Coronary artery decision algorithm trained by two-step machine learning algorithm.

Area of Science:

• Cardiovascular medicine
• Biomedical engineering
• Artificial intelligence in healthcare

Background:

• Accurate estimation of fractional flow reserve (FFR) is crucial for diagnosing coronary artery disease.
• Current computational fluid dynamics (CFD) methods for FFR calculation are complex and time-consuming.
• Machine learning (ML) offers a potential alternative for efficient and accurate FFR assessment.

Purpose of the Study:

• To introduce and evaluate a novel two-step ML algorithm for estimating coronary artery FFR and decision (DEC).
• To compare the accuracy of the proposed ML-based FFR estimation against traditional CFD methods.
• To investigate the role of flow characteristics and biometric features in improving DEC accuracy.

Main Methods:

• Development of a two-step ML algorithm incorporating Gaussian process regression and support vector machine.
• Training the first step using a synthetic model derived from CFD analysis.
• Utilizing patient-specific flow characteristics and biometric data for the second step.

Main Results:

• The first step of the ML algorithm achieved FFR estimation accuracy comparable to CFD-based methods.
• The second step demonstrated improved accuracy in the decision (DEC) analysis.
• Analysis revealed that flow characteristics and biometric features contributed to the enhanced DEC accuracy.

Conclusions:

• The proposed two-step ML algorithm provides a viable and accurate alternative for FFR estimation.
• The integration of patient-specific data significantly enhances the accuracy of clinical decision-making (DEC) in coronary artery disease.
• This ML approach holds promise for more efficient and precise cardiovascular diagnostics.