Predicting functional results of percutaneous coronary intervention using machine learning modelling.

Area of Science:

• Cardiovascular Medicine
• Interventional Cardiology
• Medical Artificial Intelligence

Background:

• Post-percutaneous coronary intervention (PCI) quantitative flow ratio (μFR) is crucial for long-term clinical outcomes.
• Accurate pre-procedural prediction of post-PCI μFR can optimize procedural planning and improve patient prognosis.

Purpose of the Study:

• Develop and validate machine-learning (ML) models to predict continuous post-PCI μFR.
• Utilize pre-procedural data (angiographic, physiological, clinical) for prediction.
• Assess ML model ability to classify PCI outcomes as optimal (μFR ≥ 0.91) or sub-optimal (μFR < 0.91).

Main Methods:

• Trained four ML models using pre-PCI variables.
• Employed internal bootstrap validation (1000 iterations) to select the best model based on the lowest root mean square error (RMSE).
• Used predicted μFR values for classifying PCI outcomes.

Main Results:

• Achieved high accuracy in continuous post-PCI μFR prediction (RMSE 0.036) using only pre-procedural data.
• Demonstrated clinically meaningful performance in classifying PCI outcomes (accuracy 0.72, AUC 0.72).
• High sensitivity (0.90) enables reliable upfront identification of vessels likely to achieve optimal physiology.

Conclusions:

• ML models accurately predict post-PCI μFR and reliably distinguish optimal from sub-optimal outcomes pre-intervention.
• This predictive capability supports personalized PCI planning and enhances strategy selection.
• The approach promises to improve patient outcomes by enabling better procedural decisions.