‘On the fly’ reconstruction and tracking system for patient setup in radiation therapy

Range imaging devices have already shown their value for patient setup and motion management in external beam radiation therapy. However current systems need several range imaging devices recording the patient’s surface from different viewpoints to achieve the required stability and accuracy. Since range imaging devices come as add-ons to regular linear accelerators, they have to share the limited space with other sensors in the treatment room to get a line of sight to the patient’s isocenter. The objective of this work is to describe a new registration framework which enables stable tracking using only one range imager. We unveil the design of our solution to the problem of tracking a patient over long trajectories and large viewpoint changes including surface acquisition, pose estimation and simultaneous surface reconstruction. We evaluate the performance of the system using three clinically motivated experiments: (i) motion management, (ii) non-coplanar patient setup and (iii) tracking over very large angles. We compare our framework to the state-of-art ICP algorithm and to a ground-truth stereoscopic X-ray system from BrainLab. Results demonstrate that we could track subtle movements up to 2.5 cm with a mean target registration error of 0.44 mm and 0.02°. Subsequent non-coplanar field setup on 30° and 2 cm motion yielded a target registration error of 2.88 mm which is within clinical tolerances. Our sensor design demonstrates the potential of simultaneous reconstruction and tracking algorithms and its use for patient setup and motion management in radiation therapy.