Mammography-based deep learning model for coronary artery calcification
Authors' conclusion
Mammography-based deep learning models showed promising results for predicting CAC, performing comparably to conventional risk models.
Study context
- Population
- Korean women who underwent mammography and coronary artery calcium (CAC) CTtraining 5235, validation 1208
- Comparison
- Mammography-based convolutional neural network (CNN) model vs Framingham risk score for predicting detectable CAC
- Outcome
- Prediction of detectable CAC (sensitivity, specificity, AUROC, accuracy)
Methodological detailsdeep learning model development and validation, Mean age 52.6 (±10.2) years; non-zero CAC 46.8%
- Deep learning model development and validation
- Mean age 52.6 (±10.2) years; non-zero CAC 46.8%
Performance
As reported in the abstract
Prediction of detectable CAC (validation dataset)
| Test or model | AUROC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|
| CNN model | 0.761 (95% CI, 0.742-0.780) | 0.764 (95% CI, 0.667-0.830) | 0.652 (95% CI, 0.614-0.710) | 70.8% (95% CI, 68.8-72.4) |
| CNN model + age and menopausal status | 0.776 (95% CI, 0.762-0.790) | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract |
| Framingham risk score | 0.736 (95% CI, 0.712-0.761) | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract |
AUROC
CNN model0.761 (95% CI, 0.742-0.780)
CNN model + age and menopausal status0.776 (95% CI, 0.762-0.790)
Framingham risk score0.736 (95% CI, 0.712-0.761)
Sensitivity
CNN model0.764 (95% CI, 0.667-0.830)
CNN model + age and menopausal status— not stated in the abstract
Framingham risk score— not stated in the abstract
Specificity
CNN model0.652 (95% CI, 0.614-0.710)
CNN model + age and menopausal status— not stated in the abstract
Framingham risk score— not stated in the abstract
Accuracy
CNN model70.8% (95% CI, 68.8-72.4)
CNN model + age and menopausal status— not stated in the abstract
Framingham risk score— not stated in the abstract
— not stated in the abstract
Original publication
Mammography-based deep learning model for coronary artery calcification. Eur Heart J Cardiovasc Imaging. Published 21 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →