Click here to

Session: Advances in MRI Image Formation [Return to Session]

An Open-Source Solution to Analyze and Correct MRI Distortion

P Liu1,2 *, S Shan1,2, D Waddington1,2, B Dong2, F Liu3, M Li3, B Whelan1,2, (1) University of Sydney, NSW, AU, (2)Ingham Institute For Applied Medical Research, (3) University of Queensland, Brisbane Qld, ,AU,


WE-A-201-1 (Wednesday, 7/13/2022) 7:30 AM - 8:30 AM [Eastern Time (GMT-4)]

Room 201

Purpose: The use of MRI in radiotherapy requires a higher degree of geometric precision than in diagnostic imaging. Geometric distortion, primarily caused by non-linear gradient fields, is a major impediment to the adoption of MRI in radiotherapy. We present an open-source library that provides the tools to design and construct a distortion phantom, characterize geometric distortion and correct for it in an end-to-end workflow.

Methods: A distortion phantom was designed and constructed from multiple foam slices. The phantom consisted of 324 oil capsule fiducial markers arranged around a 150 mm sphere. The phantom was imaged on an experimental 1T MRI-Linac and clinical CT scanner and, images were run through the automated distortion correction workflow. Markers were extracted from both image sets and matched using an adaptive search algorithm. From this data, the gradient field at each marker position was calculated and 8th order spherical harmonics were fit to characterize the gradient field. Spherical harmonics can be used to reconstruct distortion at any point within the DSV and as input for distortion correction algorithms.

Results: The distortion phantom was low-cost (~$100 USD), lightweight (<1 kg) and easy to construct in-house. All markers were clearly visible on both MR and CT and were accurately segmented and matched with automatic algorithms. The mean geometric distortion of the MRI-Linac was characterized to be 7.5±3.1 mm, with distortion of the markers ranging from 0.7 to 14.9mm. Note that these values do not include any distortion correction. The spherical harmonics were fit three experimental data to within 2% for all three gradient fields.

Conclusion: This open-source library offers a low-cost alternative to commercial distortion correction phantoms and provides a customizable and transparent workflow for the correction of geometric distortion in MRI.

Funding Support, Disclosures, and Conflict of Interest: The authors acknowledge funding from NHMRC and CINSW


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email