A Comprehensive Study on a Deep-Learning-Based Electrocardiography Analysis for Estimating the Apnea-Hypopnea Index

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Sleep apnea diagnosis relies on polysomnography, which is resource-intensive.
• Electrocardiography (ECG) signals contain physiological information relevant to sleep-related respiratory events.
• Accurate estimation of the apnea-hypopnea index (AHI) is critical for sleep apnea severity assessment.

Purpose of the Study:

• To develop and validate a deep-learning system for automatic sleep apnea detection and AHI estimation using single-lead ECG signals.
• To emphasize AHI estimation as a key metric for sleep apnea diagnosis and severity evaluation.
• To provide a non-invasive, accessible tool for sleep apnea screening.

Main Methods:

• A deep-learning model, the deep-shallow fusion network for sleep apnea detection network (DSF-SANet) combined with gated recurrent units (GRUs), was developed.
• The model was trained on 1465 ECG recordings, analyzing signals at 1-minute intervals.
• The system was designed to detect sleep-breathing disturbances and estimate the AHI.

Main Results:

• The system achieved a correlation coefficient of 0.87 with actual AHI values.
• Per-segment classification demonstrated an accuracy of 0.82, an F1 score of 0.71, and an AUC of 0.88.
• The model effectively identified sleep-breathing events and provided reliable AHI estimations.

Conclusions:

• The proposed deep-learning system demonstrates significant potential for automatic sleep apnea detection and AHI estimation using ECG.
• This approach offers a promising, non-invasive method for sleep apnea diagnosis and severity assessment.
• The system can serve as a valuable tool for medical professionals in clinical practice.