ESTRO 2024 - Abstract Book

S5588

RTT - Patient care, preparation, immobilisation and IGRT verification protocols

ESTRO 2024

of freedom as a starting point. These registrations were then evaluated and adjusted based on the contours or MR by the RTT/RO. The contour to CBCT match does not employ any information from the voxel values of the sCT. The MR to CBCT match was taken as the ground truth, as it is the only registration based on the original acquired images, in comparison to other methods, which require extrapolated imaging data. The study was performed in accordance to the cantonal ethics authorization (BASEC 2023-01508).

Results:

This analysis included 36 fractions of patients treated using an MR-only workflow. For each fraction, an MR-CBCT, sCT CBCT, and contour-CBCT match was completed. The average translational differences between the contour-CBCT and MR-CBCT registrations were found to be less than 1mm (0.52±0.79mm, -0.77±2.03mm, 0.26±1.68mm in the x-, y-, and z-direction respectively), and average rotational differences were below 0.5° (0.37±1.77°, 0.16±0.76°, -0.28±0.49° in the x-, y-, and z-direction respectively). The distribution of the deviations are reported in Figure 1. Statistical testing (TOST) demonstrated no significant deviations (p<0.05) within a 2mm/2° interval. However, when using stricter criteria of 1mm/1°, p-values of up to 0.25 were observed, indicating that the equivalence between contour-CBCT and MR CBCT should be rejected if the required accuracy is below 1mm/1°. It should be noted that this result is compatible with the resolution of the contours (2mm) and the sCT (1mm), while the direct MR-CBCT match can exploit the higher resolutions of MR (down to 0.3mm in-plane) and CBCT (down to 0.5mm in-plane).

Conclusion:

The contour to CBCT registration technique demonstrated good localization accuracy when compared to an MR-CBCT match, with average translational and rotational values below 1mm and 0.5°. These preliminary results are encouraging, and further research should be done to validate this image registration technique as a reliable method for the matching of pelvic irradiation patients within an MR-only workflow. It could also be a useful method in the quality control of standard sCT to CBCT matches, as it is comparable to matching directly on the original planning MR scan.