Aortic & peripheral vascular disease
Machine Learning Detects Symptomatic Plaques in Patients With Carotid Atherosclerosis on CT Angiography
Authors' conclusion
This study presented an interpretable machine learning model that accurately identifies symptomatic carotid plaques using computed tomography angiography derived plaque composition features
Study context
- Population
- Patients with symptomatic and asymptomatic carotid atherosclerosis undergoing CT angiography, March 2013–October 2019163 derivation, 106 testing
- Participants
- 268 patients
- Comparison
- CT angiography–based machine learning model vs traditional logit analysis (degree of stenosis, intraplaque hemorrhage, plaque composition)
- Outcome
- Detection of symptomatic carotid plaques (associated with cerebrovascular events): discrimination (AUC) and calibration
Methodological detailsretrospective, single-centre, repeated 10-fold cross-validation, dedicated testing cohort
- Retrospective
- Single-centre
- Repeated 10-fold cross-validation
- Dedicated testing cohort
Performance
As reported in the abstract
AUC for symptomatic plaque
| Test or model | AUC |
|---|---|
| Machine learning | 0.89 |
| Degree of stenosis | 0.51 |
| Intraplaque hemorrhage | 0.69 |
| Plaque composition | 0.78 |
AUC
Machine learning0.89
Degree of stenosis0.51
Intraplaque hemorrhage0.69
Plaque composition0.78
- Machine learning vs Degree of stenosis: P < 0.001
- Machine learning vs Intraplaque hemorrhage: P < 0.001
- Machine learning vs Plaque composition: P < 0.001
Further findings
Plaque components associated with symptomatic status (adjusted)
| OR (95% CI) | ||
|---|---|---|
| Intraplaque hemorrhage to lipid volume ratio ≥50% | 38.5 (10.1–205.1) | |
| Intraplaque hemorrhage volume ≥10% | 18.5 (5.7–69.4) | |
| 0.050.21520Odds ratio (log scale) |
Internal validation
- Comparable performance on internal validation
Original publication
Machine Learning Detects Symptomatic Plaques in Patients With Carotid Atherosclerosis on CT Angiography. Circ Cardiovasc Imaging. Published 18 June 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →