Author: Petr Bruza, Alexander Geiersbach, David J. Gladstone 👨🔬
Affiliation: Thayer School of Engineering, Dartmouth College 🌍
Purpose: Cherenkov images are two-dimensional projections of the surface light emissions, and lack spatial information about the radiotherapy beam delivery. We implement the first fusion of Cherenkov images and 3D patient surface meshes to quantify beam outline deviations with respect to the patient’s outer anatomy and treatment plan with objective metrics. This work simplifies clinical workflow and enables implementation of automatic thresholds that enable Cherenkov-based treatment verification.
Methods: Cherenkov images from multiple geometrically calibrated cameras were combined with 3D CT datasets (3D SGRT datasets will be tested in the future) to provide a novel Cherenkov surface image map on the patient. Raycasting was utilized to virtually image the Cherenkov surface map from the beam’s eye view, therefore enabling spatial quantification of beam outline on the patient. Several quantitative spatial metrics were tested (DICE, Hausdorff, ICP, ICP COM, etc.) to automatically detect beam outline deviations from planned deliveries.
Results: Iterative Closest Point algorithms achieve an accuracy of less than 2 mm for the flat phantom and 3 mm for the anatomical Annie phantom. Setup error and camera resolution account for up to 1 mm of inaccuracy.
Conclusion: Cherenkov surface maps can reliably detect beam deviations of 1 mm using existing Cherenkov imaging technology. Our work demonstrates the first feasible method for automatic incident detection and allows the clinical team to adapt to treatments in real time and adjust future fractions to improve patient safety and treatment efficacy.