Blood Pressure Estimation Using Time Domain Features of Oscillometric Waveforms and Recurrent Neural Network

This paper presents a novel method to estimatesystolic blood pressure (SBP) and diastolic blood pressure (DBP)from time domain features extracted from oscillometric waveforms(OWs) using a long short term memory (LSTM) recurrentneural network (RNN). The proposed LSTM-RNN is a powerfultechnique that can effectively discover the latent structure inOW sequences and automatically learn such structures. TheSBP and DBP points are then detected as the cuff pressuresat which OW sequence changes its structure. The proposedRNN-based approach can outperform traditional methods in BPestimation and can be considered as an alternative way, especiallyin DBP estimation which is basically a more challenging problemcompared to SBP estimation.