Long-term blood pressure prediction with deep recurrent neural networks

Abstract

Existing methods for arterial blood pressure (BP) estimation directly map the input physiological signals to output BP values without explicitly modeling the underlying temporal dependencies in BP dynamics. As a result, these models suffer from accuracy decay over a long time and thus require frequent calibration. In this work, we address this issue by formulating BP estimation as a sequence prediction problem in which both the input and target are temporal sequences. We propose a novel deep recurrent neural network (RNN) consisting of multilayered Long Short-Term Memory (LSTM) networks, which are incorporated with (1) a bidirectional structure to access larger-scale context information of input sequence, and (2) residual connections to allow gradients in deep RNN to propagate more effectively. The proposed deep RNN model was tested on a static BP dataset, and it achieved root mean square error (RMSE) of 3.90 and 2.66 mmHg for systolic BP (SBP) and diastolic BP (DBP) prediction respectively, surpassing the accuracy of traditional BP prediction models. On a multi-day BP dataset, the deep RNN achieved RMSE of 3.84, 5.25, 5.80 and 5.81 mmHg for the 1st day, 2nd day, 4th day and 6th month after the 1st day SBP prediction, and 1.80, 4.78, 5.0, 5.21 mmHg for corresponding DBP prediction, respectively, which outperforms all previous models with notable improvement. The experimental results suggest that modeling the temporal dependencies in BP dynamics significantly improves the longterm BP prediction accuracy.

Wearable Sensing and Telehealth Technology with Potential Applications in the Coronavirus Pandemic

“…Apart from the PTT-based BP estimation method, some other physiological features have been investigated to indicate BP changes, including PPG intensity ratio [109], Womersley number [110], radial electrical bioimpedance [111], modified normalized pulse volume [112], acceleration plethysmography (APG) [113], and diameter of a pulsating blood vessel [114]. Additionally, machine learning has also been applied to BP estimation to develop regression models between signal features and BP, and demonstrating promising estimation accuracy [115][116][117][118][119][120]. However, the interpretation of the datadriven model is nontrivial.…”

Blood pressure prediction by exploiting informative features from ICU patients’ ECG and PPG signals under a heterogeneous ensemble learning framework

“…The following seven measures were used as evaluation metrics to evaluate the performance of the Tables 2,3 wave to travel from one arterial site to another, which is widely used in related work [16], [20], [24], [27], [30], [37]. lcf_timeup capture the microscopic characteristics of PPG waveform changes, sqi_ppg reflect the skewness of the PPG signal, slr is related to peripheral resistance [55], maxv capture the pulsatile changes in blood volume caused by arterial blood flow around the measurement site [55], and rup is the average slope of ascending branch.…”

“…In recent years, there has been some progress in the utilisation of deep learning techniques for the prediction of BP [29][30][31][32][33]. Su et al [30] devised a multi-layer recurrent neural network (RNN) network named DeepRNN and multi-task training strategy is utilized to train a model to predict systolic BP(SBP), diastolic BP(DBP) and mean BP(MBP) simultaneously. Results shows that DeepRNN achieved better results for long-term predictions; Li et al [31] devised a new network called LSTM-CL, in an attempt to use both the user's contextual(profile) data and measure data to make predictions.…”

“…Slapnicar et al [33] designed a complex convolution network with raw PPG signal and its first and second derivatives as inputs, in an attempt to train the prediction model using a large database-MIMICIII. According to the signals used, above works can be grouped into three types of PPG based [10], [13], [15], [21,22], [28], [33], ECG based [14], [23] and PPG and ECG based [12], [16], [19,20], [24], [27], [30], [32]. ECG contains abundant electronic activities and can assist PPG to better predict BP.…”

3D Convolutional Neural Networks for Remote Pulse Rate Measurement and Mapping from Facial Video

“…Contact pulse signals can also be identified using restricted Boltzmann machine and deep belief networks [62]. Deep recurrent neural network architectures, and in particular multilayerl ong short-term memory (LSTM), can be trained to predict arterial blood pressure from contact PPG and electrocardiogram signals [63].…”