Gradient Nonlinearity-Corrected 4D MRI for Abdominal Motion Quantification on a 1.5T MR-Linac
Abstract
Purpose
Motion-resolved 4D MRI is essential for assessing abdominal tumor and organ motion, but gradient nonlinearity (GNL) causes geometric distortion that reduces spatial fidelity and motion accuracy. This study evaluates GNL correction for 4D MRI on a 1.5T MR Linac.
Methods
Thirty-two pancreatic cancer patients undergoing five-fraction MR-Linac treatment were included. Free-breathing 3D T1-weighted golden-angle radial stack-of-stars abdominal imaging was performed on an Elekta 1.5T Unity MR-Linac. Motion-resolved 4D MRI with 10 respiratory phases was reconstructed offline using XD-GRASP. GNL correction employed spherical harmonic coefficients to generate 3D displacement vector fields for each motion state. Motion range and displacement of the gross tumor volume (GTV) and two organs at risk (duodenum-stomach, small bowel) were analyzed in seven patients with z-direction imaging offsets >50 mm. Geometric distortion was quantified within body contours, and metrics before and after GNL correction were compared using paired t-tests.
Results
The imaging off-center distance (mean ± SD) was 27.78 ± 18.72 mm, 16.01 ± 11.99 mm, and 10.64 ± 7.30 mm in SI, AP, and LR directions. Displacement within body contours was 2.82 ± 0.75 mm. In seven patients, displacement was 3.76 ± 1.75 mm for the GTV, 1.36 ± 0.67 mm for the duodenum-stomach, and 5.56 ± 2.01 mm for the small bowel. Motion range before GNL correction was 3.37 ± 1.64 mm, 5.93 ± 2.32 mm, and 5.27 ± 2.18 mm, and after correction was 3.37 ± 1.69 mm, 6.07 ± 2.19 mm, and 7.04 ± 1.30 mm for the GTV, duodenum-stomach, and small bowel, respectively. Changes were not significant for the GTV (p = 0.829) or duodenum-stomach (p = 0.145) but were significant for the small bowel (p = 0.048).
Conclusion
GNL correction improves 4D MRI geometric accuracy on the 1.5T MR-Linac, enabling more accurate motion assessment for abdominal cancer radiotherapy.