AI CAC scoring on low-dose CT matches semi-automated scoring and speeds workflow in lung screening

Fully automated AI-based coronary artery calcification quantification on low-dose CT for lung cancer screening showed excellent agreement with semi-automated scoring (ICC 0.96) and high sensitivity (0.97) for excluding CAC, without extra reading time.

• Categorical agreement between AI and semi-automated scoring was almost perfect (weighted κ 0.92), surpassing agreement between semi-automated and visual assessment (κ 0.84 and 0.68).
• Semi-automated scoring required significantly more processing time (102.0 ± 95.7 s) than visual assessment (15.5–24.2 s; p<0.001), while AI required no additional reading time.
• AI-based CAC quantification showed high sensitivity (0.97) and specificity (0.91) for excluding CAC.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.