(Tuesday, 7/31/2018) 1:15 PM - 1:45 PM
Room: Exhibit Hall | Forum 4
Purpose: To evaluate the effect of dose calculation algorithms on lung SBRT optimization and plan quality scoring. Dose algorithms of varying accuracy are used in lung SBRT. Many studies have reported the effect of these algorithms on dose calculation, but few have evaluated that on inverse optimization and plan quality scoring.
Methods: Using the dataset of the 2017 Lung SBRT Plan Challenge and following its plan quality calculation protocol, plans were optimized with identical beam settings and optimization objectives using three dose algorithms of varying accuracy: PBC, AAA, and AXB. For higher accuracy algorithms (AAA and AXB), two plans were separately optimized with and without the “intermediate dose calculation (IDC)� turned on. For lower accuracy algorithms (PBC and AAA), the plans were also recalculated using AXB. The plan score (perfect score=150) and individual plan quality metrics were evaluated and compared among all plans using the ProKnow plan evaluation module designed for the Plan Challenge.
Results: With IDC, the plans optimized using PBC, AAA, and AXB achieved plan quality scores of 138.68, 134.59, and 129.45 respectively (the population mean score was 123 in the Plan Challenge). Without IDC during the optimization, scores dropped substantially to 108.28 (AAA) and 105.82 (AXB). When recalculating the PBC and AAA optimized plans with AXB, scores also dropped substantially to 98.14 (PBC) and 116.21 (AAA). Among all plan quality metrics, the major contributors to the score reductions were the coverage metrics for PTV and ITV doses, and the high-dose conformation number. The changes on the OAR and conformation metrics in the intermediate and low dose regions were relatively small.
Conclusion: Dose algorithm accuracy considerably impacts plan quality scoring. Lower accuracy algorithms may inflate scores via over-estimated target coverage. Intermediate dose calculation is helpful in optimization with higher accuracy algorithms.
Funding Support, Disclosures, and Conflict of Interest: We thank Dr. Benjamin Nelms of ProKnow for his help on this project.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
