DASMcC: Data Augmented SMOTE Multi-Class Classifier for Prediction of Cardiovascular Diseases Using Time Series Features

Abstract

One of the leading causes of mortality worldwide is cardiovascular disease (CVD). Electrocardiography (ECG) is a noninvasive and cost-effective tool to diagnose the heart’s health. This study presents a multi-class classifier for the prediction of four different types of Cardiovascular Diseases, i.e., Myocardial Infarction, Hypertrophy, Conduction Disturbances, and ST-T abnormality using 12-lead ECG. There are four key steps involved in the presented work: data preprocessing, feature extraction, data preparation, and augmentation, and modelling for multi-class CVD classification. The sixteen-time domain augmented features are used to train the classifier. The work is divided into three parts: extracting the features from raw 12-lead ECG signals, data preparation and augmentation, and training, testing, and validating the classifier. A comparative study of the performance of five different classifiers (i.e., Random Forest (RF), K Nearest Neighbors (KNN), Gradient Boost, Adda Boost, and XG Boost has also been presented. Accuracy, precision, recall, and F1 scores are used for performance evaluation. Further, the Receiver Operating Curve (ROC) is traced, and the Area Under the Curve (AUC) is calculated to ensure the unbiased performance of the classifier. The application of the proposed classifier in the Smart Healthcare framework has also been discussed.

Keywords

• Cardiovascular disease (CVD)
• PTB-XL data
• machine learning
• smart healthcare
• ECG
• heart failure