Study of Echocardiogram Parameters from PPG Signal Using Self-Organized Operational Map-based Network.

Abstract

Two crucial echocardiography parameters -left ventricular ejection fraction (LVEF) and myocardial performance index (MPI) are referred by clinicians to diagnose heart health. Here, an attempt was made to study the possibility of predicting values of LVEF and MPI from Photoplethysmography (PPG) signals. The classification of patients based on the LVEF and MPI values was also evaluated. After PPG signal feature extraction, the Dual Attention-Self Organised Operational Map-LSTM-Conv Network (DASLCN) was used to find the necessary results. Self-organized operational maps (SOOM) helped map the features before sending them to BiLSTM and 1D CNN layers. The results obtained were regression=0.86 with error% of 5.32±8.9 for MPI and accuracy=0.90 & sensitivity=0.89 for LVEF. This technique might help diagnose heart conditions from PPG signals without routine echocardiography.Clinical relevance- PPG is an easy cost-effective portable technique. Whereas, clinical echocardiography is possible only in specialized hospitals. Thus, exploring PPG signals to predict LVEF and MPI values were tried here. This study has been made on whether the grouping of patients based on the range of LVEF and MPI values was possible or not. Newly designed DASLCN helped to perform regression and classification in the same network.