ESTRO 2023 - Abstract Book

S1410

Digital Posters

ESTRO 2023

A. Simko 1 , M. Bylund 2 , G. Jönsson 1 , T. Löfstedt 3 , A. Garpebring 1 , T. Nyholm 1 , J. Jonsson 1

1 Umeå University, Department of Radiation Sciences, Umeå, Sweden; 2 Umeå University, Departmen of Radiation Sciences, Umeå, Sweden; 3 Umeå University, Department of Computing Science, Umeå, Sweden Purpose or Objective The use of synthetic CT (sCT) would reduce the costs and scan time while improving the accuracy of organ delineations in the radiotherapy workflow. The performance of deep learning (DL) solutions for sCT generation is steadily increasing, however most proposed methods were trained and validated on private datasets from a single scanner with the same acquisition settings for each patient. Such solutions might not perform equally well on other datasets, limiting their general usability and therefore value. Materials and Methods The generalization is best improved through a more varied dataset, however this might be impossible to achieve due to the increased price, complexity or other limitations. As an alternative, we propose to incorporate a DL model trained on a wide range of MR contrast as a pre-processing step by training a model for sCT generation using artificial PD, T1 and T2 maps. Using a dataset of only T2w MR images, the robustness of this approach is compared to a model that was trained using the MR images directly. The difference between the two approaches is shown on the low resolution figure below.

Results On T2w images acquired using the same settings as for the training dataset, both models perform similarly well. While the baseline model performs best on T2w images, our proposed robust method shows no significant change in performance between T1w and T2w contrasts, evaluated on images from three different scanners. Evaluating on five different contrast and a wide range of synthetic contrast, the robustness of our proposed method is further underlined. As a more functional evaluation method, we show that our proposed robust method also performs well with regards to radiological depth calculations. Conclusion We have propagated the robustness of a machine learning solution, by using it as a pre-processing step when training a model for sCT generation. The results show that imbalanced datasets can benefit from synthetic data augmentation, without sacrificing performance on real data. Our code and trained models are also made publicly available.

Poster (Digital): Quality assurance and auditing

PO-1699 Characterization of a novel detector array for daily quality assurance in proton therapy

H. Devendranath 1 , V. Flatten 2 , N. Chofor 3 , J. Wulff 4 , B. Timmermann 5 , A.A. Schönfeld 3

1 Heinrich-Heine-Universität, Medical Physics, Düsseldorf, Germany; 2 Marburger Ionenstrahl-Therapiezentrum, Medical Physics, Marburg, Germany; 3 Sun Nuclear, A Mirion Medical Company, Research, Melbourne, USA; 4 Westdeutsches Protonentherapiezentrum Essen, Medical Physics, Essen, Germany; 5 Westdeutsches Protonentherapiezentrum Essen, Proton therapy, Essen, Germany Purpose or Objective The aim of this project was to characterize and verify a prototype detector array for daily quality assurance (QA) in proton therapy according to TG 224. A fast array calibration method and corresponding QA dose maps were tested and evaluated. Materials and Methods The prototype array is designed to verify the spot position and size, as well as the energy range of proton pencil beams in daily routine quality assurance. This is achieved by a unique layout of air-filled ionization chambers: a central chamber surrounded by four others. The array is composed of 25 chambers, which are grouped in five clusters, each associated with different build-up depths. The calibration method, as well as the sensitivity and accuracy of the calibrated array towards changes in the aforementioned beam parameters was investigated on several active scanning proton beam lines. Results The array was able to reliably detect and characterize the changes in the spot shift, spot size and energy, while the calibration method takes less than 7 minutes and accounts for beam line variations of the pencil beam characteristics, as well as response variations of the individual chambers. The delivery of the complete QA dose map, which is used on daily