Representative multimodal CAD studies included in the main text (n=15).
Representative multimodal CAD studies included in the main text (n=15).
| First Author (Year)ref | Modalities | Fusion Strategy | Algorithm(s) | Validation | AUC/C-index | Key Contribution |
|---|---|---|---|---|---|---|
| Motwani et al. (2017)6 | CCTA + clinical | Late | XGBoost | External | AUC = 0.79 | Benchmark ML model for 5-year CAD risk |
| Betancur et al. (2018)7 | SPECT MPI + clinical | Late | Deep CNN | External | AUC = 0.81 | AI-enhanced perfusion imaging fusion |
| Sun et al. (2021)8 | PRS + clinical | Late | Cox regression | Internal | C-index = 0.722 | PRS-enhanced model with public health simulation |
| Lin et al. (2022)9 | CCTA + PET perfusion | Early | Deep learning | Internal | AUC = 0.84 | Dual-modality imaging fusion for ischemia prediction |
| King et al. (2022)10 | PRS + clinical | Late | Cox regression | Internal | HR stratification | Genetic + clinical fusion with risk stratification |
| Vassy et al. (2023)11 | PRS + clinical | Late | Cox regression | Internal | NRI = 0.38% (men) | Multi-ancestry PRS fusion with modest gain |
| Li et al. (2024)12 | EHR time series | Early | Transformer | Real-world | AUC = 0.87 | Temporal modeling of structured clinical data |
| Zhan et al. (2024)13 | PCAT + FAI + clinical | Late | ML + logistic regression | Internal | AUC = 0.83 / 0.71 | Segmental PCAT fusion with inflammation profiling |
| Pezel et al. (2025)14 | CCTA + CMR + clinical + ECG | Early | LASSO + XGBoost | External | AUC = 0.86 | Rich multimodal fusion with strong external validation |
| Zhang et al. (2025)15 | Face + tongue + waveform + lab | Early | Transformer + adaptive weighting | External | Accuracy = 85% | Non-traditional multimodal fusion with novel architecture |
| Gabriel et al. (2025)16 | CAC + ECG + lab + clinical | Late | XGBoost + SHAP | External | AUC = 0.883 | Multi-source structured data fusion for 10-year MACE |
| Zou et al. (2025)17 | PCAT radiomics + CT-FFR + clinical | Early | LASSO + LDA | Internal | AUC = 0.886 | Lesion-specific imaging fusion with clinical enhancement |
AI, artificial intelligence; AUC, area under the curve; CAC, coronary artery calcium; CAD, coronary artery disease; CCTA, coronary computed tomography angiography; C-index, concordance index; CMR, cardiac magnetic resonance; CNN, convolutional neural network; CT-FFR, computed tomography-derived fractional flow reserve; ECG, electrocardiogram; EHR, electronic health record; FAI, fat attenuation index; HR, hazard ratio; LASSO, Least Absolute Shrinkage and Selection Operator; LDA, linear discriminant analysis; MACE, major adverse cardiovascular events; ML, machine learning; MPI, myocardial perfusion imaging; NRI, net reclassification improvement; PCAT, pericoronary adipose tissue; PET, positron emission tomography; PRS, polygenic risk score; SHAP, SHapley Additive exPlanations; SPECT, single-photon emission computed tomography; XGBoost, eXtreme Gradient Boosting.