https://www.mdu.se/

This article demonstrates the feasibility of the convolutional neural network (CNN) and pulse transit time (PTT)-based approach in estimating the systolic blood pressure (SBP) and diastolic blood pressure (DBP). Electrocardiogram (ECG) and photoplethysmography (PPG) signals were employed to calculate the PTT, which is the time delay between the R-wave peak Rof ECG, and specific points of the PPG waveforms. Then, the Blood pressure (BP), which is inversely related to PTT was estimated. A total of 22 patients with available ECG, PPG and SBP data were selected from the Medical Information Mart for Intensive Care (MIMIC III) dataset to validate the proposed model. A window of five minutes of recoding was chosen for each patient. Duration of each cardiac cycle was around 0.6 s, centred at R-peaks and sampled at 125 Hz. A CNN-based model was developed with four convolutional layers. The results showed that the average root mean square error (RMSE) of 5.42 mmHg and 7.81 mmHg were achieved for SBP and DBP, respectively.