S637
Clinical – Head & neck
ESTRO 2026
challenge. The tool is designed to rapidly quantify anatomical changes and automatically flag patients with significant geometric deviations, providing a practical and efficient "adaptive decision" support system. Material/Methods: An automated C# ESAPI workflow (Fig. 1) was developed and validated on 48 NPC patients with initial (CT0) and mid-treatment (CT1) scans (avg. interval 36 days). OARs on both CTs were contoured using a FDA-cleared AI algorithm. The workflow's "engine" then executes two scripts. Script A automatically quantified geometric changes ( Δ Volume, Δ Center of Mass) for 24 structures using the clinically- approved rigid registration. To validate the tool's clinical relevance, its geometric findings were compared against dosimetric outcomes. This was done by Script B, which automatically extracted "Pass"/"Fail" statuses for 35 clinical goals, comparing the physician-approved adapted CT1_Plan with a non- adapted CT1_Hybrid (CT0_Plan recalculated on CT1). McNemar's test was used for statistical comparison.
Conclusion: To our knowledge, this is the first study to integrate germline polymorphisms with longitudinal patient- reported symptom trajectories in HN radiotherapy. Symptom trajectories followed expected acute and late patterns. Although not statistically significant, the data suggest more persistent symptoms in C-allele carriers, possibly reflecting slower fibrotic remodelling and tissue repair. Findings suggest multifactorial drivers of RT-induced effects beyond TGFB1 variation, underscoring the need for larger multi-institutional, multi-gene studies integrating multi-modality data. Keywords: patient-reported outcomes, NTCP, TGF- beta Digital Poster 3817 An Automated AI-Contouring Workflow to Quantify Deformations and Guide Adaptive Decisions in Nasopharyngeal Carcinoma Ti-Hao Wang 1,2 , JinHuei Ji 3 , Yi-Ru Chang 2 , Yen-Jung Chen 4 , Lin-Shan Chou 4 1 Department of Medicine, China Medical University, Taichung, Taiwan. 2 Department of Radiation Oncology, China Medical University Hospital, Taichung, Taiwan. 3 Research and Development, EverFortune. AI, Taichung, Taiwan. 4 Department of Heavy Particles & Radiation Oncology, Taipei Veterans general hospital, Taipei, Taiwan Purpose/Objective: The clinical decision to replan in NPC adaptive radiotherapy (ART) is complex, hindered by the high- effort, manual process of re-contouring and evaluation. This bottleneck makes routine, proactive ART impractical for many clinics. We developed and validated an automated workflow, based on AI auto- contouring and scripted analysis, to solve this
Results: The automated workflow (Fig. 1) successfully and rapidly quantified significant anatomical changes across the cohort (Fig. 2A, 2B). The tool automatically detected significant parotid shrinkage (Mean Δ Vol: - 6.46cc L, -6.71cc R) and clinically-relevant medial CoM shifts (-1.94mm L, +2.03mm R, radiological
Made with FlippingBook - Share PDF online