Investigation of a 3D system distortion correction method for MR images

Interest has been growing in recent years in the development of radiation treatment planning (RTP) techniques based solely on Magnetic Resonance (MR) images. However, it is recognized that MR images suffer from scanner-related and object-induced distortions that may lead to an incorrect placement of anatomical structures. This subsequently may result in a reduced accuracy in delivering treatment dose fractions in RTP. To accomplish the precise representation of anatomical targets required by RTP, distortions must be mapped and the images rectified before being used in the clinical process. In this work, we investigate a novel, phantom-based method that determines and corrects for 3D system-related distortions. The algorithm consists of two key components: an adaptive control point identification and registration tool and an iterative method that finds the best estimate of 3D distortion. It was found that the 3D distortions were successfully mapped to within the voxel resolution of the raw data for a 260 x 260 x 240 mm3 volume.