Super-Resolution Radiography by Mechanical Supersampling and Model-Based Iterative Reconstruction using High-Z Photon-Counting Detectors
Abstract
This study presents a practical and dose-efficient strategy for resolution enhancement in planar radiography, based on mechanically supersampled acquisition with high-Z photon-counting detectors (PCDs). Unlike prior event-based or cluster methods, our approach operates in true photon-counting mode and supports clinical flux rates. Using detector trajectories spanning multiple pixels and image registration-based shift estimation, we achieve sub-pixel sampling without requiring mechanical precision, while also compensating for motion and geometric instabilities. An iterative reconstruction framework based on Maximum Likelihood Expectation Maximization (MLEM) with a distance-driven ray model further enhances resolution and noise robustness. Long-range supersampling additionally mitigates pixel defects and spectral inhomogeneities inherent to high-Z detectors. Phantom studies demonstrate substantial resolution improvement and image uniformity. In comparison with a clinical mammography system, the method reveals sharper detail and more homogeneous contrast at comparable or reduced dose. The resolution gain also reduces the need for geometric magnification, enabling smaller and more cost-effective PCDs. These results establish mechanically supersampled radiography as a clinically viable approach for micron-scale imaging, with strong potential for digital mammography and other high-resolution applications and with scan times compatible with clinical workflow.