A Novel Cuff-less Measurement Method for Noninvasive Blood Pressure Prediction using Body Vital Signals

Abstract

Hypertension or the abnormal increase of blood pressure is a chronic disease which can damage the other parts of the body such as the kidneys, heart, and vessels. The high cost of treating the injuries caused by hypertension is undeniable. Various techniques exist for measuring the blood pressure. In recent years, machine learning models became more popular due to being non-invasive and their continuous supervision, remote use, and low cost. Several analyses were performed by the audio signals of cardiac palpitations, electrocardiograms, on photo plethysmogramy on software and hardware platforms. Researchers used machine learning techniques to present the alternative methods for aggressive and costly methods. Among the presented methods, regression algorithms, support vector machine (SVM), and neural network (NN) are highly popular. This study presented a method for analyzing ECG and PPG signals for diagnosing hypertension. The proposed method can improve the classification accuracy regardless of the classification algorithm by providing the combined features. In the conducted evaluation, the neural network algorithm was proposed for the data with continuous label while the C4.5 tree was proposed for the data with discrete label. In addition, the proposed generalized method was provided by calculating the cosine distance and optimizing the genetic algorithm for low data and noise conditions.