Sparse-view image reconstruction with total-variation minimization applied to sparsely sampled projection data from SiPM-based photon-counting CT

• Published in: Journal of instrumentation Vol. 19; no. 2; p. C02010
• Main Authors: Sato, D., Arimoto, M., Kotoku, J., Kawashima, H., Kobayashi, S., Okumura, K., Murakami, K., Lucyana, F., Tomoda, T., Kataoka, J., Sagisaka, M., Terazawa, S., Shiota, S.
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.02.2024
• Subjects: Algorithms; Computed tomography; Computer simulation; Computerized Tomography (CT) and Computed Radiography (CR); Contrast agents; Image contrast; Image quality; Image reconstruction; Image reconstruction in medical imaging; Iodine; Mathematical models; Medical imaging; Optimization; Photons; X-ray detectors
• ISSN: 1748-0221; 1748-0221
• DOI: 10.1088/1748-0221/19/02/C02010

We constructed a sparse-view computed tomography (CT) system that combines a compressed sensing (CS)-based image-reconstruction algorithm and SiPM-based photon-counting (PC) CT. CS-based image-reconstruction algorithms have been extensively studied for X-ray CT image reconstruction using fewer projections because they are expected to reduce CT imaging time and radiation exposure while maintaining CT image quality. In most previous studies, CS-based image-reconstruction algorithms have been applied to data obtained through numerical simulations or conventional dual-energy CT. However, studies on PC-CT have been scarce. Therefore, we applied a CS-based image-reconstruction algorithm to the projection data obtained using our previously established SiPM-based PC-CT system and evaluated its image quality. We prepared static phantoms equivalent to iodine-containing contrast agents and a mouse model injected with iodine-containing contrast agents as subjects. Thereafter, CT scanning was performed. The obtained projection data were downsampled to simulate a sparse-view situation, and a CS-based image-reconstruction algorithm with total-variation minimization was applied. Consequently, sparse-view CT images were successfully reconstructed, and the image quality was maintained even after downsampling the projection data (downsampling ratios of 1/10 and 1/2 for the rod phantom and mouse model, respectively). Thus, the imaging time and exposure dose could be remarkably reduced (by a factor of 10 or 2), indicating that the CS-based image-reconstruction algorithm is effective for PC-CT.