Quantifying MRI geometric distortion in tissue

Abstract

We present a method to quantify the MR field inhomogeneity geometric distortion to subpixel accuracy without using objects of known dimensions and without using an external standard such as CT. Our method may be used to quantify the geometric accuracy of MR images of anatomical structures of unknown geometry and also to test any geometry correction scheme. We have quantified the distortion in a tissue phantom and found the largest error to be approximately 2.8 pixels (1.8 mm) for Bo = 1.5 T, G = 3.13 mT/m and FOV = 160 x 160 x 70.7 mm3. We also found that our previously published correction technique reduced the largest error to 0.3 pixels (mu = 0.02 and sigma = 0.07 pixels).