2574 - Comparative Evaluation of Tile Estimation Workflows for STaRT Pre-Planning in Patients with Recurrent Brain Metastasis
Presenter(s)
V. Chaswal1,2, E. Y. Akdemir1, M. A. Garcia3, R. Herrera1,2, V. Mishra1, M. W. McDermott4, V. Siomin5, M. D. Hall1,2, R. H. Press1,2, Y. Weiss1, M. P. Mehta1,2, A. Gutierrez1,6, R. Kotecha1,7, and R. P. Tolakanahalli1; 1Department of Radiation Oncology, Miami Cancer Institute, Baptist Health South Florida, Miami, FL, 2Herbert Wertheim College of Medicine, Florida International University, Miami, FL, 3GT Medical Technologies Inc, Tempe, AZ, 4Department of Neurosurgery, Miami Neuroscience Institute, Baptist Health South Florida, Miami, FL, 5Department of Neurological Surgery, Miami Cancer Institute, Baptist Health South Florida, Miami, FL, 6Florida International University, Herbert Wertheim College of Medicine, Miami, FL, 7Miami Cancer Institute, Baptist Health South Florida, Miami, FL
Purpose/Objective(s): Surgically Targeted Radiation Therapy (STaRT) places Cesium-131 seeds embedded in collagen tiles within the surgical resection cavity (RC) to provide locoregional radiation. The quantity of tiles implanted depends on the tumor bed surface area (ISA)—often requiring estimate-based user-input(s). A pilot trial, introducing a look-up-table (LUT) for tile estimation based on the longest pre-op resection cavity dimension, was introduced by the manufacturer. This retrospective study compared pre-planning estimated tile numbers, using three available methods, to the actual implanted tile number to determine the optimal method.
Materials/Methods: Tile estimates for thirty-two (n=32) GTM101 registry patients with thirty-three (n=33) RCs who underwent STaRT (60 Gy at 5 mm depth) were included. To determine ISA, Workflow-1 (WF1), an empirical approach, utilized three cardinal dimensions of the pre-operative Gross Tumor Volume (pre-op GTV) with a ~20-40% reduction for the operative pathway and mass effect; whereas, Workflow-2 (WF2), a contour-based workflow, employed a ~2-3 mm uniform contraction of the pre-op GTV. Total tiles were calculated as ISA/4—rounded-up to the nearest 0.5 tile. Workflow-3 (WF3) utilized the manufacturer-provided LUT based on the largest dimension of pre-op GTV. Tile estimates from three WFs were compared to actual implanted tiles. A Friedman test followed by Bonferroni- post-hoc comparisons identified differences between WFs.
Results: Thirty-three RCs with a median volume [Range] of 5 cc [2-54 cc] were implanted with a median of 3.5 [1.5-8.0] tiles. Overestimate by >1 tile was seen in 66.7 % (22/33), 33.3 % (11/33) and 96.7 % (32/33) of implants using WF1, WF2, and WF3, respectively. Underestimation by at least 1 tile was seen in 39.4 % (13/33) with WF2 as compared to 9.0 % (3/33) and 0% for WF1 and WF3. Median over-estimation by 3 [1-4.5] tiles was observed for WF3. A spot-on estimate (estimated = implanted), of 24.2 % (8/33), 27.3 % (9/33) and 3.0% (1/33) was seen for WF1, WF2 and WF3, respectively. Overall Friedman test yielded significant differences among the workflows (p= 0.01) but post-hoc comparisons showed that only the WF2/WF3 difference was statistically significant (p = 0.0096).
Conclusion: This study shows that preplanning tile estimation WFs, employing an estimate-based user-input(s) on % shrinkage (WF1) or shrinkage margin (WF2) to derive ISA, are not immune to discrepancies resulting in underestimation of tiles needed for STaRT, occasionally. WF2 performed the worst (~39% underestimations). Slight overestimation of tiles remains a preferred strategy to ensure an adequate number of tiles for the procedure; therefore, WF3 (LUT approach) is more promising as it minimizes estimates-based user-input(s) error at the expense of a median 3 tiles overestimation.