Unsupervised ECG Analysis: A Review.

Area of Science:

• Cardiology and Artificial Intelligence
• Biomedical Signal Processing

Background:

• Electrocardiography (ECG) is vital for detecting heart abnormalities, but supervised learning faces data limitations.
• Unsupervised learning, specifically ECG clustering, offers a label-free approach to analyze ECG data.
• ECG clustering reveals patterns related to emotions, mental health, and metabolic levels, extending beyond cardiac analysis.

Approach:

• This study reviews unsupervised ECG clustering techniques, focusing on machine learning and deep learning algorithms from the last decade.
• The review critically examines algorithms, their practical applications, and inherent limitations.
• Future research directions in ECG clustering are also discussed.

Key Points:

• Unsupervised ECG clustering addresses the scarcity of labeled data in supervised learning.
• ECG clustering uncovers novel inter- and intra-individual patterns with broader health implications.
• This technique can mitigate challenges like imbalanced data and enhance biometric systems.

Conclusions:

• A comprehensive review of unsupervised ECG clustering techniques, particularly recent ML/DL methods, is presented.
• The study provides insights into the applications, limitations, and future scope of ECG clustering.
• This resource aids in selecting appropriate algorithms for specific ECG analysis applications.