Wearable Multi-wavelength Photoplethysmography System for Monitoring Vital Signs with Machine Learning Application Nguyen Mai Hoang Long

Abstract

Remote monitoring is a future trend for healthcare providers in our era. Taking advantage of the latest achievements in computing, network communications, data science and breakthroughs in machine learning, health monitoring becomes more feasible, comfortable, speedy, and convenient than ever. Photoplethysmography (PPG) is an optical-based technique that allows us to measure the volume changes of flow in arteries. It’s also able to estimate the ratio of compositions fluid by adopting multi-wavelengths PPG. This promising technique therefore appears on most smart devices such as smartphones and smartwatches. In this dissertation, we proposed a monitoring system using PPG with a novel design on the optical sensor for monitoring vital signs including heart rate (HR), blood pressure (BP) and glucose concentration level (GCL). Our design aims to monitor the heart signals that occur on the radial artery of human wrist. This position is well-known location for health diagnostics in traditional medicines in a widespread of many Asia countries such as China, Korea, Japan, Vietnam and India. Inspiration of the valueless experiences in oriental science and the advance of sciences, we would like to develop a wrist-wearable system using PPG for human heath monitoring. The development process of our system has solved two problems which were addressed in this dissertation. First, the natural appearance of two types of signals with their phases countered each other when using PPG technique for measurement on the wrist. We named them In-Phase and Invert-Phase PPG signals and found the origin of these signal formations. Second, PPG technique is extreme sensitivity to motion artifacts (MAs). How activities in our daily life affect the PPG signals at the wrist will be also explored. In addition, our vision on the power of Generative Artifact Intelligence (Gen AI), we proposed a Generative Adversarial Network (GAN)-based deep learning system allows us to reconstruct the corrupted PPG by MAs which promised an improvement in health monitoring system in daily life activities.