Automated detection of sleep apnea from electrocardiogram signals using nonlinear parameters

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Sleep apnea is a prevalent sleep disorder causing significant daytime impairment.
• Electrocardiography (ECG) signals from patients exhibit non-stationary and nonlinear characteristics.
• Current monitoring methods like polysomnography (PSG) can be cumbersome for patients.

Purpose of the Study:

• To develop a non-invasive method for detecting sleep apnea using ECG signals.
• To classify ECG segments into apnea, hypopnea, and normal breathing states.
• To evaluate the accuracy and efficiency of an artificial neural network (ANN) classifier for sleep apnea detection.

Main Methods:

• Nonlinear signal processing parameters (approximate entropy, fractal dimension, etc.) were extracted from ECG signals.
• An artificial neural network (ANN) classifier was trained using these extracted physiological features.
• Recurrence plots were uniquely generated for different breathing states (normal, hypopnea, apnea).

Main Results:

• The ANN classifier achieved an average classification accuracy of 90%.
• Specificity reached 100%, and sensitivity was 95% for sleep apnea detection.
• The proposed method demonstrated potential for accurate sleep apnea classification.

Conclusions:

• ECG-based nonlinear analysis combined with ANN offers a highly accurate method for sleep apnea detection.
• The developed system is a less cumbersome and potentially more accessible alternative to traditional polysomnography (PSG).
• This approach can improve patient monitoring and reduce the burden of sleep disorder diagnosis.