Paper Title
DEEP LEARNING-BASED PREMATURE VENTRICULAR CONTRACTION RECOGNITION WITH ATTENTION MECHANISM AND INCEPTION BLOCK

Abstract
Abstract - Premature ventricular contractions (PVCs) are abnormal heartbeats that require accurate recognition for effective diagnosis and treatment. In this study, we suggest a deep learning-based approach for PVC recognition, leveraging the power of attention mechanism and inception block within a convolutional neural network (CNN) model. Our method aims to exploit the informative characteristics of ECG signals and improve classification performance. To capture the distinctive features of PVCs, we extract 10 features, involving three features (QRS width, QR amplitude, and RS amplitude) obtained by morphological approaches and seven statistical features. These features effectively highlight the diverse parts of ECG signals and contribute to accurate discrimination.The CNN model is designed with two paths, each consisting of three convolutional layers. An inception block is incorporated after the first convolutional layer in each path, enabling the extraction of more unique features. The four parallel convolutional layer routes within the inception block further enhance the discrimination capability of the model.Additionally, we employ an attention mechanism to focus on important regions of the input signals and prioritize relevant features during classification. This mechanism improves the model's ability to capture relevant patterns and boosts classification performance. Experimental results on the MIT-BIH arrhythmia database demonstrate the effectiveness of our proposed approach. Comparative analysis with state-of-the-art strategies reveals the superior performance of our method, achieving a higher true positive rate and reduced false positive rate. Keywords - Deep learning, Attention mechanism, Inception block, ECG signal analysis, Premature ventricular contraction.