Author: Hilary Louisa Byrne, Owen Thomas Dillon, Paul J. Keall, Hunor Kertesz, Ricky O'Brien, Tess Reynolds 👨🔬
Affiliation: University of Sydney, Image X Institute, Faculty of Medicine and Health, The University of Sydney, Medical Radiations, School of Health and Biomedical Sciences, RMIT University 🌍
Purpose: Imaging of lung motion and function has typically required multiple volumetric CT images to create deformation vector fields (DVF) for motion and ventilation images (function). However, CT scanners are expensive, high dose, and not always readily available. To enable low cost, low dose and accessible 3D lung motion and lung function we developed a novel method called motion tomography. Motion tomography calculates the 3D DVFs directly from sparse 2D x-ray projection data acquired on a standard x-ray system.
Methods: 2D x-ray projection data was acquired in inhale and exhale states at imaging angles of 0, ±36 and ±72 degrees. The motion in the projection is estimated using deformable images registration (DIR). The 3D DVF was reconstructed from the 2D motion estimate by using the total-variation iterative image reconstruction algorithm. The method was tested in silico using projections of the XCAT digital thorax phantoms and experimentally by acquiring x-ray projection data of the PixelPrint inhale/exhale anthropomorphic lung phantoms. The 3D DVFs obtained from the motion tomography method were compared with the DVFs from the DIR of the inhale/exhale volumes using the metrics of vector magnitude correlation and mean directional accuracy.
Results: The 3D DVFs were successfully reconstructed from 5 projection angles. The vector magnitude correlation was 79% and 82% for the XCAT computational phantom and PixelPrint 3D printed phantom respectively. The mean directional accuracy was 79% for the XCAT and 78% for the PixelPrint phantom. A two-order of magnitude reduction in imaging dose was estimated for x-ray motion tomography compared to CT imaging.
Conclusion: The initial proof-of-concept study on virtual and experimental phantoms proved the feasibility of reconstructing the 3D DVFs directly from only 5 x-ray projection image pairs without the need to reconstruct the anatomical information.