[Detection model of atrial fibrillation based on multi-branch and multi-scale convolutional networks]

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Context:

• Atrial fibrillation (AF) detection traditionally relies on time-consuming ECG analysis.
• Early detection and treatment of AF are crucial for patient outcomes.
• Existing machine learning methods often overlook valuable signal features.

Purpose:

• To design an automated AF detection model using an Inception module.
• To process raw, gradient, and frequency ECG signals through multi-branch channels.
• To enhance feature extraction for improved AF identification.

Summary:

• The Inception module model utilizes multi-branch inputs (raw, gradient, frequency ECG signals) for comprehensive AF analysis.
• It extracts QRS complex, RR intervals, P-wave, and f-wave features for robust detection.
• The model demonstrated high inter-patient accuracy (96.89%), sensitivity (97.72%), and specificity (95.88%) on the MIT-BIH AF database.

Impact:

• Enables faster and more accurate early detection of atrial fibrillation.
• Provides a novel deep learning approach by integrating diverse ECG signal features.
• Facilitates automatic AF detection, potentially reducing physician workload and improving patient care.