Systematic comparison of different algorithms for apnoea detection based on electrocardiogram recordings T. Penzel 1 J. McNames 2 P. de Chazal 3 B. Raymond 4 A. Murray 5 G. Moody 6 1 Department of Respiratory Critical Care Medicine, Hospital of Philipps University, Marburg, Germany 2 Electrical & Computer Engineering, Portland State University, Portland, Oregon, USA 3 Department of Electronic & Electrical Engineering, University College, Dublin, Ireland 4 Department of Respiratory Physiology, Birmingham Heartlands Hospital, UK 5 Regional Medical Physics Department, Freeman Hospital, Newcastle upon Tyne, UK 6 Harvard-MIT Division of Health Sciences & Technology, Cambridge, Massachusetts, USA Abstract—Sleep apnoea is a common disorder that is usually diagnosed through expensive studies conducted in sleep laboratories. Sleep apnoea is accompanied by a characteristic cyclic variation in heart rate or other changes in the waveform of the electrocardiogram (ECG). If sleep apnoea could be diagnosed using only the ECG, it could be possible to diagnose sleep apnoea automatically and inexpensively from ECG recordings acquired in the patient’s home. This study had two parts. The first was to assess the ability of an overnight ECG recording to distinguish between patients with and without apnoea. The second was to assess whether the ECG could detect apnoea during each minute of the recording. An expert, who used additional physiological signals, assessed each of the recordings for apnoea. Research groups were invited to access data via the world-wide web and submit algorithm results to an international challenge linked to a conference. A training set of 35 recordings was made available for algorithm development, and results from a test set of 35 different recordings were made available for independent scoring. Thirteen algorithms were compared. The best algorithms made use of frequency-domain features to estimate changes in heart rate and the effect of respiration on the ECG waveform. Four of these algorithms achieved perfect scores of 100% in the first part of the study, and two achieved an accuracy of over 90% in the second part of the study. Keywords—Heart rate variability, Sleep apnoea, Physiologic signal database, PhysioNet, ECG, Estimated respiration Med. Biol. Eng. Comput., 2002, 40, 402–407 1 Introduction SLEEP APNOEA is a common sleep disorder, with a reported prevalence of 4% in adult men and 2% in adult women (YOUNG et al., 1993). Excessive daytime sleepiness is the most common complaint. An increased risk of accidents and a link between sleep apnoea and arterial hypertension have been proven in recent large-cohort studies (NIETO et al., 2000). Sleep apnoea is now regarded as an important risk factor for the development of cardiovascular diseases (YOUNG et al., 1997). It is successfully treated with home ventilation using nasal continuous positive airway pressure (NCPAP). If patients are treated at an early stage of the disease, their night-time and daytime blood pressure can be lowered, and the adverse health effects can be reduced (DIMSDALE et al., 2000). The traditional methods for assessment of sleep-related breathing disorders are sleep studies (polysomnography), with the recording of electro-encephalography (EEG), electro-oculo- graphy (EOG), electromyography (EMG), electrocardiography (ECG), oronasal airflow, respiratory effort and oxygen satura- tion (AMERICAN ACADEMY OF SLEEP MEDICINE (AASM), 1999). Sleep studies are expensive for patients, because they require overnight evaluation in sleep laboratories, with dedi- cated systems and attending personnel. Limited and less expensive studies are increasingly performed in a home setting. According to the AASM (1999) criteria, patients are diag- nosed with obstructive sleep apnoea if they have five or more apnoea events per hour of sleep during a full night sleep period (AASM, 1999). Each apnoea event is defined as a respiratory pause lasting at least 10 s. During each event, respiration ceases owing to upper-airway obstruction. If the upper-airway obstruction is only partial and flow is lower than 50% of normal, the resulting airflow limitation is called a hypopnoea. A patient with severe sleep apnoea can have up to 600 single apnoea events per night, with a typical duration of 40 s each, and few, if any, sustained periods of normal (unobstructed) breathing. In 1984, cyclical variation in heart rate was described as being characteristic of obstructive sleep apnoea (GUILLEMINAULT et al., 1984). Until now, this ordered variation in heart rate has been applied to the detection of sleep apnoea by only a few groups (PENZEL et al., 1990; HILTON et al., 1999; ROCHE et al., 1999). This paper describes a comparison of different algorithms to detect sleep apnoea from ECG recordings alone. Correspondence should be addressed to Dr Thomas Penzel; email: penzel@mailer.uni-marburg.de Paper received 3 August 2001 and in final form 23 April 2002 MBEC online number: 20023682 # IFMBE: 2002 402 Medical & Biological Engineering & Computing 2002, Vol. 40