ESTRO 2024 - Abstract Book

S4575

Physics - Machine learning models and clinical applications

ESTRO 2024

1. S. Momin, Y. Fu, Y. Lei, J. Roper, J. Bradley, W. Curran, T. Liu, and X. Yang, Knowledge-based radiation treatment planning: A data-driven method survey., J of App Clin Med Phys 22, 16 – 44 (2021).

2. Y. Ge and Q. J. Wu, Knowledge-based planning for intensity-modulated radiation therapy: A review of data-driven approaches., Med. Phys. 46, 2760 – 2775 (2019).

3. B. Zhang, A. Babier, T. Chan, and M. Ruschin, 3D dose prediction for Gamma Knife radiosurgery using deep learning and data modification., Physica Medica 106, 102533 (2023).

4. J. Sjölund, S. Riad, M. Hennix, and H. Nordström, A linear programming approach to 404 inverse planning in Gamma Knife radiosurgery, Med. Phys. 46, 1533–1544 (2019)

3170

Digital Poster

Impact of Tumor Margins on Radiation Pneumonitis Incidence in Advanced Stage NSCLC Radiotherapy

Ming Chao, Alyssa Gadsby, Robert Samstein, Tian Liu

Icahn School of Medicine at Mount Sinai, Radiation Oncology, New York, USA

Purpose/Objective:

Radiation pneumonitis (RP) is a common and potentially life-threatening toxicity in patients receiving radiotherapy for thoracic cancer. Finding proper predictors of RP is crucial in supporting clinician’s decision on reducing the risk and making effective interventions prior to treatment delivery. This study’s purpose is to evaluate the impact of tumor target margins on the RP incidence using a multi-institutional database of patients receiving radiotherapy for advanced stage non-small-cell-lung-cancer (NSCLC).

Material/Methods:

A total of 473 advanced stage NSCLC patients enrolled in NRG Oncology RTOG 0617 who underwent radiation therapy treatment were selected for this study [1, 2]. A binary classification was adopted for the clinical endpoints: group 1 – non-RP2 (n = 408) if the acute grade is 0 or 1; and group 2 – RP2 (n = 65) if 2 or higher. Three methods were used to assess the margin effect on the RP toxicity: total lung (gross tumor target excluded) (TL), total lung excluding clinical target volume (TL-CTV), and total lung excluding planning target volume (TL-PTV). Firstly, we examined the correlation between the commonly adopted dosimetric parameters such as V5, V20, and mean dose to lung (MLD) and radiation pneumonitis. The dosimetric variable distributions was inspected with the Shapiro-Wilk test and these variables from different margin methods were compared using repeated analysis of variance test (ANOVA). The difference of dosimetric variables between two groups of patients (RP2 and non-RP2) was compared using the Mann-Whitney U test. Both univariate and multivariate logistic regression analyses were performed to evaluate the correlation with RP incidence for these three variables separately and collectively. Secondly, based on the cumulative dose volume histograms (DVHs) from three lung volume definitions we exported 70 lung volumes with dose thresholds ranging