Vital signs monitoring device with machine learning

• Main Author: Yang, Mingqi
• Other Authors: Muhammad Faeyz Karim
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/145161
• Institution: Nanyang Technological University

According to the WHO, cardiovascular diseases (CVDs) are the number one cause of death globally, taking an estimated 17.9 million lives each year, representing 31% of all global deaths. Of these deaths, 85% are due to heart attack and stroke . Conventional approaches used in hospitals for detecting heart abnormalities has mostly relied on observation of signal outputs from Electrocardiogram (ECG) electrodes placed on the body of patients. People with diagnosed hypertension problem commonly use automated blood pressure monitors at home to monitor heart rates. The most widely accepted product for domestic blood pressure and heart rate monitoring is the Omron Blood Pressure monitors with an estimated 50% market share of total . The accuracy of both ECG electrodes and Omron Blood Pressure Monitor have been clinically validated. However, these two ways of heart rate monitoring have great limitations when it comes to convenience and portability. More and more wearable devices such as the Apple Watch, on the other hand, are beginning to support functions designed for heart rate monitoring. The accuracy of wearable heart rate monitoring devices, however, has not yet been validated. The outcome of this project is expected to give insights into a study that compares the accuracy of the Apple Watch series 2 build-in heart rate function to the Omron HEM- 6161 Wrist Blood Pressure Monitor. Heart rates will be taken on the Omron device and Apple Watch simultaneously. Readings will also be taken under various settings to evaluate how different postures will affect the accuracy of the Apple Watch. Readings will then be analysed on its distribution. After knowing how the data is distributed it will be compared by MATLAB machine learning models. In MATLAB, the KNN and SVM Classifier will be applied to show which classifier has the better accuracy rate. The data will also be further processed by trained neural network encoders to eliminate noises. Based on the results of the experiment, insights on the effects of applying machine learning on vital signs will be given as well as recommendation of future research.