An Acquisition-Based Approach for High-Fidelity Infilling of Photon-Counting X-ray Detector Pixel Gaps
Photon-counting detectors (PCDs) represent a technological advancement in X-ray CT imaging, bringing increased spatial resolution and spectral information to imaging in medical and industrial fields. Despite their potential, a critical issue arises from dead pixel gaps between detector tiles, leading to image artifacts and a reliance on imperfect computational infilling methods. Addressing this challenge, we introduced an acquisition-based solution utilizing a custom-built micro-CT system capable of laterally shifting the PCD during scans. We acquired laterally offset projection data to fill pixel gaps in unshifted projection data. The approach's inherent robustness not only bypasses the need for traditional inpainting or interpolation algorithms but also maintains high quantitative fidelity. Our method shows a marked decrease in low-frequency ring artifacts, surpassing conventional methods in performance. With the potential to be integrated into existing systems or combined with emerging deep learning techniques, our contribution opens promising prospects for future research and applications. Ultimately, this work underscores a significant step toward enhancing image quality and diagnostic precision in X-ray CT imaging, offering a practical and innovative solution to a longstanding problem.
