Coronary artery calcification assessment in National Lung Screening Trial CT images (DeepCAC2)

A research team from institutions including Harvard and the Technical University of Munich has published DeepCAC2, a breakthrough AI resource for cardiovascular risk assessment. The system uses a fully automated deep learning pipeline, trained on expert-annotated cardiac CT data, to detect and quantify coronary artery calcification (CAC) from standard low-dose chest CT scans. This is significant because CAC is a powerful predictor of heart attacks and strokes, but it's rarely analyzed in routine thoracic imaging due to the need for specialized protocols and labor-intensive manual annotation. DeepCAC2 processed the entire imaging cohort of the National Lung Screening Trial—127,776 CT scans from 26,228 individuals—to create standardized CAC segmentations and risk estimates for each scan.

The resulting DeepCAC2 dataset is being released as a transparent, large-scale public resource to accelerate medical AI research. It includes DICOM-compatible segmentation objects and structured metadata to support reproducible analysis in cardiovascular risk assessment, opportunistic screening, and imaging biomarker development. The team has already launched a public dashboard allowing visual inspection of a 200-patient subset, and will release the full pipeline as a DICOM-compatible Docker container. This approach demonstrates how AI can extract additional, life-saving diagnostic information from existing medical imaging data collected for other purposes (like lung cancer screening), creating what's known as 'opportunistic screening' without requiring additional scans, radiation exposure, or healthcare costs.