2254 - Evaluation of Scriptable Surface Area Informed Optimization (SAIO) to Enhance a Vendor Automated Multi-Metastases Planning Solution

D. D. Desai¹, I. L. Cordrey², E. L. Johnson³, A. Magliari⁴, P. Halvorsen⁵, R. A. Schulz⁶, H. Liu⁷, and S. Beriwal⁸; ¹Memorial Hospital, Chattanooga, TN, ²Thompson Cancer Survival Center, Cumberland Radiation Oncology, Crossville, TN, ³University of Kentucky Medical Center, Lexington, KY, ⁴Varian Medical Systems Inc, Palo Alto, CA, ⁵Varian Medical Systems, Boston, MA, ⁶Varian Medical Systems, Palo Alto, CA, ⁷Department of Radiation Oncology, University of Pennsylvania, Philadelphia, PA, ⁸Varian, Palo Alto, CA

Purpose/Objective(s): Previous work has demonstrated that using surface area information in the design of a plan optimization strategy can generate high quality multi-met SRS plans without leveraging a purpose-built Stereotactic Radio Surgery Normal Tissue Objective (SRS-NTO) or patient specific static collimator rotation settings. In this work we evaluate combining the strength of SRS-NTO and patient specific collimator rotation from HyperArc (HA) with adding Surface Area Informed Optimization (SAIO) objectives and related structures. We also announce the public availability of this SAIO automated treatment planning system method in open source. Materials/

Methods: Three multi-target cases and three single target cases were each planned with both HyperArc (HA) alone and with SAIO objectives added (HA+SAIO). The plans were evaluated with three metrics: CI_RTOG, CI_Paddick, to assess prescription dose conformality, and R50% to assess intermediate dose spill outside the targets. Wilcoxson sign ranked tests were performed to assess statistical relevance of any difference in the metrics.
Results: The aggregate results show HA+SAIO demonstrated improvement over HA alone for all three indices. The CI_RTOG, CI_Paddick, and R50% Wilcoxson results show that SAIO improves each index with a p-value < 0.000014.
Conclusion: Adding SAIO objectives improved multi-met SRS plans in all circumstances. This is a statistically significant improvement to HA alone. Adding the optimization objectives and related structures required for SAIO can be fully automated utilizing the free ESAPI tool and source code on GitHub. Abstract 2254 – Table 1

Case	Target(s)	Method	Mean CIRTOG	Mean CIPaddick	Mean R50%
A	1	HA alone	1.05	0.93	2.92
		HA+SAIO	1.05	0.94	2.78
B	1	HA alone	1.16	0.82	4.31
		HA+SAIO	1.18	0.83	3.94
C	1	HA alone	1.45	0.68	5.42
		HA+SAIO	1.28	0.76	4.54
D	2	HA alone	1.01	0.93	2.62
		HA+SAIO	0.97	0.95	2.30
E	5	HA alone	1.02	0.91	3.94
		HA+SAIO	1.00	0.93	3.34
F	10	HA alone	1.14	0.84	6.18
		HA+SAIO	1.07	0.89	4.43
All	All	HA alone	1.11	0.87	4.79
		HA+SAIO	1.05	0.90	3.84
		p-value	0.000004	0.00001	0.00001