Continuous blood pressure measurement based on photoplethysmography

Blood pressure (BP) is an important physiological sign for one's health system. Blood pressure monitoring and blood pressure measurement are of great significance for the prevention and treatment of cardiovascular diseases (CVD). In this paper, we proposed a high precision, continuous arterial blood pressure measurement method based on machine learning and photoplethysmography (PPG) signal. We obtained more than 5000 periodic signals from the MIMIC II database. We chose 9 morphological characteristics from each PPG period signal. Our method used fewer features than other methods, but our performance was almost the same as other methods. We used five algorithms to build a blood pressure measurement model and compared the performance of each algorithm. The results revealed that the K-nearest-neighbor (KNN) method had the best performance. The KNN method achieved an accuracy of systolic blood pressure with a root mean square error 3.18 (mmHg) and diastolic blood pressure with a root mean square error 2.47 (mmHg). This accuracy was much higher than the Association for the Advancement of Medical Instrumentation (AAMI) required.