Fractional Flow Reserve-Based Patient Risk Classification.

Area of Science:

• Cardiology
• Medical Informatics
• Biomedical Engineering

Background:

• Cardiovascular diseases (CVDs) are a primary cause of mortality globally, with acute myocardial infarction (AMI) causing millions of deaths annually.
• Timely intervention is crucial to prevent severe complications, disability, and loss of work associated with CVDs.
• Fractional flow reserve (FFR) is a key metric for assessing coronary artery stenosis severity.

Purpose of the Study:

• To develop a novel technique for evaluating patient FFR and assessing mortality risk using demographic and clinical data.
• To implement a machine learning approach for predicting cardiovascular risk.
• To utilize 3D reconstruction for coronary artery stenosis monitoring.

Main Methods:

• A random forest machine learning algorithm was employed to build a classification ensemble model for risk prediction.
• Patients were classified into high-risk (FFR < 0.8) and low-risk (FFR > 0.8) groups based on FFR values.
• A numerical approach involving 3D reconstruction of coronary arteries was used for stenosis monitoring.

Main Results:

• The final classification ensemble achieved an estimated prediction accuracy of 76.21%.
• Mean prediction accuracy ranged from 74.1% to 83.6% across different test sample sizes (5% to 20%).
• The methodology demonstrated satisfying results even with limited data points.

Conclusions:

• The developed machine learning model shows promise for early detection and risk stratification of cardiovascular diseases.
• The combination of machine learning and 3D reconstruction offers a valuable approach for stenosis monitoring.
• Future improvements can be achieved by incorporating more data to explore advanced machine learning algorithms.