A Data Mining Model for Coronary Artery Disease Detection Using Noninvasive Clinical Parameters

Coronary Artery Disease (CAD) is one of the major cause of death as well as disability worldwide. Suspected cases of CAD go through invasive and non-invasive tests to get CAD detected. Angiography is a noninvasive method for detection. Not only it is costly, time consuming and risky, but it also needs technical expertise and not suitable for the screening of large population. Hence researchers are looking for better alternatives using non-invasive clinical tests. Objectives: To construct Artificial Neural Network based model for CAD identification and adjudging its accuracy with respect to other models. Method: Data mining techniques are being employed to identify CAD cases based on non-invasive clinical tests. Early detection of disease is necessary in order to avoid the risk being exaggerated further. Benchmark Cleave land heart disease data is collected from UCI machine repository and Probabilistic Neural Network is employed and trained and tested using non-invasive clinical parameters. Finding: Neural Network based model is presented that uses non-invasive clinical parameters of the subjects to model CAD cases and achieves the diagnosis accuracy of 96% and misclassification error rate of 4%.The models’ performance is also compared with other classifiers such as RBF Network, AD Tree. Improvement: Neural network based model showed the superiority over other methods in terms of accuracy. Results are promising and reproducible and therefore the model can be valuable adjunct tool in clinical practices.