Rough local transfer function for cardiac disorders detection using heart sounds

Abstract

The heart is truly successor to the brain in being the most significant vital organ in the human body. The heart, being a magnificent pump, has its performance orchestrated via a group of valves and highly sophisticated neural control. While the kinetics of the heart are accompanied by sound production, sound waves produced by the heart are reliable diagnostic tools to check heart activity. Chronologically, several data sets have been put forward to observe heart performance and lead to medical intervention whenever necessary. The heart sounds data set utilized in this article provides researchers with an abundance of sound signals classified using different classification algorithms; neural network, rotation forest and random forest are a few that can be mentioned. This article proposes an approach based on rough sets and a local transfer function classifier for heart valve disease detection. In order to achieve this objective, and to increase the efficiency of the predication model, a Boolean reasoning discretization algorithm is introduced to discrete the heart signal data set, then the rough set reduction technique is applied to find all reducts of the data which contain the minimal subset of attributes that are associated with a class label for classification. Then, the rough sets dependency rules are generated directly from all generated reducts. A rough confusion matrix is used to evaluate the performance of the predicted reducts and classes. Finally, a local transfer function classifier was employed to evaluate the ability of the selected descriptors to discriminate whether they represent healthy or unhealthy. Alternative classifiers were applied to the same data for comparison including Support Vector Machine (SVM), Hidden Naive Bayesian Network (HNB), Bayesian Network (BN), Naive Bayesian Tree (NBT), Decision Tree (DT), Sequential Minimal Optimization (SMO), Decision Table (DT), Rotation Forest (RoF), and Random Forest (RF); however, their performance for the same diagnostic problems was lower than the proposed rough local transfer function.