ESTRO 2021 Abstract Book

S1580

ESTRO 2021

Figure 2: Paddick conformity index difference (AAA-AXB)

Conclusion The present study found a breakpoint in terms of PTV-ring median density that splits into two ranges the differences between AAA and AXB plans for SBRT lung cancer for 3D technique. The results suggested that patients with ring densities lower than the BP may need special attention. PO-1853 Assessment of the influence of shoulder movement on the Vmat plan quality in Head&Neck treatments S. Hofer 1 , M. Maffei 2 , P. Ferrari 3 , S. Bou Selman 4 , J. Waskiewicz 4 , M. Haller 3 1 Südtiroler Sanitätsbetrieb, Dienst für medizinische Strahlenphysik, Bozen, Italy; 2 Südtiroler Sanitätsbetrieb, Dienst für onkologische Strahlentherapie, Bozen, Italy; 3 Südtiroler Sanitätsbetrieb, Dienst für medizinische Strahlenphysik , Bozen, Italy; 4 Südtiroler Sanitätsbetrieb, Dienst für onkologische Strahlentherapie , Bozen, Italy Purpose or Objective Purpose of the study is to assess the influence of the shoulder movement on the plan quality and to investigate which Vmat technique is the most robust in terms of target coverage and maximum dose. Materials and Methods In the Raystation treatment planning system (v.9B, Raysearch Laboratories) the motion of organs can be simulated by the creation of virtual CTs using deformable registration. In this study the shoulder movement was simulated in craniocaudal and anterior posterior direction with a motion amplitude of 5mm. Plans with four different Vmat techniques for 5 patients where created. Subject of the investigation was the first phase of head and neck treatments with a dose of 54 Gy in 27 fractions. The following VMAT techniques where used: 2 full arcs in the standard Vmat technique (a), 2 full arcs using a sliding window technique (b), 6 arcs sparing the shoulder region with sliding window technique (c) and a technique able to automatically spare the shoulders in the beam entrance during the optimization procedure (d). Results In terms of the percentage of the PTV volume receiving 95% of the prescribed dose ( V95%(%) ) only technique (d) is significantly worse than technique (b) and (c) with the best PTV coverages ( paired bilateral T test – p < 0.05 ) . The average coverage ( V95%(%) ) for technique (a) and (d) is less than 95% and does on average not respect the quality standards required in our department. The dose maximum in terms of D1% for all four techniques was less than the limit value proposed by ICRU Report 50 of 107% of the prescribed dose. To assess better the influence of shoulder movements on the dose to the target only part of the PTV in the range of the cranial vertebrae C6-C7 was evaluated. On the radar chart showing the average V95%(%) coverage of the C6-C7 target volume under shoulder movement (5mm) it is visible that the coverage is influenced most by the shoulder movement in superior direction. In that case only for the 6 arc technique (c) the difference in coverage respect to the coverage of the undeformed plan is not statistical significant. For the other 3 techniques the coverage was significantly worse. In all other three movement directions there was no significant difference in C6-C7 target coverage and in no case of shoulder movement the dose maximum on the target (D1%) exceeded the 107% limit.

Conclusion Only the 6 arcs shoulder sparing technique (c) was robust enough to maintain coverage also in the case of a shoulder movement in superior direction. For all other directions all the techniques showed no significant loss in target coverage. Only technique (a) and (d) in some cases gave a plan quality which was not able to fulfill the criteria requested in the department. However the technique with the best plan quality (c) has with almost 7 minutes also the longest average irradiation time (average irradiation time estimated by the TPS for the different techniques: (a) 195s, (b) 315s, (c) 416s, (d) 328s ) PO-1854 Brachytherapy treatment planning for vaginal cylinder applicator using Deep Learning Neural Network H. Safigholi 1 1 Islamic Azad University, Shiraz Branch, Shiraz, Fars, Iran Islamic Republic of Purpose or Objective To evaluate the plan quality of 192Ir high-dose-rate (HDR) cylinder applicator using a Neural Network Deep Learning (DL) approach for cervix brachytherapy. Materials and Methods The CT DICOM patients data of conventional HDR cervix planning were previously treated using cylinder applicator are used to extract the normal/tumor tissues and cylinder applicator features to train a DL neural network. For 5 new patients, the previous trained DL network is used for automatic reconstruction and delineations of organ at risks (OARs), high-risk clinical target volume (CTV), and applicator. Then for each case, 3D dose distribution matrices of each source inside the applicator were calculated using Monte Carlo simulations and imported into an in-house inverse planning optimization tool to generate optimal plans. Plans are normalized to the CTV D90. Treatment time, accuracy, and dose metrics for CTV/OARs were