3375 - Quantitative and Qualitative Evaluation of an Automated Planning Solution for Prostate Radiotherapy

W. Green¹, R. McBeth², G. Güngör³, M. Moll⁴, S. Cozzi⁵, V. G. Gregoire⁶, B. Ungun⁷, M. Costea⁷, N. Bus⁸, N. Paragyos⁷, B. K. K. Teo⁹, and P. Fenoglietto¹⁰; ¹University of Pennsylvania, Philadelphia, PA, ²Department of Radiation Oncology, University of Pennsylvania, Philadelphia, PA, ³Acibadem MAA University, Maslak Hospital, Department of Radiation Oncology, Istanbul, Turkey, ⁴Medical University of Vienna, Vienna, Austria, ⁵Centre Leon Berard, Lyon, France, ⁶Centre Léon Bérard, Lyon, France, ⁷TheraPanacea, Paris, France, ⁸Therapanacea, Paris, France, ⁹Department of Radiation Oncology, Division of Physics, University of Pennsylvania, Philadelphia, PA, ¹⁰LInstitut du Cancer de Montpellier, Department of Radiation Oncology, Montpellier, France

Purpose/Objective(s): Volumetric modulated arc therapy (VMAT) is widely used in radiotherapy (RT) treatment planning, yet manual treatment planning (MP) remains labor-intensive and skill-dependent. To address this, we propose a fully automated treatment planning (AP) approach for prostate cancer treatments, aiming to streamline planning while maintaining quality. Materials/

Methods: Our pipeline takes as input a contoured planning CT, the targeted linac machine and the prescription dose. It uses a deep learning model trained on 123 clinical cases for dose prediction and direct aperture VMAT plan optimization. A clinical evaluation study was performed on another 25 cases. The RT plans generated were calculated using a Collapsed Cone Convolution engine and the obtained RTdoses were compared with the reference doses from MP. First, a quantitative evaluation was performed based on dose-volume histogram (DVH) points and plan parameters metrics (monitor units (MU) and modulation complexity score (MCS)). Paired Wilcoxon signed-rank test was used to assess significant differences between the MP and AP plans (with p<0.05 considered statistically significant). Secondly, a double blinded plan comparison study was organized, showing side-by-side the RTdoses from AP and MP plans. Three experts (from major European and US cancer centers) evaluated plan acceptability and preference.
Results: Results of the quantitative evaluation showed that the AP doses to PTV_80Gy were similar to those of MP (<0.5Gy absolute dose difference) and without statistically significant differences in the median (D50%) and maximum (D2%) doses (Table1). AP reduced toxicity to organs-at-risk (OARs), notably a 6Gy decrease in D5% to the right femoral head. Significant differences between MP vs AP were also observed in the number of MU (589.67±57.39 vs 658.48±72.52) and MCS (0.19±0.03 vs 0.17±0.03). Qualitative evaluation found both AP and MP plans clinically acceptable (87% and 85%, respectively). AP was preferred in 18/25, 13/25, and 22/25 cases by three experts, revealing varying clinical priorities. Table1. Dosimetric study results. With * are highlighted the statistically significant differences between MP and AP (p<0.05)
Conclusion: Our automated pipeline produces plans comparable to manual ones, deemed clinically acceptable and sometimes superior. Differences in expert preferences highlight variations in clinical practice across centers and countries.

Abstract 3375 – Table 1

	DVH parameter	Acceptance criteria	Manual plans (mean±SD)	Automatic plans(mean±SD)	p-value<0.05
PTV_80Gy	D98%	=72Gy	74.77±0.64	74.33±0.54	*
	D50%	=80Gy	80.25±0.45	80.17±0.29
	D2%	=85.6Gy	82.15±0.66	82.17±0.42
Rectum	Dmax	=76Gy	74.93±1.01	74.97±1.22
	D50%	=60Gy	23.52±4.77	22.47±4.11
Bladder	Dmax	=80Gy	79.46±0.55	79.63±0.65
	D50%	=70Gy	35.55±16.8	34.09±16.28
Femoral head L	D5%	=55Gy	33.84±6.68	28.86±4.72	*
Femoral head R	D5%	=55Gy	34.42±7.02	28.74±4.58	*