An Optimal Solution Study on a Non-Invasive Stereotactic Radiosurgery Instrument Optimizer for Stereotactic Radiosurgery of Single Brain Metastasis

Abstract

<h3>Purpose/Objective(s)</h3> This study looks at a fast inverse planning (FIP) optimizer that enables more efficient and less planner dependent stereotactic radiosurgery (SRS) treatment planning. FIP generates a plan using multiple user-inputted variables that affect plan quality metrics, such as target coverage (TC), maximum dose to the target and organs-at-risk, low dose (LD), and beam-on time (BOT). The objective of this study is to systematically characterize each of these variables and determine which combination leads to the best SRS plan quality metrics. <h3>Materials/Methods</h3> Thirty patients with a single brain metastasis who were previously treated with SRS were evaluated. Each lesion was planned using FIP by varying the LD and BOT variables in 0.1 increments from 0.0 to 1.0 (i.e., plan #1 [0.0, 0.0], plan #2 [0.0, 0.1], ..., plan #120 [1.0, 0.9] and plan #121 [1.0, 1.0]). All plans were generated with the same prescription dose (24 Gy/1 fraction), TC (100%), maximum allowed dose (43.2 Gy), and Gamma angle (90°). The following metrics were recorded for each plan to evaluate plan quality: Paddick conformity index (PCI), gradient index (GI), and BOT. Heat maps for each of the above metrics as a function of LD and BOT penalties were analyzed. An objective score matrix was calculated for each plan using a linear weighted combination of scaled and normalized GI, PCI, and BOT. An optimal solution space containing the 5 lowest values of the score matrix per patient was extracted. A frequency histogram was subsequently created in the LD-BOT coordinate space. <h3>Results</h3> The target volumes in the study ranged from 1.0 cc to 2.5 cc. For every patient, 121 plans were generated across the range of LD and BOT penalties, resulting in 3630 plans included in this analysis. The range of PCI, GI, and BOT across all plans were 0.71 to 0.95, 2.5 to 2.9, and 17.2 to 221 minutes, respectively. The variation of the LD/BOT penalties had a much larger impact resulting in a higher deviation about the mean for BOT (169%), followed by PCI (28%). Heat map analysis revealed that the plans that maximized PCI and minimized GI values occupied the upper triangular matrix of the LD/BOT penalty space, while plans that minimized BOT occupied the lower triangular space. The optimal solutions frequently occupied the diagonal space in between these zones. Specifically, the histogram analysis showed that 4 highest frequencies occurred at the user-inputted settings (LD, BOT) of (0.6, 0.5), (0.8, 0.6), (0.7, 0.5), and (0.7, 0.6), in order of plan quality. <h3>Conclusion</h3> In this study, the user-inputted variables (LD and BOT) were systematically varied to determine the optimal solutions to simultaneously maximize PCI while minimizing GI and BOT. The results of this study may permit SRS planners to generate single brain metastasis plans more efficiently using the study-defined optimal settings of LD and BOT.