A machine learning model in predicting hemodynamically significant coronary artery disease: A prospective cohort...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Coronary artery disease (CAD) is a leading cause of death and incurs significant healthcare costs.
• Existing noninvasive diagnostic tools for CAD have limitations.
• There is a need for improved methods to predict significant CAD.

Purpose of the Study:

• To evaluate the efficacy of machine learning (ML) in predicting hemodynamically significant CAD.
• To utilize routine demographic, clinical, and laboratory data for ML-based prediction.
• To compare ML model accuracy with current noninvasive diagnostic modalities.

Main Methods:

• A prospective cohort study of 185 patients undergoing cardiac catheterization was conducted.
• A random forest model was developed using 18 points of clinical data, including demographics, comorbidities, risk factors, and lab results.
• Model performance was assessed using the area under the receiver operating characteristic curve.

Main Results:

• The machine learning model achieved a sensitivity of 81% ± 7.8% and specificity of 61% ± 14.4% in predicting hemodynamically significant CAD.
• The model also predicted 90-day major adverse cardiovascular and renal events (MACREs) with a sensitivity of 57.13% ± 18.70% and specificity of 44.61% ± 14.39%.

Conclusions:

• Machine learning models can effectively predict hemodynamically significant CAD.
• The accuracy of these ML models approaches that of current noninvasive functional tests.
• Routine clinical data is sufficient for developing predictive models for significant CAD.