Gradient Non-Linearity Correction of MR Images for Functional Radiosurgery

This paper discusses the correction of MR images to submillimeter accuracy needed for functional radiosurgery. MR images experience non-linear distortion due to the magnetic field, which becomes more of a problem for newer machines with larger bores and stronger magnetic fields. This paper models the distortion correction parameters using a spherical harmonics basis, which avoids the need to invert the function to correct the image. The coefficients appear linearly for the spherical harmonics so they are solved for in each dimension by least squares techniques for an MR image of a phantom and the measurements of the phantom. Practical considerations in the design are also covered.