Title: Hybrid Deep Learning Assisted Neonatal Bradycardia Detection Using Ensemble Features of ECG Recording
Authors: Mr. D.Balaji and Dr.G.Geetha
Volume: 9
Issue: 6
Pages: 130-134
Publication Date: 2025/06/28
Abstract:
Neonatal bradycardia occurs when the resting heart rates are under 80 beats per minute. The bradycardia mechanisms includes abnormalities in sinus nodeand irregularities of atrioventricular transmission. This paper proposed Hybrid Dense Attention Cascaded Long Short Term Memory (HybDcL)to detect the Neonatal Bradycardia. In this work, we take input data from Preterm Infant Cardio-respiratory Signals (PICS) databasefor the normal and bradycardia ECG signals. Initially, the input ECG signal is pre-processed by the three stage Discrete Wavelet Coefficient Based Inverse Transform (DwavIT) approach. This method is performed to improve the signal quality by suppressing the noises. Then the pre-processed signals are provided for feature extractor byQ-tune empirical variational component analysis (QEmVaC) model. In the feature extractor model, for obtaining a set of feature vectorsthe Q-tune wavelet transform (QWT), Variational mode decomposition (VMD), empirical mode decomposition (EMD) and Independent component analysis (ICA) are utilized to extract the features. The MAE value of the proposed method is 0.008%, while comparing with other existing methods our proposed method yields better performance. This model is more effective using the hybrid deep learning methodologies for gaining enhanced prediction results.