ESTRO 2024 - Abstract Book

S3145

Physics - Autosegmentation

ESTRO 2024

bulb and urethra on the T2-weighted image (4). Manual contours were also created by trained clinicians on the T2 weighted MR.

Two treatment plans were created for each patient: a MR-only plan and a MR-CT plan. The MR-only plan was optimised using a manual MR-based prostate and seminal vesicles contour as the GTV, automatic MR OAR contours and the sCT for dose calculations (MR-only plan[auto,sCT]). The MR-CT plan was optimised using the same GTV, manual MR OAR contours and the deformably registered planning CT for dose calculations (MR-CT plan[man,CT]). The dose from the MR-only plan was also recalculated on the deformably registered CT to the manual OAR contours (MR-only plan[man,CT]). Differences in doses at clinically relevant DVH constraints between the MR-only plan[auto,sCT] and MR-CT plan[man,CT] were determined to assess how accurate the reported values from the MR-only pathway would be. In addition, differences between doses to the automatic contours on the sCT and manual contours on the CT from the MR-only plan were calculated (MR-only plan[auto,sCT] minus MR-only plan[man,CT]). This measured the dose difference due to inaccuracies in the automatic contours and sCT compared to manual contours and CT for the same plan.

Results:

Automatic contours and sCTs were generated for all patients (see figure 1 for example image). Mean reported DVH differences between the MR-only and MR-CT pathways were within 1.0 Gy / 1 percentage point for all organs except the bowel bag and the D5% femoral head doses (table 1). This is within the 1 Gy dose difference reported in the literature from manual contour inter-observer variability(7). Mean measured DVH differences from the MR-only plan between the automatic contours on sCT and manual contours on CT were also within 1.0 Gy / 1 percentage point for all organs except bowel bag and D5% femoral head doses (table 1).