Reconstruction of 12-lead ECG with an Optimized LSTM Neural Network

Abstract

Cardiac-related diseases take the lives of more than 17 million lives around the world. Among the most cardiac vital signs used in clinics and hospitals are Standard 12-lead electrocardiogram (ECG) signals. Nonetheless, due to many electrodes having be connected in order to acquire all 12 signals, Standard (ECG) remains highly sensitive to ambient and body noise which strongly degrades the signal quality. To meet the need for a reliable ECG system with fewer signal interferences. A reduction of recorded leads should be performed to improve the signal quality. For this purpose, An ECG reconstruction technique using an optimized Long-Short-Term Memory (LSTM) model employing the genetic algorithm will be introduced in this study. The genetic algorithm will be used to optimize the hyper-parameters of the LSTM network such as the activation function, number of units, etc. Subsequently, the optimized LSTM network will be trained using a subset of the PTB diagnostic ECG database. The evaluation of the LSTM model will be carried out using performance measures, in particular root mean square error (RMSE) and the correlation coefficient (CC). The results will be detailed and compared with traditional linear regression and standard LSTM deep learning model.