Feasibility of deep learning-based image reconstruction using TrueFidelity for coronary calcium scoring.
Authors
Affiliations (2)
Affiliations (2)
- Department of Radiology and the Research Institute of Radiological Science, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Department of Integrative Medicine, Yonsei University College of Medicine, Seoul, Republic of Korea.
Abstract
We evaluated the feasibility of deep learning-based image reconstruction (DLIR) using GE TrueFidelity at high strength for coronary calcium scoring, compared with filtered back projection (FBP) and iterative reconstruction (IR; adaptive statistical iterative reconstruction-V with 80% blending), using a cardiac phantom and clinical data. Calcium scores for FBP, IR, and DLIR were obtained with a cardiac phantom. Subsequently, 113 patients who underwent coronary computed tomography angiography were retrospectively evaluated. The Agatston method was applied to all three techniques to obtain the coronary artery calcium score (CACS). Differences, correlations, and concordance in sectional classification of calcium scores were evaluated across reconstruction methods. In the phantom study, calcium scores were comparable across reconstruction methods. In the clinical study, DLIR had significantly lower scores than FBP (p < 0.001), whereas the comparison between IR and DLIR was statistically non-significant (p = 0.064). Weighted kappa coefficients indicated high agreement in total CACS classification across the reconstruction methods. However, Bland-Altman analysis demonstrated non-negligible individual-level disagreement between FBP and DLIR, with a mean difference of 14.5 and 95% limits of agreement from -52.1 to 81.1. This disagreement may be clinically relevant near established CAC risk-category thresholds, particularly 0, 100, and 300. Among 113 patients, 109 were classified into the same CAC-DRS category by both FBP and DLIR. DLIR underclassified four patients, yielding a false-negative rate of 4.65%. In conclusion, DLIR was feasible for CACS measurement and showed high overall concordance with FBP and IR. However, near-perfect correlation should not be interpreted as interchangeability, because DLIR showed systematic underestimation and non-negligible individual-level disagreement from FBP, particularly in patients with low calcium burden or scores near CAC risk-category thresholds.