Noninvasive Arterial Blood Pressure Estimation using ABPNet and VITAL-ECG

Arterial Blood Pressure (ABP) is an important physiological parameter that should be properly monitored for the purposes of prevention and detection of cardiovascular diseases, which represent one of the leading causes of death in the world. Currently, the most common adopted noninvasive blood pressure measurement system is sphygmomanometer, which works by inflating and deflating a cuff around the arm. This work presents ABPNet, a new prediction technique, based on a multilayer perceptron (MLP), which uses ECG and PPG to estimate both systolic and diastolic blood pressure. To train the neural network, signals are gathered from the Physionet MIMIC database. The proposed architecture performances are evaluated w.r.t. both the invasive blood pressure signal and the noninvasive sphygmomanometer measurements. The experimental results are quite promising; they are compliant with the ANSI/AAMI/ ISO 81060-2:2013 for sphygmomanometer certification because the network predicted values are within +/-5 mmHg w.r.t. real invasive measurements, as imposed by the legislation. Finally, it is shown how ABPNet can be used to improve the VITAL-ECG, a wearable device designed to acquire vital parameters, such as electrocardiographic (ECG) and photoplethysmographic (PLETH/PPG) signals; indeed, by embedding the ABPNet neural network, VITAL-ECG can be upgraded to estimate, also, ABP. As a consequence, the device could be used to fight cardiovascular diseases and prevent their dangerous effects.