A Machine Learning Driven Approach to Quantifying Coronary Artery Tortuosity

Area of Science:

• Cardiovascular Imaging and Diagnostics
• Machine Learning in Medicine
• Biomedical Engineering

Background:

• Coronary artery tortuosity associations with cardiovascular risk factors are not well-established.
• Previous studies were limited by small sample sizes and heuristic metrics.

Purpose of the Study:

• To develop and validate an automated machine learning (ML)-based measure for right coronary artery (RCA) tortuosity.
• To investigate the associations of RCA tortuosity with patient demographics, cardiovascular risk factors, and coronary artery disease (CAD).

Main Methods:

• Developed an ML-enabled pipeline to quantify RCA tortuosity from 38,691 coronary angiograms.
• Validated the ML measure against interventional cardiologist review in 300 cases.
• Utilized regression models to assess associations with age, sex, hypertension, diabetes, hypercholesterolemia, smoking, and CAD.

Main Results:

• The ML tool provided a scalable, continuous measure of RCA tortuosity.
• Higher tortuosity was observed in women, and with hypertension, but inversely with diabetes.
• Increased tortuosity correlated with the presence and severity of CAD, including higher Gensini scores.

Conclusions:

• A scalable, ML-enabled measure of RCA tortuosity was successfully developed.
• Significant associations were found between RCA tortuosity and sex, hypertension, diabetes, and CAD presence/severity.