Apnoea detection using ECG signal based on machine learning classifiers and its performances

Area of Science:

• Biomedical Engineering
• Data Science
• Sleep Medicine

Background:

• Sleep apnoea is a prevalent sleep disorder characterized by respiratory airway obstruction.
• It is associated with various health conditions including stroke, depression, and neurocognitive disorders.
• Current diagnostic methods can be inconvenient and costly.

Purpose of the Study:

• To develop a cost-effective and convenient method for sleep apnoea detection.
• To explore the efficacy of machine learning techniques using electrocardiography (ECG) signals.
• To simplify the diagnosis of sleep apnoea.

Main Methods:

• Electrocardiography (ECG) signals were utilized as the primary data source.
• ECG signals underwent pre-processing to eliminate noise and artifacts.
• Time and frequency domain features were extracted, including power spectral density using Welch's method.
• Machine learning classifiers (SVM, Decision Tree, k-NN, Random Forest) were employed for sleep apnoea detection.

Main Results:

• The k-nearest Neighbour (k-NN) classifier achieved the highest accuracy of 92.85%.
• This high accuracy was obtained using only 10 extracted features.
• The proposed method demonstrated reliability and effectiveness in sleep apnoea detection.

Conclusions:

• Machine learning techniques applied to ECG signal analysis provide a reliable and promising approach for sleep apnoea detection.
• The proposed method offers a feasible, convenient, and cost-effective diagnostic solution.
• Reduced feature sets can yield high accuracy in sleep apnoea diagnosis.