ESTRO 2024 - Abstract Book

S3934

Physics - Image acquisition and processing

ESTRO 2024

Keywords: Deep learning, dose reduction, CBCT

References:

[1] Liu, H. et al. (2023). Review of cone beam computed tomography based online adaptive radiotherapy: current trend and future direction. Radiation Oncology, 18(1), 144. https://doi.org/10.1186/s13014-023-02340-2

[2] Ding, G. X. et al. (2018). Image guidance doses delivered during radiotherapy: Quantification, management, and reduction: Report of the AAPM Therapy Physics Committee Task Group 180. Medical Physics, 45(5), e84–e99. https://doi.org/10.1002/mp.12824

[3] Wang, X. et al. (2019). ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks. ECCV 2018. Lecture Notes in Computer Science, vol 11133. Springer, Cham. https://doi.org/10.1007/978-3-030-11021-5_5

Acknowledgment: DKH 70114849

2562

Poster Discussion

AI-enhanced proton radiography: deep learning for assessing treatment deviations in proton therapy

Giuliano F. Perotti Bernardini, Gabriel Guterres Marmitt, Arthur Galapon, Peter van Ooijen, Johannes A. Langendijk, Stefan Both

University Medical Center Groningen, Radiation Oncology, Groningen, Netherlands

Purpose/Objective:

Treatment deviations in the proton range can result in overdoses in healthy tissues and imprecise dosages in target volumes. Imaging techniques can pinpoint uncertainties in proton therapy, such as patient misalignment, anatomical changes, and CT image conversion issues [1]. For in vivo proton range verification, proton radiography (PR) has been proposed as a quality control (QC) method [2]. However, generating and interpreting PR images remains challenging in the fast-paced and complex clinical workflow. To address this challenge, an AI-enhanced PR tool was developed to automatically and rapidly evaluate PR images to recognize multiple sources of error affecting the proton range.

Material/Methods:

A dataset of 32 CT scans of head and neck (HN) cancer patients was processed and subjected to artificial changes programmed in RayStation to mimic proton range deviations often seen during treatment. The changes included