Heartbeat Classification and Arrhythmia Detection Using a Multi-Model Deep-Learning Technique.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Cardiac arrhythmias are life-threatening, necessitating prompt and accurate diagnosis.
• Electrocardiograms (ECGs) are the primary diagnostic tool, but manual analysis is time-consuming and inefficient.
• Current machine-learning methods for ECG analysis suffer from long training times and require manual feature selection.

Purpose of the Study:

• To develop a novel deep-learning framework for automated cardiac arrhythmia detection.
• To overcome the limitations of existing machine-learning approaches, specifically long training times and manual feature selection.
• To improve the efficiency and accuracy of arrhythmia classification from ECG data.

Main Methods:

• A novel deep-learning framework integrating various networks by stacking similar layers was proposed.
• The framework was tested on two publicly available datasets for the recognition of five micro-classes of arrhythmias.
• Performance was evaluated using sensitivity, specificity, positive predictive value, and accuracy.

Main Results:

• The proposed framework achieved high performance metrics: 98.37% sensitivity, 99.59% specificity, 98.41% positive predictive value, and 99.35% accuracy.
• The approach demonstrated superior performance compared to state-of-the-art methods across all evaluated metrics.
• Significant reduction in computational cost was observed compared to existing methods.

Conclusions:

• The novel deep-learning framework offers an efficient and accurate solution for cardiac arrhythmia detection from ECGs.
• This approach addresses key limitations of current machine-learning techniques, paving the way for improved clinical diagnostics.
• The framework's superior performance and reduced computational cost suggest its potential for widespread adoption in clinical practice.