Predrag Bakic

In the last decade, breast cancer diagnosis has seen the widespread application of digital breast tomosynthesis (DBT) in screening programs. DBT involves the X-ray tube traversing along an arc trajectory parallel to the patient’s chest wall, acquiring multiple low-dose mammograms from various angles. Reconstruction algorithms process these projection views to generate a stack of slices forming a pseudo-3D volume of the breast. While DBT offers intricate structural visualization, variations in acquisition geometry impact lesion visualization and vertical resolution of the reconstructed slices. Studies have explored optimal parameter combinations, demonstrating their impact on lesion detection. Radiation dose is a critical factor influencing image quality, with a trade-off between projection number and exposure per projection. Virtual clinical trial (VCT) tools facilitate testing different acquisition geometries without clinical trials constraints. In this study we investigate the impact of angular range and projection number on DBT reconstructed images for two different scenarios, maintaining total radiation dose and keeping the dose per projection constant. We evaluate the quality of images and identify optimal acquisition protocols for lesion visualization assessing the contrast-to-noise ratio. Results reveal that by maintaining the total radiation dose, our denoising pipeline approaches image quality comparable to higher-dose acquisitions, demonstrating the potential for optimized DBT protocols.