Synthetic CT‐enabled weekly adaptive radiotherapy for nasopharyngeal carcinoma: Optimizing plan adaptation triggers through volumetric–dosimetric monitoring
Zhenmei Cao, Chun Wu, Lixiang Han, Tingting Duan, Fengyu Zhang, Lansheng Zhang, Bin Wang, Weijie Lei, Xinye NiAbstract
Background and purpose
Anatomical changes during radiotherapy for nasopharyngeal carcinoma (NPC) frequently compromise target volume coverage while escalating dose delivery to organs at risk (OARs). In adaptive radiotherapy (ART), optimizing workflow efficiency to minimize clinical workload while maximizing patient benefit remains a fundamental challenge. Previous investigations inadequately tackled the critical aspects of identifying robust triggers and optimal timing for ART initiation. To bridge this gap, we quantified weekly sCT‐derived volumetric and dosimetric changes on a C‐arm linac workflow to determine robust, literature‐based thresholds and optimal replanning time points for NPC ART.
Materials and methods
This study analyzed 52 NPC patients undergoing VMAT. Weekly cone‐beam CT (CBCT) scans ( n = 312) were processed through ArcherQA using a CycleGAN‐based sCT generation pipeline, which enabled automated segmentation and GPU‐accelerated Monte Carlo dose recalculation. Longitudinal volumetric/dosimetric changes in targets and organs at risk (OARs) were tracked, with Spearman correlation analyzing variable relationships. Binary logistic regression and Receiver Operating Characteristic (ROC) curve analysis identified evidence‐based triggers and optimal timing for adaptive replanning, enabling quantitative criteria for ART initiation while balancing precision and workflow efficiency.
Results
Anatomical changes drove dosimetric deviations: PGTVn (nodal targets) showed ≥8% volume reduction by week 2 (OR = 2.03, p = 0.006, AUC = 0.83), while PGTVp/PTV1 (primary targets) required ≥10% reduction by week 4 (AUC = 0.76). OAR correlations included left submandibular gland volume inversely linking to mean dose (peak r = ‐0.575 at week 3, p < 0.001) and delayed ipsilateral parotid atrophy (week 4). Triggers were defined: ≥15% contralateral parotid reduction at week 2 (AUC = 0.89) and ≥19% left submandibular gland loss at week 3 (AUC = 0.97). Pharyngeal constrictor analysis revealed that week 2 ≥4% volume loss predicted ≥0.9 Gy week 3 dose rise (AUC = 0.93).
Conclusion
Weekly sCT monitoring clarified the optimal timing for offline ART, revealing that structure‐specific dosimetric changes necessitate distinct intervention windows, notably week 2 for nodal/contralateral parotid and week 4 for primary/ipsilateral parotid, providing a practical framework for prospective validation.