413 poster REDUCING TRIAL-AND-ERROR WITH PROBABILISTIC PLANNING FOR PROSTATE CANCER R. Bohoslavsky1 , M. Witte1 , T. Janssen1 , M. van Herk1
IMRT
1
T HE N ETHERLANDS C ANCER I NSTITUTE - A NTONI VAN L EEUWENHOEK H OSPITAL, Radiation Oncology, Amsterdam, Netherlands
Purpose: Probabilistic planning allows treatment optimization and evaluation without using explicit safety margins. This added flexibility generally leads to better compromises when PTV and OARs overlap. We are currently developing a practical planning system for future clinical use. Because probabilistic planning has more freedom to optimize, we expect that less user intervention will be needed to obtain a good plan. Our aim was therefore to study the potential reduction in trial-and-error steps during plan optimization for prostate cancer. Materials: Probabilistic optimization was implemented by extending the clinical cost functions of the Pinnacle TPS research version to include the effect of translational and rotational geometrical uncertainties. Random errors are incorporated by blurring the dose distribution. Systematic errors are included by rotating and shifting the CTV with respect to the blurred dose. The same geometrical uncertainties are applied in our in-house developed evaluation tool to analyze the resulting plans.In clinical practice, treatment plans are first optimized using a standard set of objectives (default plan) which are subsequently manually modified until a clinically acceptable plan is achieved (clinical plan). The actual number of additional optimizations depends on the planner’s expertise and the characteristics of the patient. In this study clinical plans were re-planned applying a set of probabilistic objectives (probabilistic plan). Comparing these three plans allows estimation of the potential reduction in trial-and-error steps needed to obtain an acceptable plan. Results: SIB treatment plans from 41 prostate cancer patients were evaluated. The typical optimization time for probabilistic and standard plans was 10 and 2 minutes respectively. The probability of delivering at least 95% of the original prescribed dose (78Gy) to 99% of the target (prostate gland and seminal vesicles) versus the rectal wall volume receiving more than 65 Gy was analyzed using our evaluation tool (PD95V99,V65). The median results were compared, keeping the original clinical plans as reference [PD95V99=87%, V65=17%]. After the first optimization, planning with probabilistic objectives lead to better plans [PD95V99=87%, V65=16%] than standard planning [PD95V99=72%, V65=16%], obtaining higher target coverage with similar rectal wall dosage (Fig.1). Conclusions: Probability based treatment planning directly includes geometrical uncertainties during optimization without using explicit safety margins. Plans obtained after one optimization run achieve acceptable dose-volume metrics more often than those using standard objectives. This indicates a potential reduction in trial-and-error steps, i.e. even though probabilistic planning is computationally more demanding, the total planning time is likely to reduce.
Figure 1. Mean DVHs produced by the Monte Carlo evaluation tool for normal and worst case scenarios. ArmA corresponds to the uniform dose escalation to the target, armB corresponds to the dose boosting of the high FDG uptake region. Conclusions: Dose escalation techniques may benefit from probabilistic planning. Increase in mean target dose is achieved both for conventional dose escalation and dose painting techniques. For equal values of systematic and random errors probabilistic plans deliver higher dose to the target compared to the standard margin based approaches. Probabilistic planning might be a useful tool designing treatment plans for dose-escalation of lung tumors.