Characterizing geometrical distortions in MRI for radiotherapy: Evaluating MR-SIM protocols and dose accuracy in stereotactic radiosurgery using AAPM TG-284 criteria

Abstract

Objective

Magnetic resonance imaging (MRI) is crucial for diagnostic imaging and radiotherapy (RT) planning due to its superior soft tissue contrast. However, geometric distortions can affect treatment accuracy. This study evaluates the geometric accuracy of MRI protocols using a 1.5T MR-Sim scanner for RT and their dosimetric impacts.

Materials and Methods

Geometric distortion was assessed using the CIRS 604-GS MR Image Distortion Phantom with 2152 control points across various MRI sequences. A central cubic structure and surrounding regions were analyzed, totaling 27 structures. Rigid registration in Eclipse V18.1 aligned MRI images to CT images, with verified accuracy and visual inspection. An evaluation of the overlap and surface distance metrics was performed through MIM Mastro. Geometric distortion was quantified using 3D distortion analysis software, comparing marker coordinates to reference CT scans. Twenty-seven VMAT full arc plans were done in Eclipse on 27 paired structures, irradiated with 16 Gy in a single fraction at 6 MV energy using the Acuros XB algorithm

Results

All 3D sequences demonstrated mean distortions of 0.4–0.5 mm, with maximum distortions of up to 3.3 mm in the TOF (time of flight) Angiography sequence. In contrast, the T2 TSE 2D sequence showed larger distortions (mean 1.4 mm, max. 9.7 mm). The center structure showed stable dosimetric performance (D_mean: 1601.6 cGy for CT, 1600.3 cGy for MR). Peripheral regions showed higher variability, with MR_normalized D_95% values ranging from 83.9% to 88.1%.

Conclusion

Precise MR protocols are essential for accurate tumor delineation and RT planning. While all sequences showed acceptable accuracy, 3D sequences are superior for high-precision RT.