Optimizing X-Ray Beam Quality for Low-Dose Upright Cone-Beam Breast CT Using Advanced Image Reconstruction Algorithm
Abstract
Purpose
To determine the optimal X-ray beam quality by varying the copper (Cu) filter thickness for upright, cone-beam breast CT (CBBCT) to achieve a mean glandular dose (MGD) of approximately 4.5 mGy that is suitable for breast cancer screening, while maintaining microcalcification visibility.
Methods
Five acquisition protocols were evaluated using a breast phantom containing 0.24 mm to 0.32 mm diameter spherical calcium carbonate microcalcifications. For all protocols, the following parameters were fixed: 60 kVp, 50 mA, 20 ms pulse width, and 210-degree scanning angle. The Cu filter thickness was varied (0.05, 0.10, 0.15, 0.20, 0.25 mm) and the number of projection views was adjusted to maintain the MGD at approximately 4.5 mGy. The specific protocols were: (a) 0.25 mm/210 views/4.27 mGy, (b) 0.20 mm/126 views/4.53 mGy, (c) 0.15 mm/90 views/4.52 mGy, (d) 0.10 mm/62 views/4.54 mGy, and (e) 0.05 mm/36 views/4.60 mGy. All datasets were reconstructed using the Fast, total variation-Regularized, Iterative, Statistical reconstruction Technique (FRIST) algorithm that is optimized for sparse-view acquisitions. Two readers with over 10 years of breast imaging expertise independently evaluated microcalcification cluster visibility, defined as detection of at least 4 out of 6 microcalcifications in a cluster.
Results
The protocol with 0.15 mm Cu filtration and 90 projection views resulted in MGD of 4.52 mGy and provided optimal 0.24 mm calcification visibility. Protocols (d) and (e) with extreme angular undersampling (62 and 36 views) exhibited streak artifacts, revealing the limitations of FRIST for such sparse acquisitions. Thicker filtration protocols had higher noise per projection and showed reduced microcalcification visibility despite increased angular sampling.
Conclusion
For low-dose upright cone-beam breast CT, 0.15 mm Cu filtration with 90 views represents the optimal balance between beam quality and angular sampling while avoiding undersampling artifacts and enables reliable microcalcification detection at MGD of 4.5 mGy.