A Method of Predicting Continuous Blood Pressure by PPG Signal Based on Neural Network

The relationship between blood pressure obtained from photoplethysmography (PPG) signals and pulse duration, is not always linear. To estimate blood pressure more accurately from PPG signals, this paper presents PPGLAB, a convolutional neural network that can predict the continuous ABP waveform of the human body from PPG Signals. In this work, we use the MIMIC-III data set. All signals are sampled at 125Hz and recorded with 8-bit accuracy. By model measurements, the comparison of estimated blood pressure to reference values shows good accuracy, meeting the Class A criteria for BHS on DBP and ABP, and also meeting the national standards of the American Medical Device Association.