Cognitive Outline Study of Electroencephalogram Signal Patterns towards Monitoring and Analyzing Sleep Behavioural Disorders

Cognitive Outline Study of Electroencephalogram Signal Patterns towards Monitoring and Analyzing Sleep Behavioural Disorders

Abstract. A better consciousness depends on how well your sleep is. Study shows that sleep disorder is become a crucial problem day by day due to work pressure and other different causes. This continuous lack of sleep or uneven sleep pattern can cause to develop various major health issues in humans. These sleep patterns are challenging to detect, analyze and experience as they exhibit at the time when people are in their deep sleep phase. This research work is focused on studying, methods for analyzing and identifying different sleep patterns based on EEG patterns. Every stage of sleep shows a different electroencephalogram pattern. This EEG pattern shows that sleep can be observed as a combination of different cyclic sleep stages. This work shows different sleep stage patterns analysis found in different sleep disorders. Based on these patterns, monitoring of quality sleep can be performed. In this research work, authors have discussed, 10 – 20 electrodes systems to capture brain signals, electroencephalogram signals, and its categorization. This study also represent the detailed description of several methods identified for monitoring sleep behavioural patterns to understand the experimental basis and forming a strong theoretical background towards analyzing these EEG signals.

Keywords: RAM, NRAM, Electroencephalogram, Sleep stages, Quality sleep

• Introduction

In recent years, there has been lots of research performed on the quality of sleep. Research related to sleep depends on how the data is collected. It is unlikely that the data collected from a wearable device or non-wearable device would give completely accurate results[1]. Collecting data from anyone method, there are some raw factors added to the data in the form of subject’s uncomfortableness concerning wearable device or environmental noise factor in terms of non-wearable device. For this type of research data collection is often a challenging task.

The quality of sleep depends on the pattern of its stages. To understand sleep stages better, EEG signals are classified based on their frequency. 8-13, 13-30, 4-8, 0.5-4 Hz is considered as Alpha, Beta, Theta, Delta, respectively. The EEG signals are like a recorder that records electrical impulses that are originating in the brain while two or more brain cells are communicating with each other. These impulses are detected with the help of the 10- 20 system of electrodes that are attached to the scalp. While designing the wearable electrode scalp, it considers a ratio of 10-20 that means every 20% of half scalp must find an electrode to measure electrical impulses. To deal with the sleep disorder, it is necessary to understand its EEG pattern. Sleep is divided into two parts (i) RAM (Rapid eye movement) (ii) NRAM (Non-rapid eye movement). Non-rapid eye movement is further categorized in sleep stages like S0, S1, S2 and S3 where S0 for basic light sleep, S1 and S2 for light sleep and S3 is considered as deep sleep. Every stage of sleep shows a different EEG (electroencephalogram) pattern. To measure sleep quality EEG pattern is considered as an input parameter. In this paper, a cognitive outline of electroencephalogram in the monitoring of sleep behaviour is drawn.

2        Literature Review

According to Kagawa et al. [1], a correct result can be measured by measuring the changes that occur in the sleep stages by the non-contact method. Nine, a 24 GHz microwave radar system, is placed under the mattress to measure the movement of the sleeping person on the bed. For better estimation of sleep stages, a body-movement index is determined. To classify wake and sleep position, LDF (Linear discriminant function) is used and obtained 79.3% accuracy. Radar impulses are the base of calculating the sleep stages (rapid eye movement, non-rapid eye movement). In view of the increasing sleep problem, the author has suggested to the medical department to evaluate and analyses real-time sleep data. Sleep is a cyclic process in which the stages often repeat at an interval of every 90 minutes. The sleep-related study comes under PSG (Polysomnography). Under PSG, various sensors are installed on the body to detect the impulses reading. Still, it is more difficult to sleep to the person who wears the sensory device on the head and other parts of the body, so these readings can’t obtain an accurate result. Because of this author, introduce the unwearable system to measure the precise results. In this paper, polysomnography and radar system accuracy is examined by taking 11 male data whose age was around 22 years. According to Barsocchi et al.[2], Explaining the process of sleep with a wearable device is quite complicated. Often people are uncomfortable due to these devices, which makes it difficult to get good results. In this paper, an unwearable system is introduced, which has 48 FSR (Force sensing resistor) grid. This grid is installed under the mattress, and it is connected to a computer through a data cable. For practical approach author installed a video recorder at night mode to check the ground reality and took almost 2 m height bed and 70 kg subject. Here Multiclass logistic regression is used on WEKA to obtain the result. In this experiment, a large amount of data is collected, so authors have calculated the result by devising the data in 3 parts. In result a better performance is shown with CCI 95.8% , 88.5%, 90.2% on 4.2%, 11.5%, 9.8% error rate respectively. This whole research is based on FSR that measures the pressure points of subjects. According to Adami et al. [3], Measuring the movement of the leg in bed is an important step in experiments conducted to assess the quality of sleep. The author has used the linear classifier to separate the body movement and the leg movement. To measure the leg or body movement, a load cell sensors grid is installed under the mattress. Here the author supports the unobtrusive process to measure sleep quality. The base of periodic leg movement measurement is conducted from the World Association of Sleep Medicine. EMG data of 17 people (29-74 years old, ten male and seven female) is considered for polysomnography. In result, 97% accuracy is obtained while classifying the leg and body movement. The base part of this experiment is to calculate the sleep quality based on PLM (periodic leg movement). According to Tseng et al. [4], Sleep quality has been tested with a Bluetooth-based EEG device. EEG device calculates the brain signals that are produced in the subject’s brain and shares to the cloud storage for further analysis. Because of brain signals have very low voltage, the author used the DAQ100 module to capture the accurate frequency of signals. In this research, the wearable device is used to measure sleep quality. This study explains the sleep disorder ratio of Taiwan society. Based on the Chinese sleep survey, 19.3%, 21.8% of insomnia patients were there in 2013 and 2009, respectively. According to the author, sleep is divided into 2 parts (1) RAM – it takes 25% of total sleep time. (2) NRAM – it is further classified in 4 parts, each part takes 5%, 45%, 45%, 13% respectively. EEG reading for each stage is different, so to identify the sleep quality, the author compares the EEG readings. According to Cheng et al. [5], Sleep plays a very important role in human life. Humans spend 1/3 of their entire life in sleep. Realizing the importance of sleep, it becomes necessary that the quality of sleep is tested on the right parameters. According to the Medical Sleep Organization of Taiwan report, insomnia patients increased from 11.5% to 21.8% in the short period of 2006 to 2009.

Collecting sleep data has always been a challenging task. Most of the devices are very expensive, and the discomfort of the people during the test spoils the sleep data. Here author collected two types of data (1) subjective – this is the bundle of subject’s information like medical history etc. (2) objective – this is recorded information during sleep. The author believes that subjective information is as important as the objective one. Here more than one types of sensors are used to collect data like acceleration sensors are used to measure the body movement. To understand the sleep pattern, it is important to calculate a quality sleep index value also. Author modified the Sleep quality index value (defined by Zeo Inc., ZQ = { ( TOTAL SLEEP DURATION *1 + DEEP SLEEP DURATION * 1.5 + REM SLEEP TIME * 0.5) – (TOTAL AWAKEN TIME * 0.5 + (AWAKENING TIMES / 15)) } * 8.5 ) with some challenges. Some challenges are: subject may be a short sleeper; sleep time varies age to age, so the values of sleep quality index should not be constant. The author modified the formula with some weight value. This new formula has weight value as per requirement. In the end, the author verified his formula with better accuracy than the traditional formula. According to Surantha et al. [6], In the era of growing technology, the Internet of things (IoT) played a miraculous role. In recent years internet of things has changed the way devices work. This is an important change, the way in which machine is interacting with another machine. All the experiments were done in the quality of sleep, collect sleep data manually with the help of sensors. In this type of data collection technique, sometimes some noise signals are added, which somehow spoil the data. Such problems can be dealt with by using IoT technology. The author explained the workflow through IOT technology in sleep assessment. According to Zaffaroni et al.[7]. There are many types of research to understand the quality of sleep. Among them, the experiment from non-wearable devices maintains a distinct identity. In this paper, ultrasonic technology is used to measure sleep quality. Here a special application has been designed that can measure the sonar generated by the movement of a person who is asleep in bed. Such experiments often claim less expensive and good results. It has been tested in the laboratory on 38 subjects to test the veracity of the application. Realizing the importance of sleep in human life, it becomes important that its use should very cheap and of good accuracy. Sonar measurement application provides this opportunity to measure sleep data in a very efficient cost or zero cost. The quality of sleep can be tested in 2 ways (1) through the wearable device (2) through the non-wearable device. The wearable device is capable of producing accurate results, but the discomfort of the tested person often worsens its results. In this experiment, one S+ device, a mobile application that is installed in Samsung S7 is used to measure the sleep data set. There are 62 subjects to collect sleep data. This research is done in Berlin under an advanced sleep research organization. To test the quality of sleep, 6 parameters have been considered, which are as follows –

(1) Total sleep time (TST), (2) Total rapid eye movement time (TREMT), (3) Total deep sleep time (TDS), (4) Total light sleep time (TLST), (5) Wake time, and (6) Sleep latency. In result, accuracy is calculated for each that are 87%, 89%, 84% for wake time, REM, and deep sleep time, respectively. Realizing the importance of quality of sleep, this experiment is the father of revolutionary results. Its cost is the lowest, and the best result is witnessed. Such experiments are often able to make human life comfortable. According to Yan et al.[8], Realizing the importance of sleep, it is important that every phase of sleep is closely examined. This research examines the movement of the body to test each phase of sleep differently. Different sleep stages are very close to each other; testing these stages separately is useful for sleep-related research. Here some sensors are placed under the bed, which generates frequency rays based on the movement of the subject. Brain signals that are measured by the EEG that supports to define sleep stages are separated through a support vector machine. The formula that is used to define sleep stage is:

Deep sleep index (D) = 0.55*number of small movements + 1.05 * medium movements + 1.02 * large movements.

Here D should be less than 0l for deep sleep stage. 0.02936, 0.2385, 0.04826 are the threshold value of small, medium, and large movements. In results, 95.7%, 90.6%, 66.3% accuracy is obtained for deep, light, and wake stage of sleep, respectively. According to Tataraidze et al. [9], There can be many parameters to test the quality of sleep; among them, testing the breathing pattern can be an important component. In this research, each stage of sleep has been understood based on breathing patterns. Quality of sleep focuses on breathing patterns. In this research, a data set has been created by measuring the breathing pattern from sensors with non-wearable devices. Polysomnography has been an important process initially for accurate measurement of sleep quality. But the discomfort of subjects hinders accuracy. Here a sleep disorder breathing algorithm was written under which the results were generated. Algorithm obtained result 93%. According to Dhamchatsoontree et al. [10], A grid of 48 pressure sensors have been installed in the bed to assess the quality of sleep accurately. The author gave a unique name to the force sensor is an i-sleep sensor. Sleep quality index is calculated based on body movement. The machine learning algorithm obtained 86.7% accuracy to detect the sleep posture. The author explained two technology, (1) sensor-based (2) camera-based. The result 95.2%, 94.9%, 94.7%, 95.8% classification accuracy is obtained through random forest, K-nearest neighbour, logistic regression and support vector machine, respectively. The data set contained 30 subjects (15 males and 15 females). According to Liao et al. [11], The quality of sleep is based on different readings for different populations, such as location, gender, and age. In this research, EEG readings on different genders have been examined. EEG patterns have been introduced in the frequency form, and each frequency range is given a name such as 0.5-4, 4-8, 8-12, 12-32Hz to the delta, theta, alpha, beta respectively. These frequency ranges give different readings for different sleep stages and also show some bigger changes in their fluctuation while considering different gender, age, location. Frequencies of beta range and delta range were found to be increased during the non-rapid eye movement. During the experiment, it was found that the delta fluency range was higher in females than males. In the result, it was found that sex differences should be considered as an important parameter for testing the quality of sleep. The author also explained the limitations of the experiment and gave the future aspect of this experiment. In this experiment, the author did not consider high-frequency range Gamma so that results that are obtained can vary while considering Gamma range frequencies. This experiment was based on wearable 10-20 system to detect the brain signals and applied a weighted phase lag index machine learning algorithm to classify the frequency range. According to Ravan et al.[12], Realizing the complexity of the process of sleep quality research, many experiments are associated with low cost, one of them being vagus nerve therapy. This experiment raises a silver lining for epilepsy patients. Through Vagus therapy, it was found that the sleep quality of epilepsy patients can be increased. Here the data of 23 people based on the EEG readings taken from Physiobank has been used. The system is trained with 22 epilepsy patients data that are recorded in a month. Support vector machine algorithm is used to classify the sleep stages. In result, support vector machine gives 90% accuracy. This research proved that vagus nerve therapy could play an important role in quality sleep. According to Cheng et al.[13]. An intelligent system is designed to test the quality of sleep showing the stages of sleep through smartphone application. Oximeter has been used in this experiment, based on which the quality of sleep is measured through a smartphone application which is specially designed for classifying the sleep stages on the basis of heart rate and other parameters. The author explained the architecture of the system which has four steps (1) data collection through oximeter or other devices, (2) Pre-analysis by algorithm, (3) smartphone application role, (4) obtain the result. This research is based on a wearable device but in a less expensive mode. According to Terzano et al.[14], To understand the process of sleep, it is necessary to examine its pattern and its duplication after a time interval. To understand the whole sleep process, it is necessary to examine the changes in EEG readings with respect to the change of physical patterns of subject occurring in a time interval. To measure the quality sleep index author emphasizes to examine the EEG pattern during periodic K- alpha (cyclic alternative pattern). Authors have established a relationship among electrooculogram, electromyogram, electrocardiogram and EEG. Every alternative change in EEG pattern is examined and compared with electrooculogram, electromyogram and electrocardiogram readings. In result, author proved that EEG pattern could calculate the sleep quality index. According to Poryazova et al.[15], Cyclic alternative pattern (CAP) is an important parameter that helps to understand sleep quality. This research examines the CAP pattern of narcolepsy patients. Narcolepsy patients were founded with less movement compared to a healthy human when the CAP patterns were tested. A summary of a systematic review is shown in table 1.

Table 1. Summary of literature review

3        Background

Sleep is a natural process that is a part of an important process like refreshing memory, testing immunity system, recovery of cell health. Sleep plays an important role in a person’s mental stability and physical strength. Sleep at the right time and for a right duration is essential for healthy living. Sleep plays a different role in every age stage, like physical growth and building their immunity system in children while in old sleep is beneficial in the recovery of cell health [11]. Lack of sleep or lack of quality sleep has been a common problem in the last few years. Lack of quality sleep leads to fatal consequences such as physical and mental debility. In view of increasing sleep patients, many sleep study centres have been established in many countries. According to a Taiwan report, there has been an increase in the number of sleep patients from 11.5% to 21.6% in a gap of 3 years, from 2007 to 2009[5]. Such reports are worrisome. Sleep is classified into two stages based on quality of sleep such as (1) RAM (rapid eye movement), (2)  NRAM (non-rapid eye movement). Rapid eye movement is that state of sleep in which a person’s brain functions almost like a waking state. The movement of eyes in this state is exactly like the waking state. A person often dreams in this state of sleep. Moreover, the movement of eyes keeps on according to the film running in the dream. While Non-rapid eye movement is considered a state of complete rest. In this state of sleep, signals of low frequency are found in the brain. NREM is further classified in N1,

N2, N3 where N1 and N2 for light sleep and N3 for deep sleep[16,17].

Fig. 1. Human brain internal structure and their working area

3.1      System and Sensors

Electroencephalogram is a process by which every electrical impulse generated during communication between neurons in the brain can be examined. A human brain is of 1.5 kg and has almost 10 ^ 20 neurons. Every Neuron is connected to the rest of the neurons. In the human brain, each nerve communicates to each other in the form of electrical signals or impulses. These impulses are generated even though in sleeping mode. The voltage of these impulses is very low up to 10µv to 100µv [17] so to examine the impulses it should be amplified. For a better understanding of these EEG impulses, firstly, these are passed through an amplifier then readings are examined for further study. To better understand of EEG impulses, it is important to understand human brain sections shown in Fig.1 and working areas, as shown in Table 2. As per Fig.1, the human brain is divided into ten sections, and each section has its responsibility. At the particular action, the particular section activates, and the nerves in that section communicate through electrical signals. These electrical signals are detected by the EEG device. There is a need for a generalized system to measure electrical impulses that are generated in the human brain.

  Table 2: Human brain section name, symbolic representation, and working area

3.2     10 – 20 System

10-20 system is a generalized system in which every 20% of every half-scalp has an electrode, as shown in Fig.2 that measures every brain impulse[17]. Different size brains are found in different age groups, genders and locations because of that designing of a fixed system is too difficult. To overcome from this problem, a generalized system is designed which do not have any specific electrode for a specific section of the brain, but it have a fixed parameter that is every 20% of every half scalp has an electrode. This system is internationally recognized. With the help of this 10-20 scalp system EEG measures the brain impulses. Every animal has 2 types of electrical signal (1) skin impulse, (2) internal nerve signal[17]. Here sensor A1, A2 measures the skin impulse. So when EEG detects the brain impulse, then additional skin impulse is founded in actual brain impulse due to externally wearable 10-20 system. So to measure the accurate result, skin impulse reading is subtracted from the total reading. On behalf of brain impulse, pattern sleep quality is measured [16, 17]. To better understanding of brain impulses generally, frequency form of signal is used [14,17].On the basis of frequency range, some categories are defined such as : 8-13, 13-30, 4-8, 0.5-4 Hz is considered as Alpha, Beta, Theta, Delta respectively[17] shown in Fig.3.

Fig. 2. 10-20 system

Fig. 3. Electrical impulses categorization based on the frequency range

A generalized result is frequencies of beta range, and delta range was found to be increased during the non-rapid eye movement [11].

4          Discussion and Findings

This paper includes both types of researches (1) through the wearable device (2) through a non-wearable device. Both methods have their own advantages. Some researches support non-wearable device [2, 3, 6, 7 and 9] because Subject feels somewhat uncomfortable with the wearable device when testing sleep quality so that there may be chances of some raw factors are added to the data. However, we cannot ignore the accuracy of collecting data through the wearable device [4, 11, 12 and 13]. Because of the subject’s uncomfortableness with wearable device researchers, collect data from the same subject in 3 – 4 days by which the raw factors in data can easily be measured and removed it [13]. However, there is always some chances to add raw factors in data. Now talk about the non-wearable device, data collection through the non-wearable device is a critical job. There are more chances to add raw factors as noise in data from the environment. There are some processes to remove noise from data, but it depends on the accuracy rate of the method.

5        Conclusion

Sleep is an important aspect of the living world. Literature shows the direct connectivity between many diseases with the quality of sleep. Early diagnosis can be one of the major remedial steps towards preventing disease caused be different sleep disorders. There have been many researches have been performed by different researchers to understand sleep disorder patterns. These research work have been performed by majorly with two types of devices wearable and non-wearable. Each has its own advantages and disadvantages. This research, the paper includes deep study about both wearable and non-wearable devices and sensors in order to understand their applicability and scope of all possible reduction for getting a higher degree of accuracy. Authors have studied and analyzed the available literature work to identify the gaps in the area of sleep pattern identification for detecting sleep disorders. Authors also discussed all sensors and systems used and signal analysis regarding identifying hidden sleep patterns in detail, which can further be helpful for other researchers in this field also.

References

• Masayuki Kagawa, Noriyuki Sasaki, Kazuki Suzumura and Takemi Matsui (2015), “Sleep stage classification by body movement index and respiratory interval indices using multiple radar sensors”, in proc. of 2015 37th Annual international conference of the IEEE engineering in medicine and biology society, DOI: 10.1109/EMBC.2015.7320153, INSPEC accession number: 15584678, 2015
• Paolo Barsocchi, Monica Bianchini, Antonino Crivello, Davide La Rosa, Filippo Palumbo and Franco Scarselli (2016), “An unobtrusive sleep monitoring system for the human sleep behaviour understanding”, in proc. of 2016 7th IEEE international conference on cognitive infocommunications, DOI: 10.1109/CogInfoCom.2016.7804531, INSPEC accession number: 16564236, 2016
• Adriana M. Adami, André G. Adami, Tamara L. Hayes and Zachary T. Beattie (2013), “A system for assessment of limb movements in sleep”, in proc. of 2013 IEEE 15th international conference on e-health networking, applications and services, DOI: 10.1109/HealthCom.20136720712, INSPEC accession number: 14060943, 2013
• Hsien Wei Tseng, Chien-Da Huang, Liang-Yu Yen, Yu-De Liao and Yung-Wen Lee (2016), “A method of measurement sleep quality by using EEG”, in proc. of 2016 IEEE international conference on consumer electronics-Taiwan, DOI: 10.1109/ICCE-TW.2016.7521000, INSPEC accession number: 16177900, 2016
• Shiao-Ping Cheng and Hsing Mei (2012), “A personalised sleep quality assessment mechanism based on sleep pattern analysis”, in proc. of IEEE, DOI: 10.1109/IBICA.2012.55, 2012
• Nico Surantha, Gede Putra Kusuma and Sani M. Isa (2016), “Internet of things for sleep quality monitoring system: a survey”, in proc. of 2016 IEEE 11th international conference on knowledge, information and creativity support systems, DOI: 10.1109/KICSS.2016.7951426, INSPEC accession number: 16969170, 2016
• Alberto Zaffaroni, Sam Coffey, Stephen Dodd, Hannah Kilroy, Graeme Lyon, Damien O’Rourke, Katharina Lederer, Ingo Fiete and Thomas Prenzel (2019), “Sleep staging monitoring based on sonar smartphone technology”, in proc. of 2019 41st Annual international conference the IEEE engineering in medicine and biology society, DOI: 10.1109/EMBC.2019.8857033, INSPEC accession number: 19127036, 2019
• Quya Yan and Chao Xu (2018), “A method of sleeping state recognition based on pressure-body movement- sleeping model”, in proc. of 2018 IEEE 3rd international conference on cloud computing and big data analysis, DOI: 10.1109/ICCCBDA.2018.8386539, INSPEC accession number: 17844178, 2018
• Alexander B, Tataraidze, Lesya N. Anishchenko, Lyudmila S. Korostovtseva, Mikhail V. Bochkarev and Yurii V. Sviryaev (2018), “Non-contact respiratory monitoring of subjects with sleep-disordered breathing”, in proc. of 2018 IEEE international conference “Quality management, transport and information security, information technologies, DOI: 10.1109/ITMQIS.2018.8525001, INSPEC accession number: 18262535, 2018
• Supaki Dhamchatsoontree, Chaiyapat Sirisin, Monika Proncharoensukkul and Konlakorn Wongpatikaseree (2019), “i-Sleep: intelligent sleep detection system for analyzing sleep behavior”, in proc. of 2019 IEEE 4th international conference on information technology, DOI: 10.1109/INCIT.2019.8912047, INSPEC accession number: 19211584, 2019
• Yuanyuan Liao, Guolin Zhou, Jiuxing Liang, Xiangmin Zhang, Xinwen Guo and Yuxi Luo (2019), “Overall population generalities, sex differences, and individual differences in sleep electroencephalography functional connectivity”, in proc. of IEEE, DOI: 10.1109/ACCESS.2019.2950066, Volume 7, 2019
• Maryam Rvan, Jason Begnaud (2019), “Investigating the effect of short term responsive VNS therapy on sleep quality using automatic sleep staging”, in proc. of IEEE Transactions on biomedical engineering, Volume: 66, Issue: 12, DOI: 10.1109/TBME.2019.2903987, INSPEC accession number: 19189150, 2019
• Shou-Hsiung Cheng, Jui-Chen Huang (2017), “An intelligent sleep quality detection system based on wearable device”, in proc. of 2017 IEEE international conference on machine learning and cybernetics, DOI: 10.1109/ICMLC.2017.8108966, INSPEC accession number: 17372997, 2017
• Mario Giovanni Terzano, Liborio Parrino, Arianna Smerieri, Ronald Chervin, Sudhansu Chokroverty, Christian Guilleminault, Max Hirshkowitz, Mark Mahowald, Harvey Moldofsky, Agostino Rosa, Robert Thomas and Arthur Walters (2002), “Atlas, rules, and recording techniques for the scoring of cyclic alternating pattern (CAP) in human sleep”, in proc. of Elsevier, DOI: 10.1016/s1389-9457(01)00149-6, 2002
• Rositsa Poryazova, Esther Werth, Liborio Parrino, Mario Giovanni Terzano and Claudio L. Bassetti (2011), “Cyclic alternating pattern in narcolepsy patients and healthy controls after partial and total sleep deprivation”, in proc. of clinical neurophysiology 122 (2011) 1788-1793 Elsevier, DOI: 10.1016/j.clinph.2011.02.028, 2011
• Carole L. Marcus, John L. Carroll, David F. Donnelly and Gerald M. Loughlin (2008), “Sleep in children – Development changes in sleep pattern”, in proc. of Lung biology in health and disease, Second edition, Publisher: Informa healthcare USA, Inc., ISBN-10:1-42006080-5 and ISBN-13:978-1-4200-6080-5, 2008
• Charles P. Pollak, Michael J. Thorpy and Jan Yager (2010), “The encyclopaedia of sleep and sleep disorder, Third edition”, Publisher: Facts On File, An imprint of infobase publishing, New York NY 10001, ISBN-13:978-08160-6833-3, ISBN-10:0-8160-6833-X and ISBN: 978-1-4381-2577-0, 2010