PPG-based End-to-End Framework for Cuffless Blood Pressure Estimation Using CNN-LSTM

Monitoring blood pressure is an essential factor in the management of hypertension and the occurrence of associated dysthymia. Blood pressure is a vital and robust sign that changes every day. Short time and occasional blood pressure monitoring have commonly been achieved by measurement using classical sphygmomanometers. However, a growing interest is developing a non-invasive blood pressure monitoring system to measure BP persistently without any cuff. As photoplethysmography (PPG) technology has been widely used in wearable sensors, indirect measurement of blood pressure (BP) using the PPG method has yielded significant results. This paper developed PPG’s fiducial point-independent framework for systolic and diastolic BP measurement. This work includes using a Convolutional Neural Network (CNN) for feature extraction followed by a Long Short Term Memory (LSTM) network which automatically estimates blood pressure with remarkable accuracy. Compared to the earlier reported methods, this work has an encouraging result in terms of MAE±STD of 4.19±6.66 for Systolic BP and 2.36 ± 3.42 for Diastolic BP experimented with dataset having 220 patients of MIMIC-II. The obtained results fulfill the criteria of BHS with grade A and the standard of AAMI.