Preliminary study of image reconstruction from truncated CT data using hybrid image constraints

• Published in: IEEE conference record - Nuclear Science Symposium & Medical Imaging Conference. p. 1
• Main Authors: Zhang, Z., Chen, B., Xia, D., Sidky, E. Y., Pan, X.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26.10.2024
• Subjects: Accuracy; Computed tomography; Microwave integrated circuits; Robustness; Semiconductor detectors; Temperature distribution
• ISSN: 2577-0829
• DOI: 10.1109/NSS/MIC/RTSD57108.2024.10657332

In CT, data with transverse truncation arises when the transverse size of the X-ray beam's field-of-view (FOV) received by the detector is smaller than that of the subject's support. While existing efforts focus primarily on image reconstruction within the FOV, our work delves into achieving accurate reconstructions not only within the FOV but also within a substantially larger region. We investigate the reconstruction problem from truncated data by formulating it as a convex optimization program with hybrid constraints on image-region total-variation (TV) and imageregion L1-norm. To address this, we introduce the TVL1 algorithm, a primal-dual-based method aimed at achieving accurate reconstructions within the subject's support. Simulated and real-data studies, covering various truncation degrees, validate the levels of algorithm's performance and robustness. Our results showcase the TVL1 algorithm's accuracy in reconstructing images within and beyond the FOV for various practical truncation extents. The methodology and TVL1 algorithm developed in this study hold promise for extending to cone-beam CT, thus broadening their applicability in medical imaging.