International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024
p-ISSN: 2395-0072
www.irjet.net
ARRHYTHMIA CLASSIFICATION USING 2D CNN K. Sai Pushvan1, G. Aiswarya2 , Ch. Manohar Reddy3, Md. Naseera4 & Ms.Ramya Asalatha Busi5 1234Undergraduate students, Department of Computer Science and Technology,
Vasireddy Venkatadri Institute of Technology, Guntur, Andhra Pradesh
5Assistant Professor, Department of Computer Science and Technology,
Vasireddy Venkatadri Institute of Technology, Guntur, Andhra Pradesh ---------------------------------------------------------------------***--------------------------------------------------------------------CNN 2D and MobileNet architectures underscores the focus Abstract - This project focuses on developing and
on capturing spatial dependencies within ECG signals while optimizing computational efficiency, thus bridging the gap between technological advancements and cardiac health diagnosis.
evaluating an arrhythmia classification system using CNNs, emphasizing smaller input sizes for computational efficiency. It transforms ECG signals from the MIT-BIH Arrhythmia database into images for CNN-based classification. The simplified CNN classifier prioritizes practicality and efficiency, exploring trade-offs between performance and resources. It investigates optimal input sizes balancing computational efficiency and accuracy. Anticipated outcomes include advancing arrhythmia detection with insights into CNNs' efficacy with smaller inputs and offering practical recommendations for efficient classification models. A refined 2D CNN, integrating MobileNet models, is tailored for specific arrhythmia classes, optimizing efficiency and interpretability for accurate categorization. Classes include Atrial Fibrillation, Murmur, Ventricular Tachycardia, PVC, Supraventricular Tachycardia, and Normal rhythms.
1.1 Classification of Arrhythmias:
Key Words: Convolutional Neural Network, CNN, CNN 2D, Mobile Net Image Classifier, Electrocardiogram (ECG), Arrhythmia
1.INTRODUCTION This project aims to develop an efficient arrhythmia classification system using Convolutional Neural Networks (CNNs), particularly focusing on exploring smaller input sizes to enhance computational efficiency. By simplifying the image classifier and grouping classes based on transformed ECG signals from the MIT-BIH and PTB Arrhythmia databases, the objective is to achieve high accuracy in classifying arrhythmia while minimizing computational resources. The initiative addresses the challenge of accurately classifying arrhythmia, a critical cardiac abnormality, by investigating the feasibility of a simplified CNN 2D classifier with smaller input sizes. The ultimate goal is to contribute to more effective arrhythmia diagnosis by leveraging advanced computational methods. The project is driven by a commitment to improving healthcare practices and advancing precise and accessible arrhythmia detection. By transforming ECG signals into image data and implementing a simplified CNN 2D image classifier, the research aims to provide practical insights into the efficacy of CNNs with smaller input sizes and offer recommendations for efficient arrhythmia classification models. The use of
© 2024, IRJET
|
Impact Factor value: 8.226
2. Literature Review: Machine learning is a powerful tool used in various aspects of our lives, from analysing images and predicting future trends to improving recommendations and enhancing healthcare, banking, defense, education, and robotics. Researchers across different fields rely on machine learning algorithms to make tasks smarter and more efficient. In this research, we employed different machine learning techniques like - Fractional Fourier Transform, KNN Algorithm, Logistic Regression, Dual-Channel 1D-CNN. [1]. In this model, the authors have introduced a ThreeHeartbeat multilead ECG Recognition method for Arrhythmia Classification with the utilization of 1D-CNN
|
ISO 9001:2008 Certified Journal
|
Page 1004