Software Automation and Uncertainty Analysis for Ring Applicator Quality Assurance
Abstract
Purpose
To develop and validate an automated quality assurance (QA) methodology for verifying ring applicator dwell positions using kV imaging and radiochromic film, with emphasis on quantifying measurement uncertainty, reducing observer-dependent variability, and evaluating the impact of ring curvature on dwell localization accuracy.
Methods
Dwell positions were evaluated for 30°, 45°, and 60° ring applicators using kV imaging and radiochromic film. kVs were acquired of the Bravos afterloader dummy wire at 5-mm increments. A Python-based image-processing pipeline detected the dummy wire tip and applied a fixed offset (3.45 mm) to localize the radiation center based on source geometry. Radiochromic films were acquired with 0–3 mm offsets and analyzed using an automated algorithm to determine dwell positions relative to a reference angle defined at the ring lumen endpoint. TPS-planned dwell positions were generated using nominal applicator geometry with 5 mm spacing. All uncertainties were characterized. The mechanical uncertainty was identified using the positional verification test (PVT) results for dummy and source cables and compared to uncertainty with added user dependency.
Results
Daily PVT measurements demonstrated mechanical deviations within ±0.02 cm. Estimated measurement uncertainty was 0.5–1.0 mm for kV imaging and 1–2 mm for radiochromic film. Agreement between kV-derived dwell positions and film measurements was within 1 mm across evaluated positions with the largest discrepancies observed at dwell locations in regions of maximum ring curvature. The results are consistent with the combined uncertainty range of 0.73-1.43 mm. The combined uncertainty increases to 1.14-2.25 mm when adding the uncertainty associated with manual film interpretation.
Conclusion
This study presents an automated QA methodology for ring applicator dwell verification that combines kV imaging and film-based measurements while eliminating observer-dependent error, decreasing the overall uncertainty range for ring applicator QA.