Delineation of ECG waveform components using encoder-decoder architecture with Postprocess algorithm

With the exponential increase in heart disease cases, it is essential to construct models (algorithms) that can be used to delineate Electrocardiogram (ECG/EKG) wave components. ECG delineation is the process of attaining structural and biological information of every wave component in a signal in terms of finding out the endpoints. This work intends to develop a deep learning model to delineate the P, QRS, T, and U segments and a post-processing algorithm to remove redundant peaks. Thus, the model will help cardiologists achieve better prediction of diseases efficiently. The proposed model is based on encoder-decoder based deep learning (DL) architecture developed to find the onsets and offsets of the various waves occurring in a single heartbeat. Here, a publicly available QT dataset (QTDB) is used as this is the only dataset with U wave annotation. The post-processing method uses the ECG signal's morphological information to remove the effect of incorrectly categorized focal points of various wave components. It identifies the prominent peak by eliminating redundant peaks. The results of the delineation performance are satisfying, with an average sensitivity of 98.45%, and a precision of 91.40% respectively. These findings point to potential uses for wireless and wearable health monitoring technology.