ESTRO 2025 - Abstract Book

S3839

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

(4) Tawfik A, et al. Discrepancy between clinical and pathological staging of laryngeal carcinoma: a dilemma to be solved. European Archives of Oto-Rhino-Laryngology. 2024 May 1;281(5):2507–13

3694

Digital Poster AI-assisted quantitative CT analysis of pulmonary changes after single-fraction, breath-hold SABR for peripheral lung tumors Omar Bohoudi 1 , John R van Sornsen de Koste 1 , Miguel A Palacios 1 , Letícia Rittner 2 , Diedre Carmo 2 , Famke L Schneiders 1 , Anna M.E Bruynzeel 1 , Suresh Senan 1 1 Radiation Oncology, AmsterdamUMC, Amsterdam, Netherlands. 2 School of Electrical and Computer Engineering, Universidade Estadual de Campinas, Campinas, Brazil Purpose/Objective: Quantitative CT (qCT) and AI-based segmentation are increasingly used to study structural changes in airways, pulmonary vessels and parenchyma in patients with lung diseases. To study pulmonary changes following lung stereotactic ablative radiotherapy (SABR), we applied an AI-assisted qCT analysis to images from patients treated with single-fraction (SF) breath-hold SABR. Material/Methods: An ethics-approved patient database was used to identify patients with a peripheral tumor who completed breath hold SF-SABR (28-34 Gy) on an MR-linac. Ten patients with high-resolution pre- and post-SABR CT scans (HRCT <1 mm) and no prior lung radiotherapy were included. Airway diameter changes were analyzed using an AI-based airway segmentation method[1]. The airways were divided into segments (Boyden nomenclature) for dosimetric analysis. Pulmonary vessels were segmented and measured on CT using intensity- and shape-based methods, including Hessian-based vesselness filters. Total blood vessel volume (TBV) and volumes of small (cross-sectional area <5 mm²; BV5) and larger vessels (5–10 mm²; BV5-10) were calculated per lobe. Lobe-specific BV5 and BV5-10 were normalized to TBV as BV5/TBV and BV5-10/TBV ratios. Lung parenchymal changes, including ground-glass opacities (GGO) and consolidations (CON), were segmented using the same AI tool[2]. Lobe-specific GGO and CON volumes were calculated. Pre- and post-SABR metrics were compared to identify structural changes. Results: A total of 48 high-resolution CT scans from 10 patients were analyzed. AI-assisted detection of airway stenosis and occlusion was observed in sub-segmental bronchi: 5% in 4th generation, 41% in 5th, 35% in 6th, and 19% in 7th generation branches. Figure 1B highlights a dose-dependent effect, with higher radiation doses linked to increased stenosis and occlusion risk (p < 0.0001). Significant post-SABR vascular changes were detected in 7 of 10 patients. Figure 2 illustrates pre- and post-SABR vasculature in the treated upper lobe, showing pruning of small vessels and dilation of larger peritumoral vessels after SABR . On average, BV5/TBV ratios decreased by 6% (p < 0.001), while BV5–10/TBV ratios showed a 1% increase. Parenchymal GGO increased by 15cc (CI: 0–36 cc), and CON by 3cc (CI: 1– 4.5 cc), primarily in high-dose regions.