Data-driven Optimisation of in vivo Radioactive Source-tracking for Real-time Cancer Radiotherapy Treatment Verification

Precise treatment delivery in high dose rate radiation therapy for prostate cancer requires comprehensive treatment verification to ensure effective tumour control and patient safety. Our group has developed a system that tracks the radiation source as it moves inside the patient’s tumour. The major aim of this study was to optimise the source-tracking algorithm to improve accuracy without excessive cost in processing time to enable real-time analysis. The source is tracked by analysing the distribution of radiation from the brachytherapy source (Iridium-192) that exits the patient’s skin and reaches a Flat Panel Detector (FPD) mounted in the couch beneath the patient. The radiation distribution in this 2-dimensional ‘image’ is analysed to estimate the source position. In this study, measurements were conducted in a ‘phantom’ - an artificial surrogate for the patient - in which the ground-truth positions were known by an independent means that cannot be achieved in a live patient. Various algorithms were examined for accuracy, efficiency, and the influence of asymmetric radiation scattering caused by inhomogeneous media interfaces e.g. air/tissue boundaries. The most accurate algorithm was identified, and some tunable parameters were able to be optimised for accuracy. The comparison of measured source positions revealed some skewing of measured positions due to asymmetric scattering existing in the proximity of the phantom edge. The algorithm with the lowest sensitivity to asymmetric scattering was identified. Computation times were compared for suitability in the clinical environment where evaluation at up to 30 frames per second may be required. The optimised algorithms could improve the quality assurance value of source-position tracking in high dose rate brachytherapy.