Poster - Thurs Eve-35: Error reduction in variable angle implant reconstruction by optimization of imaging geometry

Movement towards intraoperative dosimetric analysis of permanent seed brachytherapy implants has driven seed localization techniques that may be applied in an operating suite. Variable angle reconstructions with a C-arm are a widely available, cost effective and easily integrated solution but are subject to mechanical limitations on imaging geometry accuracy. Phantom reconstructions of known seed locations were used to show optimization of the imaging geometry for consistency with the observed datasets in multiple variable angle images can be used to obtain more accurate source and detector positions and reduce uncertainty in reconstructed seed positions. Furthermore, the most difficult part of backprojection methods is the matching of corresponding seeds between the variable angle images. With improved accuracy in imaging geometry, tighter constraints on possible seed matches can be used. This reduces the number of potential matches between images. Application to typical clinical data has shown a reduction in the number of possible matches. The benefits are observed when optimizations are performed on very small subsets of matched seeds, allowing a small number of obvious matches due to orientation or peripheral positioning to greatly simplify matching over the remainder of the implant.