Towards magnetic resonance imaging guided radiation therapy (MRIgRT)

The goal of this work is to address key aspects of the magnetic resonance imaging guided radiation therapy (MRIgRT) process of cancer sites. MRIgRT is implemented by using a system comprised of a magnetic resonance imaging (MRI) scanner coupled with a radiation source, in our case a radiotherapy accelerator (Linac). The potential benefits of MRIgRT are the real-time tracking of the tumor and neighbouring healthy anatomy during treatment irradiation leading to on-line treatment plan optimization. Ultimately, this results in an increased accuracy and efficiency of the overall treatment process. A large research effort is conducted at Cross Cancer Institute to develop a hybrid MRI-Linac system consisting of a bi-planar 0.2 T permanent magnet coupled with a 6 MV Linac. The present work is part of this project and aims to address the following key components: (a) magnetic shielding and dosimetric effects of the MRI-Linac system, (b) measure and correction of scanner-related MR image distortions, and (c) MRI-based treatment planning procedure for intracranial lesions. The first two components are essential for the optimal construction and operation of the MRI-Linac system while the third one represents a direct application of the system. The linac passive shielding was achieved by (a) adding two 10 cm thick steel (1020) plates placed at a distance of 10 cm from the structure on opposite sides of the magnet; and (b) a box lined with a 1 mm MuMetal(TM) wall surrounding the Linac. For our proposed MRI-Linac configuration (i.e. 0.2 T field and rotating bi-planar geometry) the maximum dose difference from zero magnetic field case was found to be within 6% and 12% in a water and water-lung-water phantom, respectively. We developed an image system distortion correction method for MRI that relies on adaptive thresholding and an iterative algorithm to determine the 3D distortion field. Applying this technique the residual image distortions were reduced to within the voxel resolution of the raw imaging data. We investigated a procedure for the MRI Simulation of brain lesions which consists of (a) correction of MR images for 3D distortions, (b) automatic segmentation of head sub-structures (i.e. scalp, bone, and brain) relevant for dosimetric calculations, (c) conversion of MRI datasets into CT-like images by assigning bulk CT values to head sub-structures and MRI-based dose calculations, and (d) RT plan evaluation based on isodose distributions, dosimetric parameters, dose volume histograms, and an RT ranking tool. The proposed MRI-based treatment planning procedure performed similarly to the standard clinical technique, which relies on both CT and MR imaging modalities, and is suitable for the radiotherapy of brain cancer.