Predictive value of automated coronary calcium scoring in lung cancer screening with low dose computed tomography

Coronary artery calcium (CAC) is a known risk factor for cardiovascular events, but not yet routinely evaluated in Low Dose Computed Tomography (LDCT) screening. The present analysis compared the accuracy of a new automated CAC quantification versus prior manual quantification on baseline LDCT screening images as predictors of all-cause mortality at 12 years. The study included 1129 volunteers of the Multicentric Italian Lung Detection (MILD) trial who underwent a baseline LDCT scan from September 2005 to September 2006, already analyzed in a previous paper on CAC scoring. The initial manual CAC (mCAC) had been scored by one operator using a dedicated software, while the new automated CAC (aCAC) score was measured by a fully automated artificial intelligence software. All CAC scores were stratified in four categories: 0, 0.1- 19.9, 20-399, and [≥] 400. The study showed a high correlation between aCAC and mCAC scores, with an Intraclass Correlation Coefficient of 0.887. Of 613 negative mCAC score, 87.6% had aCAC score >0, and 14.0% >20. A CAC score >20 revealed a higher risk of 12-year all-cause mortality both with mCAC and aCAC. Focusing on the 535 individuals with false negative mCAC score, aCAC identified a subset of volunteers with a significantly poorer survival of 86% (aCAC 20-399, p=0.0007). CAC quantification could be accurately and safely performed with a fully automated software on baseline LDCT screening images to predict all-cause mortality risk.