ESTRO 2020 Abstract book

S830 ESTRO 2020

Sixteen complexity metrics have been have been identified as relevant: fifteen from literature (PIMV, AI, MCS, MI…) and one novel (wavelet transform of the fluence map). They can be classified into two categories: the metrics based on geometrical aspects (shapes of the fields, opening of the leaves), and based on fluence. Their calculation use dicom RT files generates by TPS (treatment planning systems). We have used machine learning techniques to create a model linking the Patient QA results and to complexity metrics. Two types of models have been created (pass/fail prediction, based on nonparametric bayesian method) and gamma-index features (passing rate, mean gamma, max gamma, based on regression). Prediction uncertainty has been implemented with a prediction model assessment using the leave-one-out method LOOCV (logarithmic, zero/one and R² scores) and the leave-pair-out method LPOCV (ROC curves and AUC). We have built a prediction model, at APHM – La Timone, France, based on the following equipment ( Elekta Synergy, Beam Modulator, Delta4, Pinnacle3, and different tumor sites (pelvis, prostate, H&N, thorax, brain, …)). 445 VMAT plans with 615 arcs have been used in the learning phase. The clinical gamma criteria used was 3% local dose/ 3mm/ treshold 20%. For the test phase, 92 plans with 146 arcs have been used. Results ROC Analysis has been performed on the learning database (Area Under The Curve = 0,85, Zero-one = 0,89), that is linked with the specific equipment used ( Elekta Synergy, Beam Modulator, Delta4, Pinnacle3, and different tumor sites (pelvis, prostate, H&N, thorax, brain, …)), and the specific acceptance criteria used. (3% local dose/ 3mm/ treshold 20%).

Conclusion We have shown that with this deep learning method, we could reduce the number of pre treatment patient QA by 45% (True OK/KO) and have a good efficiency of the global prediction of 89% (True OK/KO/TEST). The 11% left being good plans that will be tested. Further investigations in other radiation therapy centers are ongoing, and the Gamma values will also being studied, in order to give additional analysis tools to reduce the time spent for pre treatment patient QA. PO-1461 Benchmarking proton therapy water equivalent path length calculations against TPS algorithms K. Busch 1 , A.G. Andersen 1 , J.B.B. Petersen 1 , P.S. Skyt 1 , O. Nørrevang 2 , L.P. Muren 1 1 Aarhus University Hospital, Department of Oncology, Aarhus, Denmark ; 2 Aarhus University Hospital, Danish Centre of Particle Therapy, Aarhus, Denmark Purpose or Objective Proton therapy is sensitive towards inter-fractional organ motion and density variations due to the finite range of protons, potentially causing dose degradations. Range variations can be estimated by calculating the water equivalent path length (WEPL), which can be done quicker than full-fledge dose calculations using treatment planning system (TPS) algorithms. WEPL calculations have the potential to become a useful tool for applications involving online calculations with respect to plan robustness towards organ motion and density changes. Our aim was to compare WEPL calculations with dose re-calculation using a TPS algorithm. Material and Methods The WEPL-based isodose levels were calculated using two image modalities e.g. a planning CT (pCT) and a repeat CT (rCT), which were translated to stopping power values. A field isodose was extracted from the plan as physical points in the pCT space. We assumed that the pCT field configuration had the same physical positions in rCT space. From each point in the isodose, WEPL was calculated as the sum of linearly interpolated values, at each step along the opposite beam direction, until the edge of the image. The in-house WEPL program calculated isodose levels of 15%, 25%, 35%, 45% and 55% of the target dose, which were compared using Dice similarity coefficient to isodose levels created using the CERR platform in Matlab (Figure 1c/d). For the treatment planning calculations, a cube of size 20cmx20cmx20cm with the Hounsfield units (HUs) of water was created in Eclipse TPS with a cylindrical target in the middle. Two multi-field optimised proton plans were created with a target dose of 78 Gy. Plan 1 had fields at zero and 270 degrees and Plan 2 had fields at 270 and 325 degrees. The plans were recalculated on the same cube, where one and/or two inserts were created with HUs of either bone (300 HU), air (-1000 HU) or water (0 HU) in front of each field (Figure 1a/b). In total, each plan was

(Figure 1)

This model has been used for the test patient. Figure 2 shows the results of good prediction and time sparing for a given confidence level, that can be adjusted. The figures are calculated with a False OK rate >5% and a False KO rate > 15% and 95% of the probability distribution fulfilling the 2 criteria. If not, the advise would be to do the QA.

(Figure 2)