Overview on prediction, detection, and classification of atrial fibrillation using wavelets and AI on ECG

Area of Science:

• Cardiology and Biomedical Engineering
• Signal Processing and Machine Learning

Background:

• Atrial fibrillation (AF) is a prevalent cardiac arrhythmia linked to increased stroke and heart attack risks.
• Timely and automated detection of AF is critical for effective patient management, especially given its often asymptomatic and paroxysmal nature.

Purpose of the Study:

• To review recent advancements in predicting, detecting, and classifying atrial fibrillation episodes.
• To focus on methodologies employing wavelet transforms (WT) and artificial intelligence (AI) over the past decade.

Main Methods:

• Systematic review of 45 articles published in the last decade.
• Analysis of studies utilizing wavelet transforms (WT) combined with artificial intelligence (AI), including deep learning (DL) and conventional machine learning (ML).
• Comparison of WT with Fourier Transform (FT) in signal analysis for AF.

Main Results:

• A significant increase in research on WT and AI for AF detection, classification, and prediction, with 30 of 36 relevant studies published after 2015.
• Wavelet transform combined with AI shows promise for addressing the challenges of early AF detection.
• Deep learning and conventional machine learning approaches are both actively explored within this domain.

Conclusions:

• The integration of wavelet transforms and AI represents a rapidly advancing and promising area for improving atrial fibrillation management.
• Future research holds significant potential for developing more effective AF prediction, detection, and classification tools.
• The growing body of literature underscores the clinical importance and technical viability of AI-driven AF analysis.