ESTRO 2021 Abstract Book

S58

ESTRO 2021

Results All dwells could be identified and the dwell time was measured with an accuracy better than 0.1s. The high accuracy in dwell time determination allowed for finding errors in dwell time and verify the number of dwell positions. The dwell time measurement was not affected by the position of the panel, whilst the dwell position reconstruction shows better accuracy in set ups where the source was closer to the surface of the phantom. Determining which dwell was skipped proved difficult when the panel was located on the back of the head, because the interdwell distance (0.2 cm) was smaller than the maximum deviation of the source location reconstruction. However, with the panel positioned on the side of the head, in which the source is closer to the phantom surface, this was possible. In all cases IrIS could determine swapped catheters. Conclusion IrIS can track the source accurately during HDR BT treatments in a realistic phantom. Position and dwell time determination allows for detection of several common treatment errors, as wrong dwell times, skipped dwells and swapped catheters. With known source positions and dwell times it is possible to recalculate the dose that was deposited. Further research on phantoms with multiple materials and non-static phantoms are the next step to validate the use of IrIS for IVD for HDR BT. OC-0085 Development and quantitative evaluation of AI-based pelvic MRI autocontouring for adaptive MRgRT M. Nachbar 1 , M. Lo Russo 2 , S. Boeke 2 , D. Wegener 2 , J. Boldt 2 , S. Butzer 2 , T. Roque 3 , A. Lombard 3 , L. De Vitry 3 , N. Paragios 3 , D. Zips 2 , D. Thorwarth 1 1 University Hospital and Medical Faculty. Eberhard Karls University Tübingen, Section for Biomedical Physics. Department of Radiation Oncology, Tübingen, Germany; 2 University Hospital and Medical Faculty. Eberhard Karls University Tübingen, Department of Radiation Oncology, Tübingen, Germany; 3 TheraPanacea, TheraPanacea, Paris, France Purpose or Objective In MR-guided radiotherapy (MRgRT) online plan adaptation can account for tumor shrinkage, interfraction motion and allows thus daily sparing of relevant OARs close to the target. Due to the high interfraction variability of bladder and rectum, patients with tumors in the pelvic region may strongly benefit from adaptive MRgRT. However, manual annotation of relevant structures limits the potential of online adaptive MRgRT, as time spent for MRI contouring results in intrafraction motion. Therefore a deep neural network was generated on MRIs of the 1.5 T MR-Linac and evaluated for the implementation in an online adaptive workflow. Materials and Methods From 47 patients, T2w MRI data acquired on a 1.5 T MR-Linac (Unity, Elekta) at five treatment fractions were contoured, identifying prostate, seminal vesicles, rectum, anal canal, bladder, penile bulb, and bony Proffered papers: Proffered Papers 6: Adaptive RT: MR-guided and SBRT