Monday, December 2 | 9:00 a.m.-9:30 a.m. | M2-SPCA-5 | Learning Center
Using AI with chest CT scans improves quantification of coronary artery calcium (CAC) volume, Texas researchers have found.
In this Monday morning poster session, Fernando Uliana Kay, MD, PhD, of UT Southwestern Medical Center in Dallas will share results from a study he and colleagues conducted that implemented AI software for this purpose in September 2022. Their goal was to set detection thresholds for actionable coronary artery calcium and to compare AI results with radiologist CAC scoring.
The study included 333 patients who underwent noncontrast chest CT imaging and a cardiac CT exam with Agatston CAC scoring. The investigators calculated CAC volume using AI Rad Companion (Siemens Healthineers) and compared these results to the radiologists' scores.
The group found that median AI CAC volume was 21.0 mm³ and median CAC score was 30.7 AU. It also reported that the area under the receiver operating curve for detecting a CAC score equal to or less than 100 AU was 0.97 for the AI algorithm and 0.92 for the radiologist readers. With an optimal threshold of 70.1 mm³ CAC volume, the AI algorithm showed 100% sensitivity and 88% specificity for identifying high-risk patients, compared to 71% sensitivity and 100% specificity shown by the radiologist readers.
"[AI-based coronary artery calcium volume quantification] correlates strongly with the Agatston method and could be a precise, accessible metric for opportunistic cardiovascular risk screening," Kay and colleagues concluded.