Bi-LSTM based Interdependent Prediction of Physiological Signals

Abstract

Two of the most essential physiological signals obtained during patient care are the photoplethysmogram (PPG) and electrocardiogram (ECG). Due to recent technological advancements the correlation between the two has come to light. The significance of each signal warrants a solution to predict one when the other is absent. Also, the inexpensive and non-invasive approach of PPG provides a cheaper and comfortable way of monitoring instead of installing ECG. Thus, This study proposes a Bi-LSTM model to predict the two physiological signals. The model was successful in predicting long term data after filtering and aligning the signals efficiently with an MSE value of 0.092 and 0.065 for ECG and PPG respectively.