Machine Learning to Predict Long-Term Cardiac-Relative Prognosis in Patients With Extra-Cardiac Vascular Disease

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Patients with ischemic stroke (IS), transient ischemic attack (TIA), and peripheral artery disease (PAD) have a higher risk of coronary artery disease.
• Accurate prognostic risk assessment is crucial for identifying high-risk patients who may benefit from intensified treatment.
• Current risk stratification methods for this population remain challenging.

Purpose of the Study:

• To evaluate the feasibility and predictive capability of machine learning (ML) models.
• To predict long-term adverse cardiac outcomes in patients with IS, TIA, and/or PAD.
• To compare ML performance against established clinical risk scores and CCTA metrics.

Main Methods:

• Analysis of 636 patients with a history of IS, TIA, and/or PAD who underwent coronary CT angiography (CCTA).
• Automated feature selection from 35 clinical variables and 34 CCTA metrics for ML model development.
• Clinical outcomes included all-cause mortality (ACM) and major adverse cardiac events (MACE) over a mean follow-up of 3.9 years.

Main Results:

• ML models demonstrated significantly higher predictive accuracy for ACM (AUC 0.92) and MACE (AUC 0.84) compared to modified Duke index, segment stenosis score, segment involvement score, and Framingham risk score.
• Traditional risk scores showed lower predictive values for both ACM and MACE.
• The study identified ML as a superior tool for risk prediction in this patient cohort.

Conclusions:

• Machine learning models show superior capability in predicting long-term adverse cardiac outcomes (ACM and MACE) in patients with IS, TIA, and/or PAD.
• ML surpasses traditional clinical risk scores and CCTA-derived metrics in prognostic accuracy for this high-risk population.
• These findings suggest ML can enhance risk stratification and guide treatment decisions for patients with cerebrovascular and peripheral artery disease.