Classıfıcation of sleep apnea by using wavelet transform and artificial neural networks

Abstract

This paper describes a new method to classify sleep apnea syndrome (SAS) by using wavelet transforms and an artificial neural network (ANN). The network was trained and tested for different momentum coefficients. The abdominal respiration signals are separated into spectral components by using multi-resolution wavelet transforms. These spectral components are applied to the inputs of the artificial neural network. Then the neural network was configured to give three outputs to classify the SAS situation of the patient.

The apnea can be broadly classified into three types: obstructive sleep apnea (OSA), central sleep apnea (CSA) and mixed sleep apnea (MSA). During OSA, the airway is blocked while respiratory efforts continue. During CSA the airway is open, however, there are no respiratory efforts. In this paper we aim to classify sleep apnea in one of three basic types: obstructive, central and mixed.

Introduction

Sleep apnea syndrome (SAS) is characterized by apne periods; the repeated temporary cessation of breathing to the lungs during sleep (Guilleminault, van den Hoed, & Mitler, 1978). Clinically, apnea is defined as the complete cessation of breathing for more than 10 s in adults (American Academy of Sleep Medicine, 1999). In general, when a person is wide awake, except for momentary closures during swallowing and speaking, the upper airway remains open permitting airflow to the lungs (Shelton & Bosma, 1962). During sleep, however, the throat lumen may be physically obstructed from time to time (Remmers, deGroot, Sauderland, & Anch, 1978) and give rise to SAS. Three types of SAS is as follows:

Obstructive sleep apnea (OSA): This is the more frequent pattern, characterized by the presence of thoracic effort for continuing breathing while air flow completely stops.

Central sleep apnea (CSA): This is characterized by a complete cessation of both respiratory movements and airflow during, at least, 10 s.

Mixed sleep apnea (MSA): This pattern is a combination of the previous two, defined by a central respiratory pause followed, in a relatively short interval of time, by an obstructive ventilatory effort.

In fact the mechanisms underlying these different types of apnea are interacting (White, 2000).

The prevalence of these kinds of sleep disorders, commonly the SAS, is approximately 2% in women and 4% in men whose age ranging from 30 to 60 years (Young, Palta, & Dempsey, 1993). It has been reported that in individuals with SAS, throughout the night there can be 5–15 episodes per hour in mild cases, and more than 30 episodes per hour in severe cases (Kryger, 2000).

Beside genetics (Guilleminault et al., 1995, Mathur and Douglas, 1995, Redline et al., 1995), physical obstruction of the airway can result from a variety or combination of anatomical factors (Schwab, Pasirstein, & Pierson, 2003) such as enlarged tonsils (Shintani, Asakura, & Kataura, 1996), enlarged uvula (Hamans, Van Marck, & De Backer, 2000), increased tongue size (Do, Ferreyra, Healy, & Davidson, 2000) and abnormal craniofacial morphology (Cakirer, Hans, & Graham, 2001).

These respiratory disturbances may lead to hypoxia and hypercapnia, which can trigger arousal from sleep by increasing ventilatory drive (Ayas et al., 2000, Benlloch et al., 1995, Gleeson et al., 1990). As a result of such sleep disruption, excessive daytime sleepiness is the most common presenting complaint (Dement, Carskadon, & Richardson, 1978). Other symptoms of sleep apnea include loud snoring, not feeling well-rested in the mornings, chronic fatigue (Kales, Cadieux, & Bixler, 1985), falling asleep at inappropriate times of the day, morning headaches, recent weight gain, limited attention span, memory loss, poor judgment, personality changes and lethargy (Bassiri & Guilleminault, 2000). These symptoms can significantly decrease the quality of life and increase the risk of accidents (Findley et al., 1988, George et al., 1987).

Unfortunately, because of person’s unawareness, sleep apnea may go undiagnosed for years (Ball et al., 1997, Kryger et al., 1996). The case is often recognized via patient’s spouse or a roommate or a family member who has witnessed the apnea periods alternating with arousals and accompanied by loud snoring (Hoffstein, 2000, Stradling and Crosby, 1990). Although snoring is the most common complaint associated with sleep apnea the reverse hypothesis is not true (Bearpark et al., 1995, Hoffstein, 2000). Therefore, patients reporting symptoms of SAS should be consulted to a sleep center for an overnight sleep study that is usually achieved through polysomnograph, an integrated device comprising EEG, EMG, EOG, ECG, oxygen saturation, airflow through the mouth and nose, thoracic and abdominal respiration measurement units (Kryger, 2000).

From overnight sleep studies a respiratory disturbance index (RDI) and an apnea-hyponea index (AHI) which holds the sum of apneas, hypopneas and respiratory arousals per hour during sleep have been standardized (Chervin, 2000). While the RDI value is used to diagnose and grade the severity of the sleep apnea (Kryger, 2000), the AHI is used to assess the severity of apnea according to the Chicago criteria: AHI<5, normal; AHI =5–15, mild; AHI=15–30, moderate; and AHI >30, severe (The Report of an American Academy of Sleep Medicine Task Force, 1999). The risk factors for SAS include upper airway abnormalities (Schwab et al., 2003), male gender (Pillar et al., 2000, Mohsenin, 2001, Malhotra et al., 2002), alcohol use (Issa and Sullivan, 1982, Scrima et al., 1982, Taasan et al., 1981), snoring (Hoffstein, 2000), obesity (Strobel & Rosen, 1996), overweight and a neck circumference which ranges more than 17 in. in men or 16 in. in women (Davies et al., 1992, Mortimore et al., 1998).

The most common method for the diagnosis of the SAS is based on nocturnal polysomnography. It consists of a polygraphic recording during sleep of the electrophysiological and pneumological signals. Thus, this method uses the electroencephalogram (EEG), electrocardiogram (ECG), electro-oculogram (EOG), electromyogram (EMG), airflow, thoracic breathing movements, abdominal respiration, the position of the body during sleep, and arterial oxygen saturation signals (American Academy of Sleep Medicine, 1999, Penze et al., 2002).

The treatment of SAS can range from conservative methods, such as oral appliances (Lowe, 2000) or continuous positive airway pressure (CPAP) (Grunstein & Sullivan, 2000), to more radical approaches, such as surgical removal of anatomic obstructions (Riley, Powell, Li, & Guilleminault, 2000).

Early recognition and classification is the most important step in treating SAS. Automated systems for monitoring, recognition and classification of SAS, particularly using ANNs, have been introduced (Akin, Kurt, Sezgin, & Bayram, 2007). Concerning implementation ease and fast learning with smaller training sets, multilayer perceptron neural networks (MLPNNs) with the most frequently used backpropogatin algorithm can be employed.

The abdominal effort signals are separated into spectral components by using multi-resolution wavelet transforms. These spectral components are applied to the inputs of the artificial neural network. Then the neural network was configured to give three outputs to classify the SAS situation of the patient: OSA, CSA and MSA. An apneic event is shown in Fig. 1.