Artificial intelligence-assisted longitudinal assessment of coronary artery calcification in the Korean lung cancer screening CT program

Abstract

Purpose

The clinical implications of coronary artery calcification (CAC) growth remain underexplored. This study aims to assess CAC growth and its association with adverse cardiovascular events (ACEs) in individuals undergoing lung cancer screening (LCS) using artificial intelligence (AI)-assisted evaluation.

Methods

We included patients who underwent LCS low-dose chest CT (LDCT) between April 2017 and December 2023 with available follow-up LDCT scans. CAC severity was quantified using AI-based software. CAC growth was defined as incident CAC in those with baseline CAC = 0 or annual progression > 15 % in those with baseline CAC > 0. ACEs were categorized as major or minor events. Associations between CAC growth and ACEs were evaluated using Cox regression models, adjusting for baseline age and CAC status.

Results

Male patients (n = 193; mean age, 61.6 ± 5.2 years) were analyzed. Over a 4-year mean follow-up, 15.5 % experienced ACEs (major event: 4.1 %, minor event: 11.4 %). Greater baseline CAC severity correlated with a higher annual CAC increase (p < 0.001). Age (adjusted hazard ratio (HR) (95 % confidence interval (CI)) = 1.08 (1.00, 1.17); p = 0.041), CAC growth (adjusted HR (95 % CI) = 2.40 (1.13, 5.09); p = 0.023), and moderate to severe baseline CAC (adjusted HR (95 % CI) = 2.86 (1.11, 7.38); p = 0.030) in the three-tiered classification were significantly associated with a higher occurrence of total ACEs.

Conclusions

AI-assisted CAC growth tracking using serial LDCT scans provides prognostic value in LCS populations and may guide risk-based cardiovascular follow-up and prevention strategies in clinical practice.