Room: AAPM ePoster Library
Purpose:
To validate two distortion correction techniques for diffusion-weighted single-shot echo-planar imaging (DW-ssEPI) on a low-field MR-guided radiotherapy (MRgRT) system.
Methods:
The effects of sequence optimization using parallel imaging (PI) and Appropriate bandwidth selection on spatial distortion was first evaluated on the 0.35T MRgRT system using a spatial integrity phantom. Field map correction (FM) and reversed gradient (RG) correction were then performed on the optimized protocol to further reduce the distortion[1,2]. An open-source toolbox was used to quantify the spatial displacement before and after distortion correction[3].
Five patients were recruited to assess the in-vivo performance. The optimal distortion correction technique was identified by visual assessment. To quantify distortion reduction, body contours were drawn on the TSE image (reference image), the DW-ssEPI image, and the distortion corrected DW-ssEPI images, respectively. Mean distance to agreement (MDA), DICE coefficient, and Jaccard index between contours on the reference image and the test image were calculated.
Results:
Sequence optimization plays a vital role in improving the spatial integrity, and spatial distortion is proportional to the total readout time. Before the correction, distortion of the optimized protocol (PI and high bandwidth) was 1.50±0.89mm in 100mm radius and 2.21±1.39mm in 175mm radius. FM correction reduced the distortion to 0.42±0.27mm and 0.67±0.49mm respectively, and RG reduced to 0.40±0.22mm and 0.64±0.47mm respectively.
For the patient study, the field map correction was unable to handle chemical shift artifact whereas the reversed gradient method successfully mitigated the chemical shift. MDA reduced from 2.41±0.70 to 0.70±0.08 after the RG correction. DICE coefficient increased from 0.96±0.01 to 0.99±0.00, and Jaccard index increased from 0.92±0.02 to 0.97±0.00.
Conclusion:
Two distortion correction techniques were validated on a low-field MRgRT system. Overall, the reversed gradient correction was able to reduce the mean distortion to less than 1mm on both phantom and in-vivo.
Funding Support, Disclosures, and Conflict of Interest: The authors would like to acknowledge the research support from ViewRay Inc.