(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: To evaluate the clinical adequacy of region-of-interest (ROI) auto-segmentation by evaluation of probabilistic dosimetric metrics and probabilistic biological endpoints for VMAT prostate patient radiation treatment plans created in different scenarios.
Methods: Physician drawn and auto-segmented (Velocity, Varian Medical Systems) ROIs were created for 10 prostate patients. Pinnacle auto-planning module were used to create treatment plans (78 Gy/39 fraction; two dynamic arc with 45 and 135 degree collimator angles) in three contour scenarios; P_MD using manually delineated (MD) ROIs, P_AS using auto-segmented ROIs, and P_AM using auto-segmented organs-at-risk (OAR)s and the manually delineated target. Assuming published models for patient setup uncertainty, plan robustness analysis was performed by our in-house radiation treatment robustness analyzer. The coverage probability (CP) of important dose-volume metrics, expectation values for normal tissue complication probability (NTCP), and tumor control probability (TCP) metrics for all possible combination of ROI types and planning scenarios were reported and compared with base-line clinical scenario.
Results: Difference in CP of the femoral heads was zero in all cases except one case in which ΔCP was less than 0.03. The observed variation in ΔCP was the most in CTV and rectum dosimetric objectives. CP of D_{95} prostate, D_{15} rectum, and D_{15} bladder sustained heavy losses. While maximum loss ΔE(NTCP) was observed in scenario (MD,P_{AS}) in bladder, maximum loss in ΔE(TCP) was 0.032. Among evaluation scenarios, in overall, (MD,P_{AM}) and (MD,P_{MD}) were the closest.
Conclusion: In the presence of daily setup uncertainty, the statistical analysis of the important biological and dosimetic indices underline the demand for increased accuracy in auto-segmented ROIs of abdominopelvic cavity, in particular prostate and rectum.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
