Commissioning of a 4D MRI phantom for use in MR-guided radiotherapy.

Abstract

Systems for integrated magnetic resonance guided radiation therapy (MRgRT) provide real-time and online MRI guidance for unequaled targeting performance of moving tumors and organs at risk. The clinical introduction of such systems requires dedicated methods for commissioning and routine machine quality assurance (QA). The aim of the study was to develop a commissioning protocol and method for automatic quantification of target motion and geometric accuracy using a 4D MRI motion phantom. The commissioning was performed on a clinically used 3T MR scanner. The phantom was positioned on a flat tabletop overlay using an in-house constructed base plate for a quick and reproducible setup. The torso-shaped phantom body, which was filled with mineral oil as signal generating medium, included a 3D grid structure for image distortion analysis and a cylindrical thru-hole in which a software-controlled moving rod with a hypo-intense background gel and a decentralized hyper-intense target simulated 3D organ motion patterns. To allow for sequence optimization, MR relaxometry was performed to determine the longitudinal T1 and transverse T2 relaxation times of both target and background gel in the movable cylinder. The geometric image distortion was determined as the mean and maximum 3D Euclidean distance (Δmean , Δmax ) of grid points determined by nonrigid registration of a 3D spoiled gradient echo MRI scan and a CT scan. Sinusoidal 1D/2D/3D motion trajectories, varying in amplitude and frequency, as well as an exemplary 1D MR navigator diaphragm motion pattern extracted from a healthy volunteer scan, were scanned by means of 2D cine MRI and 4D MRI. Target positions were automatically extracted from 2D cine MRI using an in-house developed software tool. The base plate enabled a reproducible setup with a deviation of <1mm in all directions. Relaxometry yielded T1 /T2 values for target and background gel of 208.1±2.8/30.5±4.7ms and 871±36/13.4 ± 1.3ms, respectively. The 3D geometric image distortion increased with distance from the magnetic isocenter, with Δmean =0.58±0.30mm and Δmax =1.31mm. The frequencies of the reconstructed motion patterns agreed with the preset values within 0.5%, whereas the reconstructed amplitudes showed a maximum deviation to the preset amplitudes of <0.5mm in AP/LR direction and <0.3mm in IS direction. A method and protocol for commissioning of a 4D MRI motion phantom on a 3T MR scanner for MRgRT was developed. High-contrast and geometrically reliable 2D cine MR images of the phantom's moving target could be obtained. The preset motion parameters could be extracted with sufficient spatio-temporal accuracy from 2D cine MRI in all motion directions. The overall 3D geometric image distortion of <1.31mm within the phantom grid confirms geometric accuracy of the clinically utilized 3D spoiled gradient echo sequence. The method developed can be used for routine QA tests of spatio-temporally resolved MRI data in MRgRT.