ESTRO 2025 - Abstract Book

S1299

Clinical - Lung

ESTRO 2025

Conclusion: Moderate hypofractionated radiotherapy in combination with chemotherapy as a radical treatment for LS-SCLC is effective and safe . However, further powerful studies are needed.

Keywords: SCLC, Hypofractionation, Chemoradiation

References: Saeed NA, et al. Hypofractionated vs. standard radiotherapy for locally advanced limited-stage small cell lung cancer. J Thorac Dis. 2022 Feb;14(2):306-320. doi: 10.21037/jtd-21-1566. Qiu B, et al. Moderately Hypofractionated Once-Daily Compared With Twice-Daily Thoracic Radiation Therapy Concurrently With Etoposide and Cisplatin in Limited-Stage Small Cell Lung Cancer: A Multicenter, Phase II, Randomized Trial. Int J Radiat Oncol Biol Phys. 2021 Oct 1;111(2):424-435. doi: 10.1016/j.ijrobp.2021.05.003. Agolli L, et al. Hypofractionated Image-guided Radiation Therapy (3Gy/fraction) in Patients Affected by Inoperable Advanced-stage Non-small Cell Lung Cancer After Long-term Follow-up. Anticancer Res. 2015 Oct;35(10):5693-700.

497

Digital Poster a planning CT radiomics model to predict the risk of radiation pneumonitis after SBRT in early stage lung cancer patients or lung metastasis patients. Yuen Ying Lau 1 , Jing Cai 2 , W.S. Leung 2 1 Department of Radiotherapy, Hong Kong Sanatorium & Hospital, Hong Kong, Hong Kong. 2 Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, Hong Kong Purpose/Objective: This study analyzed the predictive value of different radiomic features extracted from different meaningful lung regions, including but not limited to target volume, from planning CT images (pCT) in patients treated with SBRT using different deep learning algorithm to obtain the best clinically applicable predictive model for radiation pneumonitis (RP). Material/Methods: We conducted a retrospective study on 122 patients who completed SBRT treatment to their lung tumors at two public hospitals in Hong Kong from 2012 to 2021, median follow-up time 36 months. They were classified as high risk RP group when they experience grade 3 or above acute radiation-induced lung injury according to CTCAE. A pool of 600 predictive models were generated with: (1) 3 feature selection methods; (2) 4 optimal number of feature; (3) 10 classification algorithms; and (4) features extracted from 5 regions of interest (ROIs). The ROIs included tumor itself, region likely under high dose, low dose lung regions and normal lung structures regardless of any radiation received. Of all models generated, the one with best AUC in both training and validation cohorts was chosen to further investigate its predictive value on RP after SBRT to lung tumors. Results: The incidence of grade 3 or above RP events was 18.2%. Features extracted from “Whole lung – PTV”, which is an ROI disregarding radiation dose, were found to have strongest predictive value to RP regardless of number of features used. Performance of predictive models generated from features in “(PTV+2cm) – Chest Wall – Cardiovascular tissue”, which is the normal lung region most likely to receive high dose (more than 50% of prescribed dose) was the worst. This highlights the fact that the more wholesome we look into the complex and heterogeneous architecture of normal lung volume, which now can be quantified from pCT, the better we could predict how well the lung can still function after treatment. The best predictive models for RP after SBRT to lung tumors was developed using GAUSS and 40 features predicting with the best radiomic feature mentioned; mean AUC=83.7%, maximum AUC=93.8%, mean ACC=83.9 and