Respiratory Event Detection During Sleep Using Electrocardiogram and Respiratory Related Signals: Using Polysomnogram...

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Signal Processing

Background:

• Respiratory event detection is crucial for diagnosing sleep-related breathing disorders.
• Current methods often rely on polysomnography (PSG), which can be cumbersome and expensive.
• There is a need for accessible and accurate automated methods for respiratory event detection.

Purpose of the Study:

• To develop and evaluate an automatic algorithm for detecting respiratory events using electrocardiogram (ECG) and respiratory signals.
• To estimate the apnea-hypopnea index (AHI) using the developed algorithm.
• To assess the performance of the algorithm in terms of accuracy and correlation with PSG-based AHI.

Main Methods:

• Utilized handcrafted features from ECG and respiratory signals.
• Applied machine learning algorithms, including Support Vector Machine (SVM), for respiratory event detection.
• Developed and validated the algorithm using data from 1,285 subjects, including PSG and patch-type device recordings.

Main Results:

• The SVM-based algorithm achieved 83% accuracy and 0.53 Cohen's kappa for minute-by-minute respiratory event detection.
• A high correlation coefficient of 0.87 was found between the reference AHI (from PSG) and the estimated AHI.
• Patient classification based on AHI cutoff (15) showed 87% accuracy and 0.72 Cohen's kappa.

Conclusions:

• The proposed automatic algorithm effectively detects respiratory events and estimates AHI using ECG and respiratory signals.
• Simultaneous recording of ECG and respiratory signals enhances performance.
• The use of open datasets can potentially lower the development cost of commercial sleep apnea detection software.