Study of a Deep Learning Reconstruction Algorithm for Displaying Small- and Medium-sized Blood Vessels in Upper Abdominal Energy Spectrum CT

Objective: To investigate the effectiveness and clinical value of using a deep learning reconstruction algorithm (DLIR) to display small blood vessels in upper abdominal computed tomography (CT) with an enhanced energy spectrum. Methods: Using three reconstruction algorithms, a retrospective analysis was performed on 28 patients with upper abdominal discomfort who underwent enhanced CT spectrum examination at the West China Hospital of Sichuan University from February 2021 to June 2022. The three reconstruction algorithms were adaptive statistical iterative reconstruction (ASIR-V), DLIR-M, and DLIR-H. Simultaneously, 40 keV and 70 keV single-energy images were extracted using energy spectrum post-processing software, and four groups of images were generated, which were labeled as 40 keV-AV, 40 keV-DL-M, 40 keV-DL-H, and 70 keV-AV, respectively. The CT and standard deviation (SD) values of common hepatic, left gastric, splenic, and superior mesenteric arteries were measured, and the CT and SD values of the vertical spinal muscle at the same level were measured. In addition, the signal-to-noise (SNR) and contrast-to-noise (CNR) ratios of each branch vessel were calculated. Two radiologists provided subjective scores on image noise, image artifacts, target blood vessel contrast, image “waxiness.” and overall image quality. Differences in SNR, CNR, and background noise among the four groups of images were compared using one-way analysis of variance (ANOVA) and paired t-tests. The kappa test was used to compare differences in the consistency of the subjective evaluations. Results: In both objective and subjective evaluations, the SNR, CNR, overall image quality score, and noise of the DL-H images were superior to those of the DL-M images, where the latter in turn were superior to those of the AV images. The SNR, CNR, and image quality score increased and the noise decreased with an increase in DL intensity. In the subjective scores of the four groups of images, the DL-H score was higher than the DL-M score, and the DL-M score was higher than the AV score. Conclusion: The DLIR can improve the display of small- and medium-sized vessels in upper abdomen energy spectrum enhanced CT 40 keV single-energy images. With an increase in intensity, the image quality is improved and noise is reduced. Compared with AV, the DLIR significantly improves the display capabilities of upper abdominal energy spectrum-enhanced CT in the examination of small blood vessels.