TH‐C‐J‐6B‐02: Towards Intra‐Operative Dosimetry in Prostate Brachytherapy

Purpose: Intra‐operative dosimetric optimization of TRUS‐guided prostate brachytherapy implants requires localization of seeds relative to prostate [1]. Analytical tools are available to intra‐operatively tailor an implant‐plan, thereby accounting for inevitable deviations [2]. Method and Materials: The majority of the practitioners have C‐arm fluoroscopes in the treatment room, making intra‐operative dosimetry feasible with little additional cost. The obstacles towards intra‐operative dosimetry are: (a) discerning the 3D poses of fluoro images, (b) registering fluoroscopy to TRUS, and (c) establish seed correspondences in multiple fluoro images. We address the first two issues by single‐image‐based fluoroscope tracking (FTRAC) fiducial with salient attributes: small dimensions (3×3×5cm); need not be close to the anatomy of interest; auto‐segmentable; and mathematically robust to segmentation, calibration, and image distortion errors. The 3D coordinates of the segmented seeds are calculated upon resolving the correspondence of seeds in the multiple X‐ray images, achieved by an algorithm called MARSHAL. We formalize seed‐matching as a network flow problem, which has salient features: (a) extensively studied exact solutions, (b) performance claims on the space‐time complexity, (c) optimality bounds on the final solution. Results: The FTRAC fiducial tracks a C‐arm to an accuracy of 0.56mm in translation and 0.33° in orientation, comparable to commercial tracking systems. On pre‐segmented images, MARSHAL achieved 100% correct matching in simulation experiments. In experiments on a precision‐machined phantom, the FTRAC and MARSHAL correctly matched and reconstructed 98.5% of the seeds with a mean 3D accuracy of 0.63mm, while the mean error for the mismatched 1.5% of seeds was only 0.91mm. Conclusion: Performance of FTRAC and MARSHAL appear to be adequate for intra‐operative dosimetry in brachytherapy.