A novel approach for atrial fibrillation-related obstructive sleep apnea detection using enhanced single-lead...

Area of Science:

• Cardiology
• Sleep Medicine
• Artificial Intelligence

Background:

• Atrial fibrillation (AF) and obstructive sleep apnea (OSA) are common, interrelated conditions increasing cardiovascular risk.
• Current diagnostic methods struggle with concurrent detection, creating a clinical unmet need.
• Existing home sleep apnea tests (HSAT) have limited ECG monitoring, hindering OSA-associated AF diagnosis.

Purpose of the Study:

• To introduce SHHDeepNet, a deep learning framework for simultaneous AF and OSA detection using single-lead ECG.
• To leverage advanced signal processing and deep learning for enhanced diagnostic accuracy.
• To address the clinical need for integrated AF and OSA detection.

Main Methods:

• Developed SHHDeepNet, a deep learning framework utilizing enhanced ECG features.
• Employed Reconstruction Independent Component Analysis (RICA) for ECG signal preprocessing and feature isolation.
• Validated the framework on the Sleep Heart Health Study (SHHS1) and Osteoporotic Fractures in Men (MrOS) databases.

Main Results:

• SHHDeepNet achieved high accuracy in binary and multi-class classification on internal validation (e.g., 98.22% binary accuracy, 98.36% multi-class accuracy with 5-fold CV).
• Leave-one-subject-out cross-validation demonstrated robust performance (e.g., 86.42% binary accuracy).
• External validation on the MrOS database showed promising results (88.51% multi-class accuracy).

Conclusions:

• Simultaneous AF and OSA detection is crucial for comprehensive cardiovascular health evaluation.
• The SHHDeepNet framework shows significant potential for clinical decision support.
• This tool can enhance management strategies and improve patient outcomes for individuals with AF and OSA.