SU-F-303-04: Characterization and Correction of Gradient Non-Linearity Induced Distortion On a 1.0 Tesla Open Bore MR-SIM

Authors


Abstract

Purpose:

Implementation of MRI for single-modality simulation is limited by system-level distortions that may degrade treatment planning accuracy. We performed technical characterization of gradient non-linearity (GNL) for large field-of-views (FOVs), developed a correction scheme, and quantified temporal stability.

Methods:

A 36×43×2cm3 phantom with 255 known landmarks was scanned at isocenter in transverse, sagittal, and coronal planes (T1W-FFE, TE/TR/α=5.54ms/30ms/28°, voxel size=1×1×2mm3, BW=191Hz/Pix, FOV=450×450×26mm3)using 1.0T open MR-SIM with vertical magnetic field. Two scans were obtained in each orientation (positive/negative read gradient polarity of ±4.48mT/m) to isolate GNL using reverse gradient techniques. Four off-axis positions (-125 to 125mm) in superior-inferior direction were evaluated. An automated MATLAB program was developed to identify control-point locations relative to phantom schematics using masking, thresholding, and connectivity analysis. Distortion maps were interpolated between control points using singular value decomposition to fit to a sixth-degree polynomial. In addition to vendor corrections, images were corrected via warping with inverse distortion maps. Temporal stability was assessed over 6 months.

Results:

Over the phantom FOV, mean±stdev (P5/P95) in mm were −0.07±1.10 (−1.83/1.92) in x-direction and 0.10±1.10 (−1.5/2.15) in y-direction for the transverse plane. Similar results were obtained in the sagittal plane. For coronal images, distortions were 0.4±1.16 (−1.32/2.50) and 0.04±0.40 (−0.52/0.77) in x- and y-directions, respectively. Although negligible near isocenter (<1mm pixel width), 35% of pixels had distortions >1mm and 7% >2mm. Distortions were more pronounced 12.5cm off-axis, with 44% of voxels >1mm, 17% >2mm, and 5% >3mm with largest distortions (∼8mm) occurring 22cm radial distance from isocenter. In-plane distortions were reduced to <1mm up to ∼22.5cm from isocenter post-correction. Distortion maps remained stable over 6 months with distributions varying <0.5mm between measurements.

Conclusion:

GNL distortion was non-negligible. Further corrections were necessary to maintain <1mm accuracy for large FOVs. Statistical analysis of temporal stability revealed GNL distortion was consistent over 6 months.

Research supported in part by a grant from Philips HealthCare (Best, Netherlands).

Ancillary