Dosimetric Impact of Seed Segmentation in GammaTile Surgically Targeted Radiation Therapy for Gliomas and Brain Metastases Cases

Abstract

To assess the variability of dose reporting variability due to uncertainty in segmentation of Cs-131 seeds in GammaTile therapy for gliomas and brain metastases. Ten patients with either glioma or brain metastases had 4-11 GammaTiles placed along resection bed during craniotomy. A dose of 60 Gy is prescribed to 5 mm depth. Each GammaTile has four Cs-131 seeds imbedded in a biodegradable collagen sponge. GammaTile Post-Op workflow in MIM Symphony software is used for post-implant dose evaluation and reporting. This workflow requires a post-surgery CT to identify seeds, and a post-surgery MR for residual disease and OAR contours. Seeds are segmented using a threshold tool. Threshold levels may change depending on the CT used, thus users need to manually change the HU threshold value in each data set. Since GammaTiles are lined along the resection bed, PTVs are generated automatically by adding 8 mm expansion on the seed contours and later combined with residual disease contours. We simulate the seed contour uncertainty by applying -0.5 mm, -1.0 mm, +0.5 mm & +1.0 mm concentric margins to the current seed contours to create 4 new seed contours per patient. New PTVs are generated by adding 8 mm expansion on the new seed contours combined with residual disease contours. PTV volume, PTV volume receiving 100% and 150% of prescription dose (V100, V150), and percentage of the prescription dose received by 90% of the PTV (D90) are calculated to evaluate dose reporting variability due to seed segmentation uncertainty. Mean PTV volume decreases by 8.4 cc & 10.2 cc for PTVs generated from seed contours with -0.5 mm & -1.0 mm margin, respectively, and increases by 5.8 cc & 8.2 cc, respectively, when +0.5 mm & +1 mm margins are applied to the original seed contours. We observe up to 10% change in V100 due to seed segmentation uncertainty. Mean V100 increases by 4.0% (range: 0.2% - 8.9%) & 4.9% (range: 0.5% - 11.0%) for cases with -0.5 m & -1.0 mm seed margin, respectively, and reduces by 4.2% (range: 0.5% - 6.7%) & 5.9% (range: 0.6% - 10.4%) for cases with +0.5 mm and +1.0 mm seed margin, respectively. Mean D90 increases by 7.7% (range: 4.0% - 12.6%) & 9.9% (range: 4.0% - 17.4%) for cases with -0.5 m & -1.0 mm seed margin, respectively, and reduces by 5.5% (range: 3.6% - 7.8%) & 7.4% (range: 5.2% - 9.6%) for cases with +0.5 mm and +1.0 mm seed margin, respectively. We also observe up to 8.0% changes in mean V150 when margins are applied to the seed contours. Our results show significant impact of seed segmentation uncertainty on dose reporting in GammaTile therapy. Variability in dose reporting parameters highlight the need for a more standardized and automated approach to seed segmentation to ensure consistent and accurate dose reporting. The current manual threshold adjustment method is subject to user dependence and therefore unreliable. Development of a more robust tool could help to minimize variability and improve reliability of dose reporting.