SU-E-T-506: Low Dose Fluoroscopy Based Detection of Implanted Marker Position on the VERO System for Real-Time Tumor Tracking

Abstract

Purpose: To determine the balance between kV fluoroscopy imaging dose and image quality, for robust detection of implanted fiducial markers in lung on the VERO real-time tumor tracking system. Methods: Tumor visibility in fluoroscopic images can be improved with an implanted fiducial marker in or near the tumor. In total 5 types of fiducial markers were used in this study: solid gold markers, helical coils, multi-gold-marker assemblies, spherical markers and stent-like markers. To investigate automatic detection performance of markers projected on high density structures in the thorax images, an anthropomorphic phantom was placed on a moving platform behind the static markers. Images were acquired with variable imaging settings and in different anatomical circumstances. A marker detection algorithm was developed using maximization of normalized cross correlation with a template library incorporating marker rotations. The templates were created from a 3D model of the marker. The marker detection was applied on 25 images per sequence and detection scores calculated. Additionally imaging dose TLD measurements were conducted for different imaging settings. Results: Differences were seen in the marker detection for different marker types in the different anatomical situations. Beside the intrinsic attenuation of the marker, also its spatial structure played a role. Increasing the mAs did not always increase detection scores. For 0.5 mAs, 100kVp and an SSD of 900mm a skin dose was measure of 0.031 mGy/image and an exit dose of 0.001 mGy/image. At a frame rate of 14 frames/sec this would mean an imaging skin dose of 26.04 mGy/min. Conclusions: It is feasible to use marker-based fluoroscopy based real-time tracking with imaging doses comparable to CBCT. Markers with more elaborate structures made from high density materials have, beside an advantage of being less prone to migration, also an advantage in terms of detectability. This collaborative work was supported by the Flemish government through the Hercules foundation and the “Fonds voor Wetenschappelijk Onderzoek - Vlaanderen” grants G.0486.06 and G.0412.08, and corporate funding from BrainLab AG. There are no other conflicts of interest.