Photoplethysmography-Based Blood Pressure Estimation Using Deep Learning
Weinan Wang, Li Zhu, Fatemeh Marefat, Pedram Mohseni, Kevin L. Kilgore, Laleh Najafizadeh
Abstract
Blood pressure (BP) measurement is an important measure of health status, yet simple and accurate measurement techniques have remained elusive. In this paper, we present a novel transfer learning-based blood pressure estimation algorithm that requires only few seconds of the photoplethysmography (PPG) signal as input. The proposed algorithm utilizes visibility graph to create images embedded with features related to the waveform morphology. The algorithm is evaluated using the data from the Multi-parameter Intelligent Monitoring in Intensive Care (MIMIC) II database. Results show that the difference between the estimated and reference BP for the systolic BP (SBP) and for the diastolic BP (DBP) are -0.080 ± 10.097 mmHg and 0.057 ±4.814 mmHg, respectively, demonstrating the effectiveness of the proposed approach for estimating BP.