Coronary artery disease (CAD), one of the leading causes of mortality worldwide, necessitates effective risk assessment strategies, with coronary artery calcium (CAC) scoring via computed tomography (CT) being a key method for prevention. Traditional methods, primarily based on UNET architectures implemented on pre-built models, face challenges like the scarcity of annotated CT scans containing CAC and imbalanced datasets, leading to reduced performance in segmentation and scoring tasks. In this study, we address these limitations by introducing DINO-LG, a novel label-guided extension of DINO (self-distillation with no labels) that incorporates targeted augmentation on annotated calcified regions during self-supervised pre-training. Our three-stage pipeline integrates Vision Transformer (ViT-Base/8) feature extraction via DINO-LG trained on 914 CT scans comprising 700 gated and 214 non-gated acquisitions, linear classification to identify calcified slices, and U-NET segmentation for CAC quantification and Agatston scoring. DINO-LG achieved 89% sensitivity and 90% specificity for detecting CAC-containing CT slices, compared to standard DINO’s 79% sensitivity and 77% specificity, reducing false-negative and false-positive rates by 49% and 57% respectively. The integrated system achieves 90% accuracy in CAC risk classification on 45 test patients, outperforming standalone U-NET segmentation (76% accuracy) while processing only the relevant subset of CT slices. This targeted approach enhances CAC scoring accuracy by feeding the UNET model with relevant slices, improving diagnostic precision while lowering healthcare costs by minimizing unnecessary tests and treatments. Graphical
Coronary artery calcium (CAC) scoring on computed tomography (CT) acquired for other indications could provide calcium-burden information without a dedicated scoring examination. We investigate direct, segmentation-free Agatston score prediction from frozen CT patch representations using only volume-level supervision for the scoring head. A lightweight additive head maps attention-selected features to non-negative contributions, combines them within slices, and sums slice scores across each volume. In developmental evaluation, applying the nonlinear mapping before aggregation improved linearly weighted by – over a pooled multilayer perceptron across six encoders. Using spatially adaptive label-guided temperature (SALT) features, held-out was on gated CT and on contrast-enhanced coronary CT angiography (CCTA), with separately fitted heads. On external non-gated CT, a head fitted on 165 patients from the same cohort achieved on 40 held-out patients; patient-level five-fold cross-validation over all 205 patients yielded , indicating preliminary local-adaptation performance. Unchanged-head external gated transfer achieved (). Gated intraclass correlation coefficient on was . Study-level CAC scores served as CCTA reference targets; non-contrast images were not model inputs. SALT uses gated calcium masks during pretraining, not downstream scoring. Its highest-attention patch overlapped calcium in of calcium-bearing slices, but localization rankings did not directly track scoring rankings. These results support direct, spatially decomposable CAC prediction from reusable features; broader validation is needed before clinical use.