Feasibility and limitations of deep learning–based coronary calcium scoring in PET-CT: a comparison with coronary calcium score CT

Oh et al. investigated AI-based coronary artery calcium scoring (CACS) on PET-CT using **AVIEW CAC software**, comparing its results with conventional ECG-gated CaCT. In this study, 210 patients underwent both CaCT and PET-CT on the same day, and CACS was computed three ways: manually from CaCT, manually from PET-CT(reader), and automatically using PET-CT(AI) with AVIEW CAC. Results indicated a strong correlation between PET-CT(reader) and PET-CT(AI) (ICC: 0.958), showing consistent performance within PET-CT methods. CaCT and PET-CT(reader) also showed high agreement (ICC: 0.911). However, CaCT and PET-CT(AI) had only moderate correlation (ICC: 0.842), highlighting a tendency for PET-CT(AI) to underestimate CACS, with a false-negative rate of 37.3%. Misclassification rates were higher in older male patients. This suggests that while AVIEW CAC is effective for PET-CT scoring, CaCT remains the more precise method for CACS, with PET-CT(AI) serving as a supplementary tool.