Validation of Machine-Learning Angiography-Derived Physiological Pattern of Coronary Artery Disease
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
The classification of physiological patterns of coronary artery disease (CAD) is crucial for clinical decision-making, significantly affecting the planning and success of percutaneous coronary interventions (PCI).
Objectives
This study aimed to develop a novel index to reliably interpret and classify physiological CAD patterns based on virtual pullbacks from single-view Murray’s law-based quantitative flow ratio (μFR) analysis.
Methods
The pullback pressure gradient index (PPGi) was used to classify CAD patterns, with a cut-off value of PPGi=0.78 distinguishing focal from diffuse and non-focal disease. A machine learning method using penalized logistic regression models was proposed to assess CAD patterns. Scores derived from multivariate functional principal component analysis (MFPCA) of μFR and quantitative coronary analysis improved model performance. Expert panel interpretations served as the reference.
Results
A total of 179 vessels (134 patients) underwent classification. The PPGi cut-off of 0.78 achieved 70% accuracy (95% CI: 0.70 to 0.71) for focal vs. diffuse and 77% accuracy (95% CI: 0.76 to 0.77) for focal vs. non-focal classification. The penalized logistic regression model, including PPGi as a feature, provided superior accuracy: 95% (95% CI: 0.94 to 0.95) for focal vs. diffuse and 84% (95% CI: 0.83 to 0.84) for focal vs. non-focal classification. Positive predictive value (PPV) and negative predictive value (NPV) were 95% and 92% (focal vs. diffuse) and 84% each (focal vs. non-focal). Overall, the penalized logistic regression model successfully identified more focal lesions and ensured fewer diffuse or non-focal lesions were misclassified.
Conclusions
The machine learning method with penalized logistic regression outperformed the PPGi cut-off values, providing robust and generalizable classification across different study populations.
