Quantification of the geometric accuracy of MRI in tissue: a new approach using MRI itself

The authors have developed a new way to quantify the extent of the geometric distortion due to magnetic field inhomogeneities in MR images to subpixel accuracy and to test the effectiveness of any method used to correct this distortion. This method does not require an external standard such as CT nor does it require that the geometry of the imaging objects be known a priori. Therefore, it can be used to test the effectiveness of MR spatial distortion correction methods in tissue. The authors have shown that the geometric accuracy along the phase-encoding directions of MR images is uncorrupted. Thus measurements along the phase-encoding directions are used to quantify the accuracy of the measurements along the frequency encoding directions. The authors have quantified the distortion in a tissue phantom and found the largest error to be approximately 2.8 pixels (1.8 mm) for B₀=1.5 T, G=3.13 mT/m and FOV=160×160×70.7 mm³. The authors also found that their previously published correction technique reduced the largest error to 0.3 pixels (μ=0.02 and σ=0.07 pixels)