Systematic comparison of different algorithms for apnoea detection based on electrocardiogram recordings

Area of Science:

• Cardiology
• Sleep Medicine
• Biomedical Engineering

Background:

• Sleep apnoea is a prevalent disorder often diagnosed via costly sleep laboratory studies.
• Diagnosis relies on identifying cyclic variations in heart rate and electrocardiogram (ECG) waveforms.

Purpose of the Study:

• To evaluate the feasibility of diagnosing sleep apnoea using only overnight ECG recordings.
• To assess the capability of ECG analysis to detect sleep apnoea minute-by-minute.

Main Methods:

• An international challenge provided a training set (35 recordings) for algorithm development and a test set (35 recordings) for independent scoring.
• Thirteen algorithms were developed and compared, focusing on frequency-domain features and respiratory influences on ECG.
• Expert assessment using additional physiological signals served as the gold standard for apnoea identification.

Main Results:

• Four algorithms achieved 100% accuracy in distinguishing between patients with and without sleep apnoea in the first part of the study.
• Two algorithms demonstrated over 90% accuracy in detecting sleep apnoea minute-by-minute in the second part.
• The most effective algorithms utilized frequency-domain heart rate variability and respiration-induced ECG waveform changes.

Conclusions:

• Overnight ECG recordings hold significant potential for accurate and inexpensive sleep apnoea diagnosis.
• ECG-based automated diagnosis could enable remote and cost-effective screening for sleep apnoea in patient homes.
• The study highlights the efficacy of advanced signal processing techniques applied to ECG for diagnosing sleep-related breathing disorders.