Automated Detection of Hypertension Disease Using Machine Learning and Signal Processing-Based Methods

Abstract

Hypertension is a critical health issue and an important area of research because of its high pervasiveness and a remarkable risk factor for cardiovascular and cerebrovascular disease. However, it is a silent killer in all respects. Hardly any side effect can be found in its beginning period until an extreme Medical emergency like heart attack, stroke, or chronic kidney disease. Since individuals are unaware of hypertension, the identification is possible through measurement only. The detection of hypertension at the beginning stage can protect from serious health issues. Furthermore, the hypertension diagnosis by measuring blood pressure may not reflect any severe complication caused due to high blood pressure. Alternatively, automated machine learning and signal processing-based methods require to detect hypertension and its complication (syncope, stroke, and myocardial infraction) from the direct ECG signal. However, the ECG signal is non-stationary, and experts may commit mistakes in observation. As a result, the delay in the treatment of hypertension can be life threatening. Therefore, we have developed the automated detection algorithm for HPT-influenced electrocardiogram (ECG) signal using an optimal filter bank and machine learning. A total of six sub-bands were produced from each ECG signal using a filter bank. In addition, we have extracted the various linear and nonlinear features for all six sub-bands. Subsequently, a ten-fold cross-validation technique was employed for the k-nearest neighbor (KNN) classifier to classify the ECG signals. As a result, the proposed model has achieved a classification accuracy of 98.4%. Hence, the proposed work classifies hypertension from ECG signals in myocardial infarction, stroke, syncope, and low-risk hypertension. Moreover, we can install the proposed algorithm on a personal computer and diagnose the HPT-associated disease from an ECG signal.