2433 - Driving Efficiency of Lung SAbR X-Ray Based Adaptive Planning with TPS-Integrated Automation Tools

J. Visak¹, C. Y. Liao¹, H. J. Wang¹, X. Zhong¹, S. Domal¹, B. Washington¹, D. D. M. Parsons¹, D. Nguyen², D. X. Yang¹, Y. Zhang³, K. D. Westover¹, and M. H. Lin¹; ¹Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, ²MAIA Laboratory, Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, ³Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, TX

Purpose/Objective(s): Adaptive radiotherapy planning (ART) using X-ray imaging requires skilled planners to ensure robust contouring and planning strategies in lieu of potential changes between treatments. We describe our early experience with a novel X-ray-based ART treatment planning system (TPS) equipped with AI-based organ at risk (OAR) delineations, automatic beam geometry selection, and intelligent knowledge-based (KBP) optimization for adaptive lung stereotactic ablative body radiotherapy (SAbR) workflow. We hypothesized that utilizing TPS-integrated automated tools would streamline ART planning with minimal user interaction. Materials/

Methods: Ten clinical lung SAbR patients treated with X-ray imaging-based ART, who received 50 Gy in 5 fractions using intensity-modulated radiotherapy (IMRT) were selected. Planning target volume (PTV) size ranged from 9.9-88.3cc. AI-based OAR delineations were compared to clinical contours with dice similarity coefficient. An IMRT (aIMRT) and volumetric-modulated arc therapy (aVMAT) plan was generated with integrated automated beam geometry selection and KBP guidance with limited user interaction. Plans were compared for target and dosimetric coverage. Statistical significance was assessed using paired student t-tests.
Results: OAR dice scores exceeded 0.95 for the heart and lungs. Bronchus, trachea and esophagus reported 0.62-0.77 due to differences arising from AI segmentation compared to institutional protocol. Automatic plans were generated on average in under 15 minutes and typically met institutional RTOG-0813 based criteria using KBP-driven guidance. The aVMAT plans were similar to benchmark cases while requiring less planner burden. The average conformity index was 1.02 for all plans. Gradient index was 4.74±0.7 (p<0.001), 4.22±0.5 (p=0.003) for aIMRT and aVMAT compared to benchmark 3.96±0.5. No statistical differences were seen in total MU whereas optimization time was modestly increased on average by 2.9± 0.4 mins (p<0.001) in aVMAT cases. The aIMRT suffered from excessive intermediate dose spill due to the algorithm-selected 7 fields compared to 9-12 fields used clinically.
Conclusion: The integration of automated treatment planning tools within a novel X-ray-based ART TPS preserved dosimetric coverage compared to manual processes while significantly improving the efficiency of planning lung SAbR.

Table 1. Example Plan Metrics for Automated Reference Adaptive Lung SAbR Plans. Statistical Significance to Benchmark Cased Denoted With *.

Structure	Metric	aIMRT	aVMAT	Benchmark
PTV	CI	1.02 ± 0.02	1.02 ± 0.02	1.02 ± 0.03
	GI	*4.74 ± 0.73	*4.22±0.50	3.96 ± 0.45
	D0.03cc(cGy)	*115 ± 2.3	*116 ± 1.0	119 ± 3.6
Spinal Canal	D0.03cc(cGy)	992 ± 679	917 ± 429	914 ± 495
Bronchus	D0.03cc(cGy)	*2282 ± 1683	*2233 ± 1770	2435 ±1 742
Esophagus	D0.03cc(cGy)	1063±577	1115±691	1227±644
Lungs-ITV	V13.5Gy(%)	*7.6±11	*7.0±12	6.7±12
Plan Delivery	Total MU	4548±867	4326±617	4787±988
	Opti Time (s)	249.4 ± 31.0	70.8 ± 9.0	--