ESTRO 37 Abstract book

ESTRO 37

S480

We study spatiotemporal fractionation for liver SBRT, where the main clinical motivation is to improve the ratio of tumor dose to mean dose in the noninvolved liver. 5 patients have been selected that represent different tumor locations and sizes. Fractionation effects are modeled via the biologically effective dose (BED) model. Treatment planning for spatiotemporal fractionation is performed by simultaneously optimizing multiple distinct treatment plans for different fractions based on their cumulative BED. Results Figure 1 illustrates a spatiotemporal treatment plan for a 5-fraction treatment of a liver lesion that is centrally located in the right lobe. Each fraction delivers a high single-fraction dose exceeding 20 Gy to complementary parts of the target volume. Figure 1f shows that all 5 fractions combined deliver the prescribed BED equivalent of 50 Gy in 5 fractions to the GTV (red contour). At the same time, each fraction delivers a similar dose bath to the surrounding normal liver, i.e. exploits the fractionation effect. Figure 2 compares the DVH evaluated for BED of the spatiotemporal plan to a uniformly fractionated 5-fraction reference plan. It demonstrates that spatiotemporal fractionation achieves a net reduction in liver BED for the same target coverage. On average, spatiotemporal fractionation was found to achieve a mean liver BED reduction of 10-20%. Patients with large lesions centrally located in the liver show a larger benefit than patients with lesions adjacent to bowel or stomach, where these additional dose-limiting normal tissues require fractionation. Conclusion Delivering distinct dose distributions in different fractions, purely motivated by fractionation effects rather than geometric changes, may improve the therapeutic ratio. This challenges the decades-old paradigm in which the same dose is delivered in each fraction. Liver SBRT represents a promising application of this concept as it may allow for dose escalation for patients in whom the prescription dose is limited by the mean liver dose.

PO-0901 Testing the RapidPlan performance with multiple TPS and multiple H&N models A. Scaggion 1 , G. Loi 2 , M. Fusella 1 , L. Vigna 2 , M. Paiusco 1 1 Veneto Institute of Oncology IOV-IRCCS, Medical Physics Department, Padova, Italy 2 Hospital Maggiore della Carità, Medical Physics Department, Novara, Italy Purpose or Objective To test the performance of a Knowledge Based Planning (KBP) algorithm (RapidPlan, Varian Medical System, Palo Alto, CA) for oropharynx treatments across multiple TPS (inter-system validation), using multiple DVH estimation models. Material and Methods Two samples of oropharyngeal cancer patients treated with VMAT were collected from two different institutions. Sample #1 was composed by 50 patients planned with Eclipse TPS v11 at Institute #1; Sample#2 was composed by 30 patients planned with Pinnacle v9.10 at Institute#2. Two RapidPlan models, Model#1 and Model#2, were trained respectively using the two samples of patients. A third RapidPlan model (Model#3) was trained merging the two samples. The RapidPlan predictions from these models were used to drive the optimization of ten further patients collected evenly from the two institutions. The optimization process was undertaken with three different TPS: Eclipse v11 and RayStation v5 at Institute #1 and Pinnacle v9.10 at Institute#2. No manual touch-up was performed for all the plans, leading to an automatic optimization process in order to avoid any bias related to user planning experience. The outcomes of the planning were compared on the basis of several DVH endpoints and also the Plan Quality Metric formalism (based on D. Gintz, J. Appl. Clin. Med. Phys. 2016). Results TPS RapidPlan Model PQM% score Pinnacle Model #3 74.7±7.9% Eclipse Model #2 74.5±.9.% RayStation Model #2 74.4±8.3% Pinnacle Model #2 73.7±5.3% RayStation Model #3 72.5±9.3% Pinnacle Model #1 72.4±6.7% RayStation Model #1 72.2±10.3% Eclipse Model #3 71.2±11.3%