Purpose: The ability to track tumor motion without implanted markers on a standard linac will enable wide access to adaptive radiotherapy. This study presents the first prospective implementation of markerless tumor tracking (MTT) on a standard linac and its quality assurance (QA) procedure.
Methods: A QA procedure following the AAPM TG147 report was developed to measure static and dynamic localization accuracy and latency. A CIRS lung phantom was attached to the HexaMotion platform to emulate 3D motion from seven static and five dynamic motion traces that en-compass different types of lung tumor motion. For MTT a patient specific model was built from 4D treatment planning data. On the treatment day, a pre-treatment imaging arc was acquired to update for daily motion information. KV-projections were acquired during VMAT-treatment delivery. The model was exploited to find the projected tumor and an extended Kalman filter was used to infer the 3D tumor position. MTT was implemented in the VMAT-treatment workflow using only clinical imaging protocols. The linac was operated in clinical mode, while a frame-grabber-software transmitted kV-projections to a dedicated laptop. The performance was accelerated using a GPU and the OpenCV package.
Results: The prospective evaluation was successful. The mean tracking error was left-right (LR): 0.5Â±0.2mm, superior-inferior (SI): -0.7Â±0.2mm and anterior-posterior (AP): -0.0Â±0.3mm for the static localization test, and LR: -0.2Â±1.5mm, SI: -0.5Â±0.7mm and AP: -0.3Â±0.9mm for the dynamic localization test. The computational latency for each kV-projection was 108Â±8ms.
Conclusion: This study presents the first implementation of MTT on a standard linac. A QA procedure was developed to benchmark tracking accuracy and precision. This study paves the way for the first clinical use of MTT on a standard linac, which has the potential to enable real-time adaptive radiotherapy for a large population of cancer patients. Ethics have been submitted for related clinical trials.