Development of a Neural Network based model for Non-obtrusive Computation of BP from Photoplethysmograph

Blood pressure (BP) is an important vital sign that needs to be monitored regularly to maintain a healthy life. A normal blood pressure is crucial for life and a consistent variation in that can lead to critical health conditions such as kidney failure, cerebral infarction, hypertension and cardiovascular diseases which can be fatal. Hypertension is one of the major reasons for premature death worldwide. An effective non-obtrusive mechanism for continuous monitoring of BP is necessary for the early detection and prevention of fatal events. In this paper, we present the design, development and validation of a neural network based computational model for continuous BP monitoring using photoplethysmograph (PPG) data from the University of Guilin.

I. Introduction

The normal blood pressure level among adults is 120/80 mmHg. Blood pressure that is too low from this is called hypotension and pressure that is consistently high is called hypertension. A consistent hypertension can lead to many fatal events and diseases such as cardiovascular diseases, stroke, kidney failure and cerebral infarction [1]. According to WHO, hypertension has become a leading cause for premature death worldwide [2]. Hypertension has become a major global problem and this necessitates the development of an efficient live BP monitoring system that can be incorporated into a wearable device to improve self-monitoring among patients. This has given life to a non-obtrusive method of using photo-plethysmograph (PPG) signals for the prediction of BP, which has become a recent research trend.strongly correlate the PPG feature with the corresponding value, although the machine learning models have started showing promising results.