EffNet: an efficient one-dimensional convolutional neural networks for efficient classification of long-term ECG...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiology

Background:

• Early diagnosis of cardiovascular disease (CVD), particularly cardiac arrhythmia, is critical for preventing mortality.
• Manual electrocardiogram (ECG) interpretation is resource-intensive and prone to inaccuracies.
• Existing automated ECG classification methods often lack sufficient accuracy and efficiency.

Purpose of the Study:

• To develop an efficient deep learning model for automated classification of five distinct heartbeat categories in ECG signals.
• To enhance ECG datasets using unique PhysioNet databases and data balancing techniques.
• To evaluate the proposed model's performance against established deep learning architectures and machine learning algorithms.

Main Methods:

• A 12-layer deep One-Dimensional Convolutional Neural Network (1D-CNN), named EffNet, was designed for ECG classification.
• A merged dataset was created from five PhysioNet databases, segmented into 10-second ECG fragments.
• The Synthetic Minority Oversampling Technique (SMOTE) was applied for dataset balancing, followed by hyperparameter optimization for the 1D-CNN.

Main Results:

• EffNet achieved superior performance in classifying five distinct ECG categories compared to GoogLeNet, SqueezeNet, and Support Vector Machines (SVM).
• The proposed model demonstrated dominance over existing literature in ECG classification based on key performance metrics.
• Statistical analysis confirmed EffNet's effectiveness in automated cardiac arrhythmia detection.

Conclusions:

• EffNet represents a significant advancement in automated ECG classification, offering high accuracy and efficiency.
• The proposed deep learning approach provides a robust solution for early detection of cardiac arrhythmias.
• This research contributes to improving cardiovascular disease diagnosis and patient outcomes through advanced AI.