Diagnosis of coronary arteries stenosis using data mining

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Cardiovascular diseases, particularly Coronary Artery Disease (CAD), are a leading cause of mortality worldwide.
• Early diagnosis of CAD is crucial for effective treatment and reducing deaths.
• Current diagnostic methods are limited, especially in cases with subtle or absent ECG abnormalities, and few studies analyze stenosis in individual coronary vessels.

Purpose of the Study:

• To develop a method for identifying the specific lesioned coronary artery in patients with Coronary Artery Disease (CAD) when ECG changes are not significant.
• To utilize risk factors, physical examination, and paraclinical data for diagnosing coronary artery stenosis.
• To evaluate the effectiveness of machine learning algorithms in pinpointing stenosed vessels based on comprehensive patient data.

Main Methods:

• A novel dataset of 303 patients from Tehran's Shaheed Rajaei Cardiovascular, Medical and Research Centre was utilized, featuring comprehensive data without missing values.
• Key features included Function Class, Dyspnoea, Q Wave, ST Elevation, ST Depression, and T inversion.
• Data was processed using C4.5, Naïve Bayes, and k-nearest neighbour (KNN) algorithms, with accuracy assessed via tenfold cross-validation.

Main Results:

• The study achieved diagnostic accuracies of 74.20 ± 5.51% for Left Anterior Descending (LAD) artery stenosis, 63.76 ± 9.73% for Left Circumflex artery, and 68.33 ± 6.90% for Right Coronary Artery using the best-performing method.
• Identification of effective features contributing to the stenosis diagnosis for each specific vessel was accomplished.
• Machine learning models demonstrated capability in differentiating stenosis across individual coronary arteries.

Conclusions:

• The developed approach shows promise for diagnosing coronary artery stenosis, particularly in challenging cases lacking clear ECG indicators.
• Risk factors and paraclinical data, when analyzed with machine learning, can effectively identify specific lesioned coronary arteries.
• This method offers a valuable tool for improving the early and precise diagnosis of Coronary Artery Disease (CAD).