A phantom-based geometric calibration method for digital breast tomosynthesis systems

Digital Breast Tomosynthesis (DBT) X-ray imaging technique generates pseudo-3D images via reconstruction algorithms, requiring precise knowledge of the acquisition geometry to ensure high spatial resolution and minimize artifacts. The objective of this thesis is to develop and validate a phantom-based method for geometric calibration of DBT systems. Projection images of a phantom with metallic markers were acquired at different angles, from which marker positions are extracted. Based on a geometric projection model, acquisition geometry parameters are estimated via iterative optimization with Tikhonov regularization. These parameters are then used to evaluate in-plane spatial resolution and microcalcifications detectability, using dedicated image quality assessment phantoms. For validation, four DBT units equipped with different detectors and two calibration phantoms are considered. The results highlight reliability in marker extraction and numerical robustness of the algorithm with respect to the initial guess and data variability. The extracted geometric parameters show a mean absolute deviation from the nominal values lower than 1% for most parameters, within the mechanical tolerances, and an average distance between extracted and predicted marker positions of 0.01mm. Consistent results were obtained across different DBT units, suggesting the calibration method’s ability to adapt to the specific geometric properties of each system. In terms of image quality, the optimized geometry yielded a limit spatial resolution comparable to that obtained with the nominal geometry. For microcalcifications detectability, similar diameters and contrasts were observed for both optimized and nominal geometries. The proposed geometric calibration method appears effective for stationary-detector DBT systems, achieving reliable image quality and demonstrating its potential for integration into industrial quality control procedures and periodic calibration in clinical settings.