Author: Hyosung Cho, Dae Yup Han, Duhee Jeon, Jiwon Park, Hyesun Yang 👨🔬
Affiliation: Department of Therapeutic Radiology, Yale University School of Medicine, Yonsei University 🌍
Purpose: To precisely detect intrafraction patient motion during spine stereotactic body radiotherapy (SBRT) treatment delivery, we propose an optimized technique for three-dimensional reconstruction from the triggered kV images, employing a dehazing algorithm.
Methods: To emulate patient intrafraction motion, triggered kV images were acquired using an anthropomorphic phantom while delivering an L-spine SBRT treatment plan. Each kV image is processed with a dehazing algorithm, regarding the scattering effect of soft tissues as analogous to fog, enabling the separation of soft tissue and bone structures for enhanced visualization of the bone. The processed kV images are reconstructed in three dimensions using optimized geometry, predefined to enhance sparse-view image quality. The geometry was determined by exploring 10–40 projections over a total rotation angle of 180°–360°, ensuring sufficient image quality for bone contour identification with a minimal number of projections. Intentional 0–2 cm shifts were applied to the phantom to assess motion artifact and its impact on image quality in the axial plane.
Results: Three-dimensional CT images with a dehazing algorithm effectively enhance the visualization of central bone structures making it more suitable for contouring fiducial markers or bones to confirm patient positioning during SBRT. When comparing image quality across 5 different geometry settings, it was concluded that reconstruction using only 20 projections over the 360° angular range (triggered angle=18°), provided sufficient quality for accurately contouring the targets. In addition, noticeable motion artifacts were observed in the cross-sectional plane when shifts exceeded 2 mm.
Conclusion: Our method not only enabled the detection of patient motion in the cross-sectional plane, potentially improving the treatment precision and post-evaluation quality, but also demonstrated enhanced clarity, even in lateral views, compared to standard triggered kV X-ray images.