Predicting Successful Chronic Total Occlusion Crossing With Primary Antegrade Wiring Using Machine Learning
Authors' conclusion
We developed an ML model with 14 features and high predictive capacity for successful primary AW in chronic total occlusion percutaneous coronary intervention.
Study context
- Population
- Primary antegrade wiring cases in CTO PCI, PROGRESS CTO registry, 2012–2023
- Participants
- 12,136 primary antegrade wiring (AW) cases
- Outcome
- Successful CTO crossing with primary antegrade wiring
Methodological detailsregistry, machine learning, multicentre, 48 centres
- Registry
- Machine learning
- Multicentre
- 48 centres
Performance
As reported in the abstract
Prediction of successful primary AW crossing
Primary AW successful
6,965 cases
57.4%
Extreme gradient boosting best of 5 machine learning (ML) models, average AUC
0.775
± 0.010
After hyperparameter tuning, average AUC, training set
0.782
After hyperparameter tuning, average AUC, testing set
0.780
Further findings
Feature importance (SHAP)
- Occlusion length had the most significant impact, followed by blunt/no stump, interventional collaterals, vessel diameter and proximal cap ambiguity
- Aorto-ostial lesion location had the least impact
Original publication
Predicting Successful Chronic Total Occlusion Crossing With Primary Antegrade Wiring Using Machine Learning. JACC Cardiovasc Interv. Published 3 July 2024. doi:10.
Summarised from the published abstract; every number is checked against it