Site Loader
Rock Street, San Francisco

In recent old ages, addition of patients with a sleep apnea syndrome has become a serious job. Clogging sleep apnea syndrome is the most common respiratory perturbation in worlds. Many new diagnosing and intervention methods are invariably being proposed. EEG signal parametric quantities, extracted and analysed are extremely utile in nosologies, and has become an of import tool for sensing of sleep apnea. This paper proposes a new method for observing sleep apnoea by dividing EEG signals into its different frequence constituents by executing power spectral analysis. A comparative survey of these constituents are done and fed to the unreal nervous web as a preparation input. Furthermore, informations from assorted other topics are besides fed to the nervous web for sensing of SAS.


Sleep is a circadian beat indispensable for human life. Good slumber is an indispensable stipulation to the care of mental and physical wellness. Sleep apnoea is a common slumber upset characterized by brief breaks of take a breathing during slumber. It is defined as sleep-disordered external respiration distinguished by perennial episodes of upper air passage prostration during sleep [ 9 ] . These take a breathing intermissions typically last between 10 to 20 seconds and can happen up to 100s of times a dark doing alterations in cardiac and neural activity and discontinuities in sleep form [ 10 ] . When people enter in-between age and beyond, the upper respiratory piece of lands of some parts shrink, which may take to obstructor of rhinal transition and saw wooding during slumber. This apnea may impact the quality of slumber and wellness when it occurs often, and may even do decease in terrible instances [ 8 ] . The different types of slumber apnea syndromes are: Clogging slumber apnea, Central sleep apnea and Complex sleep apnea. Clogging slumber apnea occurs when the soft tissue in the dorsum of your pharynx relaxes during slumber, doing a obstruction of the air passage ( every bit good as loud saw wooding ) [ 9 ] , Central sleep apnea involves the cardinal nervous system, instead than an airway obstructor. It occurs when the encephalon fails to signal the musculuss that control take a breathing. Complex slumber apnea is a combination of clogging slumber apnea and cardinal sleep apnea [ 6 ] . The most common type of sleep apnoea is Clogging Sleep apnea syndrome ( OSA ) . The oncoming of slumber is typically characterized by gradual alterations in cortical electroencephalographic ( EEG ) activity. EEG is the recording of electrical activity along the scalp and is typically described in footings of ( 1 ) rhythmic activity and ( 2 ) transients. This rhythmic activity is divided into frequence and amplitude bands- Delta moving ridges ( frequence up to 4Hz and amplitude 20-100uv ) , Theta waves ( frequence 4- & A ; gt ; 8Hz and amplitude 10uv ) , Alpha waves ( frequency 8-13Hz and amplitude 2-100uv ) , Beta moving ridges ( frequence & A ; gt ; 13-30Hz and amplitude 5-10uv ) , Gamma waves ( frequency 30-100Hz ) , Mu ( besides 4-8Hz ) .

The chief aim of this work is to research assorted possible relationships among sleep phases and apnoeic events and better on the clinical truth of algorithms for sleep categorization and apnea sensing. Surveies have shown that there are alterations in cortical activity that occur during slumber disordered external respiration ( SDB ) events like sleep apnoea. EEG signals will be assessed utilizing advanced signal processing attacks in which EEG signal is segregated into different frequence bands-delta, alpha, theta and gamma. These parametric quantities are used for the sensing of sleep apnea with the aid of unreal nervous web. Therefore, to invent a more economical method, we focus chiefly on the exclusive ability of cortical EEG for the sensing of sleep apnoea.


The normal human EEG is observed to demo activity over the scope of 1-30 Hz with amplitudes in the scope of 20-100 µV. The surface EEG shows typical forms of activity that can be correlated with assorted phases of slumber and wakefulness.

Sleep is non a unvarying province, but is characterized by a cyclic jumping form of non-rapid oculus motion ( REM ) and REM sleep. Non-REM is divided into four phases of slumber:

A individual falling asleep is observed to first enter phase 1 characterized by low-amplitude, high frequence ( 6-8 Hz ) EEG activity. Alpha beat are more prevailing in this province of sleepiness. Sleep 2 phase is the light slumber province, where the oculus motions halt and our encephalon waves become slower. Particular moving ridges ‘K-complexes and sleep spindles begin to look. In this province, EEG amplitude is average, of about 10-50 µV and EEG frequence is 4-7Hz.In phase 3 ( deep slumber ) , highly slow encephalon moving ridges called delta moving ridges begin to look, interspersed with smaller, faster moving ridges. By phase 4 ( deep slumber, slow wave slumber ) , the encephalon produces delta moving ridges. In phase 4, the amplitude of EEG will be high, but the frequence will be less than 2 Hz.

REM slumber is observed as rapid, low electromotive force, irregular EEG activity associated with musculus vellications and rapid oculus motions. Theta moving ridge is more prevailing in this sleep phase.


Some physiological observations of sleep apnea patients during apnoeic events have been recorded as follows:

During each episode of obstructor there is a lessening in O impregnation which is sometimes accompanied by a decelerating bosom rate. At the terminal of the episode the EEG is said to demo a brief ( 3-10 seconds ) explosion of alpha activity, the EMG ( EMG ) is elevated and the bosom rate is accelerated. After this, take a breathing sketchs and O impregnation returns to the degree of wakefulness. This form occurs recurrently throughout the dark which consequences in sleep atomization and therefore the upset is associated with daylight symptoms, most frequently inordinate drowsiness. This besides consequences in cognitive diminution, memory loss etc [ 10 ] .

Full polysomnography ( PSG ) , which measures slumber, respiratory variables and oxygenation, is still regarded as the criterion method to name Sleep Apnoea Syndrome. Each patient undergoes a individual dark of PSG trial utilizing standard techniques in which Central and frontal EEG ( C3-C4, CZ-PZ, F3-FP1, F4-FP2 ) , two electro-oculograms ( EOG ) , submental and left and right tibial EMGs ( EMG ) and the EKG ( ECG ) are monitored. Respiratory variables monitored include rhinal airflow, unwritten air flow measured by thermal resistor, thoraxo-abdominal motion by induction plethysmography and O impregnation ( SaO2 ) . At the terminal of the nightlong slumber survey a sleep specializer scores the polygraph recording, placing sleep phases and apnoeic and hypoapnoeic events doing O desaturation.

From this analysis an Apnoea Hypopnea Index ( AHI ) is calculated for the patient:

Apnoea-Hypoapnea Index: This index is used to mensurate the badness of the sleep apnoea. The mean figure of apneas and hypopneas during one hr of slumber is called the apnea/hypopnea index ( AHI ) or respiratory perturbation index ( RDI ) . It measures the frequence of decreases in air flow associated with upper-airway prostration or narrowing that occurs with the province alteration from wakefulness to kip. Apnoea is complete surcease of air flow for 10s or more. Hypoapnoea is associated with a lessening of respiratory volume by 50 % for more than 10 seconds. A patient holding an AHI between 5 and 15 is said to hold mild Obstructive Sleep Apnoea ( OSA ) , whereas 15 to 30 is moderate and more than 30 events per hr is diagnosed as holding terrible sleep apnea. These categorizations besides depend on factors such as sleepiness etc.


Many research workers have assorted techniques including unconventional attacks such as technology diagnostic techniques, for finding patient ‘s status. A reappraisal of literature includes, rapid electroencephalographic alterations in response to intellectual anoxia were observed every bit early as 1925. Since so many researches has been conducted on sensing of sleep apnoea by EEG signal processing affecting assorted methods and algorithms. Further on [ 1 ] Guilleminault et Al. ( 1996 ) reported a delta set amplitude addition get downing on mean 13 seconds after the oncoming of apnea. During NREM, the mean differences between initial and maximum values were found to be 268 % and 202 % between initial and concluding values during the event continuance. [ 2 ] Berry et Al. ( 1998 ) further studied the fluctuation in delta power and reported a cyclic addition in delta power which was in synchronism with increased respiratory attempt in NREM slumber. [ 3 ] Baumgaurt-Schmitt et Al. ( 1998 ) have used nervous web to sort the assorted sleep phases by pull outing the characteristics from the familial algorithms. Here, end products of nine different webs were created by utilizing informations of 9 different topics which were used at the same time for categorization. [ 4 ] Dingli et Al. ( 2002 ) have shown the spectral analysis technique for the sensing of cortical activity alterations in sleep apnoea topics. The most consistent important alteration is the lessening in theta power, during Non-Rapid Eye Movement ( NREM ) slumber which is either associated with an addition in high frequences ( alpha and sigma ) or with delta addition. [ 5 ] Lin et Al. ( 2006 ) have implemented a new technique for categorization and analysis of EEG signals, by utilizing ripple transforms and so feeding the spectral constituents to the inputs of an unreal nervous web for acknowledgment of EEG signal features of Sleep Apnoea Syndrome which was configured to give three end products to mean the SAS state of affairs of the patient. The acknowledgment threshold for all trial signals turned out to hold sensitiveness degree of about 69.64 and a specificity value of about. [ 6 ] T.Sugi et Al. ( 2009 ) proposed the method for automatic sensing of EEG rousings in SAS patients. To efficaciously observe respiratory-related rousings, threshold values were determined harmonizing to pathological events. The proposed method was applied to Polysomnographic ( PSG ) records of eight patients with SAS and truth of EEG rousing sensing was verified by comparative ocular review. [ 7 ] Chein-Chang Hsu et Al. ( 2010 ) used the Hilbert-Huang transmutation for pull outing the frequence component from Hilbert spectrum. The chief part of the system is to continue clip information in the EEG by Hilbert-Huang transmutation mechanism every bit good as find frequence fluctuation information. [ 8 ] Boyu Wang et Al. ( 2009 ) has performed the different off-line methods for two different EEG signal categorization task-motor fanciful and finger motion.


Sleep apnoea is presently diagnosed by the analysis of clinical polysomnography trial which measures EOG, EMG, to place assorted leg motions, ECG, to document cardiac arrhythmias, EEG, to document assorted sleep provinces, nasal/oral air flow, SaO2 ( oxygen desaturation ) and pectoral motion. EEG signals being non-stationary, non- invasive are rather utile since they give off no radiation and can be recorded over a long interval of clip. Here, we chiefly focus on analysis of EEG signals for sensing of sleep apnoea.


EEG slumber informations ( in.txt format ) has been obtained from sleep database ( MIT-BIH Polysomnography database ) . Datas from five topics are taken, among which all are male, age runing from 32-56 old ages and their weights runing from 89-152 kilogram. These records were analysed over 30 seconds window for a period of 1 hr. Following the 10-20 System of electrode constellation, the electrode arrangement for the EEG signal was either C4- A1 or C3-O2. The signal from the information has been sampled at a frequence of 250 Hz. The algorithm starts with acquisition of EEG informations of multiple topics which has been processed to bring forth an artefact and intervention free signal.


The informations holding sleep apnoea events, acquired from database ( ) is processed in MATLAB. We load EEG informations taken from one topic ( slpdb59 ) , in MATLAB and plot the signal. Following, we take 30 seconds Windowss for a period of 1 hr and analyze them utilizing Power Spectral Density, which is done by utilizing Welch method of averaged periodograms, utilizing Overacting window with 64 points overlap ( 50 % convergence ) . This was performed by making a Welch object in MATLAB by utilizing the map spectrum.welch and using it in the map psd written in MATLAB. This consequences in an mean power spectral denseness curve ( µV2 /Hz ) that lies over the frequences ± ? . Thus, the information is converted to its frequence sphere which is sampled at N equidistant samples, which is given by the power of two, holding value greater than the length of the signal. PSD is estimated individually for all the Windowss, at the same time segregating the signal into its delta, theta, alpha, and beta frequence sets. The power spectral denseness matching to respective frequence sets is calculated and their power ratios are calculated by spliting mean power of single slumber moving ridge frequence set by the entire mean power of all the sets. The entire mean power of each frequence set is calculated by gauging Power spectral analysis over 30s window for the full signal, and therefore taking mean of the single frequence sets of all the Windowss. Now we compare the window incorporating sleep apneic event with the non-event slumber window and note the differences between the power ratios of the several frequence sets such as Delta, Theta, Alpha and Beta. This process is repeated for the full signal continuance by taking 30 s Windowss, therefore separating between apnoeic and non-apnoeic parts

Now, that we have an established a threshold, we feed the information to an unreal nervous web, which adapts and learns by itself, and so utilize this information for the sensing of sleep apnoea for more topics or informations ( in this instance five ) .


The power spectral denseness is performed by making a welch object in MATLAB, and taking N point FFT for power spectrum appraisal. For ciphering N points, we suppose that the signal representation in the frequence sphere is X ( ? ) and is periodic with a period 2? . Sing the frequence sphere is sampled at N equidistant samples, the interval between two consecutive samples is ?? . This is written mathematically as:


Therefore, the estimated power spectrum utilizing Welch method is given by:

At frequences fn = n/N where n= 0,1, … , N-1


By the terminal of this enterprise we expect to carry through newer methods for the sensing of sleep apnea by elaborate survey and analysis of the parameters- Delta, Theta, Beta and Gamma, which offers a clinical mention value for placing Sleep Apnoea Syndrome, therefore cut downing diagnosing clip and bettering medical service efficiency.


In this paper we have tried to happen a new solution to the job of placing Sleep Apnoea Syndrome ( SAS ) episodes. In order to accomplish this end, the EEG signal features of SAS episodes were extracted utilizing power spectral denseness and so detected utilizing an Artificial Neural Network.

Post Author: admin