SU-F-J-168: Dosimetric Impact of MR Image Distortion in Treatment Planning

Authors


Abstract

Purpose:

To evaluate the dosimetric impact of MR image distortion in treatment planning.

Methods:

Based on the realistic MRI distortion measurement data from an integrated MRI-Linac system, we created a generalized distortion model and applied it to CT images to simulate distorted MRI images. Ten patients with IMRT plans generated in Pinnacle treatment planning system in 5 anatomical cancer sites (lung, prostate, spinal cord, liver, and pancreas) were retrospectively selected for the simulation. The distortion model was applied to the planning CT image and contours. A distorted plan was created in Pinnacle for each patient. The dose-volume-histogram (DVH) of targets and critical structures were compared between the original and distorted plans. For a fair comparison, the same field sizes and intensity-modulated radiation therapy (IMRT) parameters were used in generating the distorted plan.

Results:

The DVH of organs at risk (OARs) within a distance of 5cm to the imaging center did not show significant difference between the original and distorted plans. For certain OARs such as the spinal cord, the seminal vesicle (SV), and the larynx, with a distance greater than 5cm to the imaging center, larger deviations were observed between the distorted and original plans. The maximum mean dose difference was 1.92Gy (proximal SV, prescribed dose is 40Gy). Similar target coverage was observed between the distorted plans and the original plans for the 10 patients. One lung case had the maximum difference of the target mean dose between the original and distorted plans of 1.63Gy (prescribed dose is 72Gy).

Conclusion:

The MRI distortion did not lead to significant dosimetric difference compared to the original plans in most situations. However, OARs farther than 5cm from the imaging center showed larger dosimetric difference due to the distortion. Safety margin may need to account for the distortion in treatment planning.

One of the authors receive cooperate funding from the Elekta.

Ancillary