Diagnosis of coronary artery disease using data mining techniques based on symptoms and ECG features

The most common heart disease is Coronary artery disease (CAD). CAD is one of the main causes of heart attacks and deaths across the globe. Early diagnosis of this disease is therefore, of great importance. A large number of methods have thus far been devised for diagnosing CAD. Most of these techniques have been conducted on the basis of the Irvine dataset (University of California), which not only has a limited number of features but is also full of missing values and thus lacks reliability. The present study was designed to seek a new set, free from missing values, comprising features such as the functional class, dyspnea, Q wave, ST elevation, ST depression, and T inversion. Information was gathered from Shaheed Rajaei Cardiovascular, Medical and Research Center, between Fall 2011 and Winter 2012. The dataset included 303 patients and SMO, Naïve Bayes, and a proposed ensemble algorithm were used to conduct the analyses. The accuracies of the different algorithms on the dataset were calculated using tenfold cross-validation. In the best case, i.e. using the presented ensemble algorithm, up to 88.5% accuracy was achieved. Finally, several rules and relevant features to CAD, which were absent in previous studies, were extracted.