Poincaré Image Analysis of Short-Term Electrocardiogram for Detecting Atrial Fibrillation

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Atrial fibrillation (AF) poses significant risks for stroke and heart failure.
• Early AF detection is vital but complicated by asymptomatic cases and similar ectopic beats.
• Existing screening methods face challenges with short-term electrocardiogram (ECG) data and differentiating arrhythmias.

Purpose of the Study:

• To develop and validate a novel automated screening model for atrial fibrillation (AF).
• To enhance AF detection accuracy using Poincaré image-domain features from short-term ECGs.
• To differentiate AF from premature atrial contractions (PACs) and premature ventricular contractions (PVCs) in ECG signals.

Main Methods:

• A hybrid model combining a radial basis function-based support vector machine (SVM) classifier with rule-based criteria.
• Extraction and reduction of 84 Poincaré image features to four key features using minimum redundancy maximum relevance (mRMR).
• Integration of P-wave information and dRR distribution patterns for improved arrhythmia discrimination.

Main Results:

• The model achieved high accuracy, ranging from 96.35% to 99.40% across 5-fold cross-validation.
• Leave-one-dataset-out validation yielded accuracies between 96.48% and 99.33%.
• Demonstrated balanced sensitivity and specificity across eight diverse datasets comprising over 200,000 ECG segments.

Conclusions:

• The developed Poincaré image-based model offers a robust and accurate method for automated AF screening.
• Its high performance across varied datasets indicates suitability for real-world clinical applications.
• The model shows promise for computerized assessment of short-term ECGs, facilitating timely AF diagnosis and management.