation of the phantom. The imaged locations of spheres are compared to the known locations yielding up to fourth-order measurements of scanner distortion. We have surveyed MR distortion measurements acquired by groups doing routine QA and longitudinal studies over an eight year period.
Distortion fields vary in magnitude and form from between scanners and protocols. Corrections applied by the scanner typically reduce distortion to 1 mm or less throughout a 20 cm diameter volume centered on isocenter. Corrections are typically less successful in one direction depending on the protocol. Distortion increases with distance from the isocenter and can be highly non-linear.
Distortion of MR images is inherent to current scanners. Distortion is a function of many factors and although often less than 1mm can be much greater depending on protocol, distance from isocenter, and other factors. Scanners are typically stable with respect to their distortion profile but variations are not uncommon.
Research supported by Image Owl, Inc., The Phantom Laboratory, Inc. and Raforninn ehf. Authors Mallozzi and Healy employed by The Phantom Laboratory, Inc. Author Goodenough is a consultant to The Phantom Laboratory, Inc. Authors Fredriksson, Kristbjornsson, and Olafsdottir are consultants to Image Owl, Inc.