Assessment of gradient nonlinearity distortions in a low-field mrlinac using a modular large field of view phantom
Nejad-Davarani SP, Barberi E, Cole M, and Glide-Hurst C. Assessment of gradient nonlinearity distortions in a low-field mrlinac using a modular large field of view phantom. Med Phys 2018; 45(6):e481.
Purpose: MR-guided radiation therapy offers unparalleled soft tissue contrast for real-time, non-ionizing target tracking and localization. However, MRI is prone to distortions, with the dominant source arising from nonlinearities in the spatial-encoding gradients, particularly for low field strengths. This work characterizes gradient nonlinearity (GNL) for a low-field MR-linac using a modular large field-of-view (FOV) phantom to assess the necessity for GNL corrections for MR-IGRT. Methods: The phantom configuration was adjusted to 60×48×56 cm3 to fit the bore geometry including >6000 spherical landmarks. The phantom was scanned with forward/reverse read polarities using a Gradient Echo sequence (FA/TR/TE = 28, 30 ms, 6 ms). A CT reference dataset was acquired and MR images were rigidly registered using central landmarks. After background removal, landmarks were segmented using a connectivity algorithm, and centroids were calculated. Distortion was measured in Anterior-Posterior (AP) and Left-Right (LR) frequencyencoding directions based on landmark deviations from CT. By averaging landmark centroid deviations between MR images and the CT image, GNL can be isolated from B0 field and susceptibility. Our approach was benchmarked against a commercially available phantom analyzed via spherical harmonics. Results: Within 10 cm radial distance from magnet isocenter, distortions ranged from [-0.69, 0.18] mm and [-0.77, 0.23] mm in the LR and AP directions, respectively. In the LR and AP images, respectively 45% and 55% of landmarks >20 cm away shifted by >1 mm. At 10-20 cm from isocenter, GNL distortions in the AP were -0.07±0.53 mm (P5-95: -0.92 to 0.76, range: -2.04 to 2.0). Results obtained from spherical harmonics analysis were consistent (-0.04±0.54 mm (P5-95: -0.97 to 0.83, range: -1.9 to 1.7). Conclusion: GNL distortions in an MR-linac were negligible within 11 cm from magnet isocenter which was confirmed in two independent analyses. However, in the periphery, non-negligible distortions were observed, which may necessitate GNL corrections for MR-only planning, particularly away from isocenter.