Development of Semi-Supervised Learning-Based Continuous Blood Pressure Estimation System

Blood pressure is an important measurement for diagnosing diseases. Both healthy and ill patients should monitor blood pressure periodically to prevent sudden occurrence of serious illnesses. However, real-time monitoring has been impossible due to the fact that a cuff was required. To solve this problem, we herein propose a semi-supervised learning-based cuffless blood pressure estimation algorithm. To estimate the blood pressure, ECG and PPG are measured from the arm of the user with a bio-signal measurement system. Multi-scale segmentation is applied to the measured bio-signals to extract quality parameters. The diastolic and systolic blood pressure values are then extracted, and the mean arterial pressure (MAP) is calculated from the two measured parameters. When the MAP value exceeds the normal range, it transmits a message to the specialist or the user via WiFi or LTE.