Automated OSAHS detection from ECG using temporal convolutional network.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS) affects 1 billion people globally, leading to severe health risks and accidents.
• Current diagnostic methods like polysomnography are expensive and inconvenient, contributing to frequent underdiagnosis.
• There is a critical need for accessible and accurate OSAHS detection tools.

Purpose of the Study:

• To develop an automated method for simultaneous detection of apnea and hypopnea events using a Temporal Convolutional Network (TCN).
• To introduce a novel ECG-TCN model with a linearly scalable attention mechanism for enhanced diagnostic accuracy and reduced computational cost.
• To evaluate the model's performance and generalization capacity for improved OSAHS diagnosis.

Main Methods:

• Development of a novel Temporal Convolutional Network with a Linearly Scalable Attention Mechanism (ECG-TCN).
• Training and validation of the ECG-TCN model using the University College Dublin Sleep Apnea Database.
• Evaluation of model performance based on per-segment classification accuracy and generalization capacity.

Main Results:

• The ECG-TCN model achieved 91.6% accuracy in per-segment classification of apnea and hypopnea events.
• Demonstrated superior performance compared to traditional classification models.
• Exhibited high generalization capacity, indicating robustness across different datasets.

Conclusions:

• The ECG-TCN model presents a novel and effective approach for simultaneous apnea and hypopnea event detection.
• This AI-driven method offers a cost-effective and accurate alternative to conventional OSAHS diagnostic tools.
• The study highlights the potential of ECG-TCN to improve early detection and management of Obstructive Sleep Apnea Hypopnea Syndrome.