----------------------- Page 22----------------------- Sleep Science ORIGINAL ARTICLE Electroencephalographic and electrocardiographic synchronic activation during sleep apneas detected using coherence wavelet method Ativação eletroencefalográfica e eletrocardiográfica sincronizada durante apneia do sono detectada por meio do método coherence wavelet 1 2 3 4 Susana Blanco , Marcela Smurra , Hernando Sala , Cecilia Di Risio ABSTRACT visíveis em observação direta no estudo polissonográfico. Os restantes Objective: Breath perturbations associated with sleep are accom- devem ser avaliados por meio de técnicas como redes neurais ou análi- panied by cortical micro-awakenings. Seventy per cent of them are ses espectrais que utilizem o método de Fourier. O objetivo deste estu- visible by direct observation on the polysomnographic study. The re- do foi estabelecer se a aplicação de um método matemático dinâmico maining may be evaluated using techniques such as neural networks para análise do traçado polissonográfico possibilita a identificação da or by spectral analysis using Fourier’s method. The objective was to frequência de associação do eletroencefalograma e eletrocardiograma establish whether the application of a dynamic mathematical method como resposta a eventos respiratórios. Métodos: Polissonografias de for polysomnographic tracing enables the identification of an associa- 22 pacientes (14 homens e 8 mulheres) portadores de distúrbios res- tion frequency of the electroencephalogram and the electrocardiogram piratórios de sono e 3 indivíduos hígidos (2 homens e 1 mulher). Para as a response to breathing events. Methods: Polysomnographs of 22 estabelecer a relação entre ativação cortical (microdespertares) e eletro- patients (14 males and 8 females) with sleep disordered breathing cardiograma, ambos os sinais foram analisados. O método matemático (SDB) and 3 normal subjects (2 males and 1 female). In order to estab- usado consiste de uma decomposição de banda wavelet dos sinais ele- lish the relationship between cortical activation (micro-awakening) troencefalográficos e eletrocardiográficos juntos com coerência wavelet. and the electrocardiographic (ECG), both signals were analyzed. The Resultados: O padrão de sincronização mais evidente foi observado mathematical method used consisted of a wavelet band decomposition entre a banda eletrocardiográfica correspondente aos complexos QRS of the electroencephalographic (EEG) and the ECG signals together de alta frequência e as bandas eletroencefalográficas alfa. A correlação with wavelet coherence. Results: The most evident synchronisation foi encontrada entre a frequência de microdespertares e a frequência pattern occurred between the ECG band corresponding to the high dos valores máximos da correlação entre banda eletroencefalográfica frequencies QRS complex and the EEG alpha band. A correlation was alfa e banda eletrocardiográfica rápida. Essas frequências representam found between the micro-awakening frequency and the frequency of tempos de sincronização entre 1,5 e 2,8 segundos. Conclusão: Acima the maximum values of the correlation between the EEG alpha band do ponto de corte de 40 microdespertares por hora, a população que and the ECG fast band. These frequencies represent synchronization apresenta significante atividade miocárdica foi agrupada demonstran- times between 1.5 and 2.8 seconds. Conclusion: Above a cut point do significante atividade autonômica, detectada no traçado polissono- of 40 micro-awakenings/hour, the population with a significant myo- gráfico. Quando o nível de fragmentação do sono é alto, uma ativação cardial activity was grouped showing a significant autonomic activity miocárdica significante torna-se evidente. detectable in the polysomnographic tracing. When sleep fragmenta- tion level is high, a significant myocardial activation appears. Descritores: Síndromes da apneia do sono/diagnóstico; Transtornos do sono; Polissonografia; Análise de Fourier; Eletroencefalografia; Ele- Keywords: Sleep apnea syndromes/diagnosis; Sleep disorders; Poly- trocardiografia somnography; Fourier analysis; Electroencephalography; Electrocar- diography INTRODUCTION RESUMO Breath perturbations associated with sleep are accompanied Objetivos: Perturbações respiratórias associadas ao sono são acom- by cortical micro-awakenings. Seventy per cent of them panhadas de microdespertares corticais. Setenta por cento destes são are visible by direct observation on the polysomnographic 1PhD, Consejo Nacional de Investigaciones Científicas y Técnicas – CONICET, Facultad de Ingeniería, Universidad de Belgrano, Buenos Aires, Argentina. 2MD, Hospital Tornú, Buenos Aires, Argentina. 3MD, Hospital Nacional Profesor Alejandro Posadas, Buenos Aires, Argentina. 4PhD, Facultad de Ingeniería, Universidad de Belgrano, Buenos Aires, Argentina. Corresponding author: Marcela Smurra − (1431) Hospital Tornú - Combatientes de Malvinas 3002 − (5411) 45233200 − Ciudad Autónoma de Buenos Aires − República Argentina − E-mail: marsmurra@yahoo.com Received: 30/12/2010, Accepted: 28/02/2011 Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 23----------------------- 137 Blanco S, Smurra M, Sala H, Di Risio C study(1) . The remaining 30% may be evaluated using tech- METHODS niques such as neural networks(2) or by spectral analysis us- A total of 25 polysomnographic studies were analyzed, ing Fourier’s method(3) , which improve the detection of the belonging to 22 patients with SDB: 10 corresponding to cortical activity associated with breathing events that are slight-moderate SDB (5-30 apneas-hypoapneas/hour), 12 not visible for those who process the study. with severe SDB (> 30 apneas-hypoapneas/hour) and 3 Other types of responses, such as the autonomic ones, normal subjects (no apneas), that represent a 10% of the that appear together with the breathing phenomena are: sample. variations of cardiac frequency, arterial tension or variabil- Patients included were those with obstructive sleep ity of the cardiac response with the sympathetic/parasympa- apneas with less than 5 central apneas/hour and less than thetic balance. 50% of the apnea-hypoapnea index. Patients who presented In some cases, other methods of autonomic measurement neurological disorders, periodic leg movements, narcolepsy are required, like pulse transit time (PTT) or peripheral ar- or took medication that could interfere with sleep patterns terial tonometry, that represent an indirect determination of were excluded from the considered population. Moreover, autonomic phenomena such as vasoconstriction or arterial patients with arrhythmia or cardiological history were also tension variation associated with abnormal breathing events excluded. during sleep(4,5). The polysomnograph was carried out using a computer- In a direct way, the presence of oscillations of heart ised system recording: rate variability at the high frequency band (HF) (> 0.10 1) three EEG channels: two central (C3 and C4) and one Hz) is associated with parasympathetic activity related to occipital (O1) referred to right and left mastoids (A2-A1); breath, while oscillations at the low frequency band (LF) 2) two electroculography (EOG) channels with palpebral (0.01 - 0.05 Hz) are not related to respiration and are right and left electrodes; thought to correlate sympathetic activity. Intermediate 3) three electromyography (EMG) channels, two of them frequencies are still undefined in their cause(6) . In healthy in sub-chin and the other in tibial locations; individuals, a decrease in vagal tone was observed, as 4) ECG recorded from two electrode derivations in the sec- well as a sympathetic predominance during REM sleep ond right and left intercostal space in paraesternal location; in men(7) . In the sleep disordered breathing (SDB), a 5) oronasal flow sensors, one of them of the thermistor decrease of HF components is detected together with a type and the other constituted by a nose canula, with pres- relative increment of the LF components(8) . The corti- sure transductor to detect flow limitation by nose pressure cal activation without visible micro-awakenings (alpha measurement; rhythm: 8-12 Hz) in the electroencephalography (EEG), 6) thoracic and abdominal piezoelectric band to record associated with simultaneous variation of the arterial ten- breathing efforts; sion in patients with Cheyne Stokes breath(9) , suggests a 7) body position sensor; questioning of the presence of other EEG rhythms associ- 8) pulse oxymetry (as part of the polysomnograph equip- ated with cardiac autonomic events in patients with sleep ment, Praxis 18 – Line AMP18P, Lermed S.R.L., Argentina). apnea and this breathing pattern. The wavelet transform is a mathematical method(10) Visual assessment of the alpha band power behaviour that allows a decomposition of the biological signal, sep- with respect to the marked micro-awakenings was per- arating its components more effectively than the usual formed with the software associated to the AMP 18P - Le- Fourier analysis, by adding the time variable to the evalu- rmed S.R.L. equipment. The remaining algorithms required ation of non steady signals such as EEG and electrocardi- to calculate the correlation between ECG and EEG channels ography (ECG). were developed with MatLab 7.0. As concerns these topics, the objective of the present Polysomnograph tracings in non-rapid eye movement work was to apply a dynamic mathematical method for the (NREM) sleep stage were chosen and read sequentially in identification of EEG and ECG frequency interaction asso- order to avoid the bias produced by the simultaneous ob- ciated to breathing events, and to evaluate this method as servation of EEG and breathing events. At first, the EEG a tool for the association of EEG behaviour with the cardiac stratification was performed according to Rechtschaffen- variability, in order to know if the frequencies of both bio- Kales standards(11) ; micro-awakenings were then defined as logical signals may be related in the SDB. This relation- the appearance of sudden changes in alpha frequencies in the ship, as severity clinical feature, could explain some facts EEG (not sleep spindles) of at least 3 seconds and no more in the development of cardiovascular morbidity in patients than 15 seconds(12) , assuming there were sleep records 10 with SDB. seconds before and after the micro-awakening. Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 24----------------------- 138 EEG and ECG synchrony by wavelet method The rapid eye movement (REM) stage was not considered In summary, the scale identifies the frequency band, in the signal analysis due to the greater stability in NREM while the position locates such frequency components in sleep stages, considering that the cortical activity reappear- time. Their functional dependence may be expressed as fol- ing in REM with alpha type (8-12 Hz) and theta type (5-8 lows: ∞ Hz) frequencies is an intrinsic part of such stage. The second sequence was analyzed on those breathing C(scale,p osition) = ∫f (t).Ψ(scale,p osition,t).dt - ∞ events classified as follows: 1) obstructive apnea: breath flow ceased or signal drop to where f(t) is the original series and Y are the infinite ele- 85% during 10 seconds or more, with persistence of breath- ments of the wavelet base. ing effort (activity of thoracic-abdominal bands); The spectrum decomposition results asymmetric to- 2) central apnea: breath flow ceased or signal drop to wards low frequencies, which produces a numerical ad- 85% during 10 seconds or more, without evidence of breath- vantage. With this method, details have a high frequency ing effort in thoracic-abdominal bands; definition and a poor time location. Approximations, 3) hypopnea: significant decrease in breath flow signal instead, have a low frequency location but a high time during 10 seconds or more, with a decrease in the breathing location. This choice for signal decomposition is not ar- effort signal amplitude with more than 3% of desaturation bitrary. The signals usually found are time-localised bun- or presence of a micro-awakening; dles, that is to say, they have high frequency components 4) flow limitation: flattening pattern in the flow sensor during a short period of time, and low frequency compo- due to nose pressure of at least three breaths followed by the nents of long duration. normalisation of the plot morphology. As the convolution process is iterative, it could be theo- retically continued indefinitely. A reasonable number of de- Mathematical method: wavelet transform and wavelet composition steps are usually chosen based on the particular coherence needs of the actual problem. Once the decomposition level The correlation in the spectral analysis indicates the pos- required is achieved, it is possible to analyze each band to sibility of some relation between two time series through a look for information. common frequency. If the mother function used for decomposition is a base Fourier Transform can determine linear interactions in in the mathematical sense, the details squared may be con- stationary time series, but nonlinear interactions in non sidered as the energy deposited independently on each band. stationary time series are better analyzed by means of the In this work, mother wavelets of the cubic spline type are wavelet coherence method. This mathematical operation used, which, due to their morphological characteristics, are gives valuable information about when and how two signals the most adequate for biological signals. On the other hand, synchronized in time. they constitute a base in the mathematical sense. The wavelet transform (WT) uses a base of periodic func- The coherence is defined as the cross-spectrum, normal- tions, which is built from a located function named “mother ized to an individual power spectrum. It is used to identify function”. The wavelet base function changes its scale by frequency bands within which two time series are covered, “expanding” or “compressing”, and each scale is convoluted and to determine the time and frequency intervals in which with the analyzed signal, moving the signal along it. two phenomena have a strong interaction. While in the Fourier approach only the frequency pa- For coherence calculation between EEG and ECG wave- rameter was varied and the base function remained un- let decomposition bands, the steps below were followed: changed, in the WT the base function changes its shape - To determine micro-awakening, about 10 seconds of according to two parameters, one of them related to the the EEG signal was taken together with the corresponding displacement on the signal and thus to time and the other ECG portion – these portions of EEG and ECG correspond- related to translation or to scaling, that analyzes frequency ing to sleep Phases I and II, without artifacts. Channels C3 simultaneously. and C4 of the EEG were chosen. For each patient, 10 similar These operations allow decomposing the original signal micro-awakenings were used. in sub-signals called approximations, each corresponding to - Both signals EEG and ECG were decomposed with the original signal filtered in the band of the respective ex- wavelet pass bands separating them in 6 octaves. This is al- pansion scale. Each approximation is associated with a set of ways possible because both signals are sampled at the same coefficients called details, obtained from the convolution of frequency (349 Hz). the wavelet with the signal, that represent the relative val- - The coherence wavelet method was applied between ues of the band importance as a function of time. the 36 pairs of EEG and ECG bands. Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 25----------------------- 139 Blanco S, Smurra M, Sala H, Di Risio C Statistics If we plot maximum correlation frequency as a function of In order to establish the correlation between the micro- micro-awakening/hour (arousals), we obtain a good correlation, awakening frequency values and the maximum correlation with Pearson’s coefficient of 0.817. Figure 1 shows graphic data. frequency, the Pearson regression coefficient was calculated Figure 2 shows the same graph, but for slight-moderate for the analysis of the results. F- and t-tests were used to (A) and severe (B) groups of patients. Figure 2A indicates compare mean and standard deviation at the significant val- that spectral frequencies are between 0.5 and 0.6 Hz for ues of α = 0.05 and 0.01, respectively. slight-moderate group, while in the severe cases, they are between 0.6 and 0.7 Hz. RESULTS The group of studied patients (n = 25) consisted of 16 males and 9 females, ages between 26 and 78 years (54 ± 16 years), y c 2 n and body mass index (BMI) between 21.6 and 54 kg/m (29 e u q 2 e ± 7 kg/m ). Table 1 shows the results patient by patient. r f The coherence between EEG alpha band and ECG high n o i t a frequency band showed a periodic pattern with a distinc- l e r r tive frequency establishing the maximum synchronisation o c x between both of them. a M Table 1 shows the synchronization frequency in hertz (Hz) and the maximum correlation times between EEG and ECG for each individual; these are the mean values for all Micro-awakenings/hour synchronization frequencies for each patient without consid- ering the kind of breathing event. Figure 1: Synchronization frequency and micro-awakenings/hour correlation plot in the whole group. Table 1: Demographic data. Ages are expressed in years and body mass A index (BMI), in kg/m2 ) z H ( Maximum Arousals/ Time y c Patient Age BMI Severity Correlation n hour (seconds) e u Frequency (Hz) q e r f P1 32 23 N 3 0.48 2.08 n o P2 74 31.4 N 4.8 0.41 2.4 i t a P3 49 26.6 N 13.8 0.49 1.1 l e r r P4 60 25 M-L 16.8 0.55 1.81 o c P5 36 23.2 M-L 17.4 0.48 2.08 x a M P6 69 29.2 M-L 19.2 0.57 1.75 P7 50 38 M-L 21.6 0.53 1.88 P8 26 21.6 M-L 24 0.55 1.8 P9 44 23.8 M-L 30 0.55 1.81 Micro-awakenings/hour P10 67 29 M-L 30 0.59 1.69 B P11 76 30.3 M-L 33 0.53 1.88 ) P12 45 28.2 M-L 40 0.53 1.86 z H ( P13 30 24.4 M-L 45 0.58 1.72 y c P14 59 33.8 S 45 0.64 1.56 n e u P15 28 54 S 45 0.59 1.69 q e r f P16 68 24.2 S 46.2 0.57 1.75 n o P17 50 27.7 S 46.8 0.62 1.61 i t a l P18 54 23.9 S 60 0.65 2.08 e r r P19 28 37.5 S 60 0.68 1.47 o c x P20 59 37.2 S 60 0.61 1.6 a M P21 78 23.9 S 63.6 0.64 1.56 P22 65 29 S 75 0.71 1.4 P23 61 39.1 S 76 0.66 1.51 Micro-awakenings/hour P24 78 23.9 S 79.2 0.65 1.53 P25 50 27.7 S 81.6 0.7 1.42 Figure 2: Synchronization frequency and micro-awakenings/hour N: normal subject; M-L: slight-moderate; S: severe. correlation plot. A: slight-moderate; B: severe population. Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 26----------------------- 140 EEG and ECG synchrony by wavelet method Mean values were 0.546 and 0.643 Hz for slight-mod- bilities of cardiac events(16) . Also, in association with CAP erate and severe patients, respectively. To establish a level phenomena, secondary micro-awakenings responding to of significance of this difference, F-test for comparison of auditory stimuli were detected, in which a substantial car- standard deviation at a level of α = 0.05, and t-test for com- diovascular activation was observed, coinciding with patho- parison of mean (α = 0.01) were performed. Results show logical situations such as SDB or periodic leg movements(17) . that this difference is significant. Different regions of the heart are differentially activated Figure 2 also shows the existence of a virtual cut among during arousal from sleep, and they may be partially influ- patients with severe SDB and individuals with slight-mod- enced by different respiratory and non respiratory related erated or without SDB. Such cut is represented by the value sensory inputs to the neural cardiomotor centres. Cardiac of 40 micro-awakenings/hour above which the population arousals seem to have preceded cortical arousals, the pre- presenting significant myocardial activity has been grouped dominant responses to arousal were respiratory rate (RR) together. and QT interval shortening and PR interval lengthening, Synchronisation times among them − the EEG alpha although different activation patterns with potentially dif- band and the QRS band for ECG − fluctuate between 1.5 ferent arrhythmogenic potential were also observed(18) . The and 2.8 seconds. increase in cardiac activity during micro-awakenings is mainly a reflex activation response, which implies a decrease DISCUSSION in the vagal tone and an increase of the sympathetic during Polisomnography, in its usual interpretation, offers clear and the period of the micro-awakening(19,20) . reliable information about sleep stages and breathing events. Respiratory patterns that need correction activate the Nowadays, there are methods of sleep signal analysis central nervous system (CNS); the autonomic nervous sys- that allow a reduction on the visual reading time of the tem is enhanced when an arousal occurs, which explains the polisomnographic tracings and let us find, in an automat- greater increase in heart rate with EEG arousal than without ed way, sleep stages and disease related patterns linked to arousal(21) . them. The regularly used methods are sleep analysis system The slow waves determine a softer vegetative reac- to challenge innovative artificial networks (SASCIA), which tion, which in certain pathologic conditions may be strong works with neural networks, or cyclic alternating pattern enough to overcome a disturbing factor. Faster EEG activi- (CAP)(13-15) . The use of the wavelet decomposition (WPM) ties guarantee more powerful activation of autonomic func- plus wavelet coherence in this project helped us to evalu- tions(22) . ate simultaneous variations in the EEG signals, and in those The techniques that used spectral analysis with time who show an autonomic response of the organism as a result variation or fractal dimensions have analyzed variability in of apneas, hypopneas or limiting flow events. It makes this the cardiac frequency, but not the QRS changes in terms of variations appear in the polisomnographic tracings, as for analysis of signal power(23,24) . its high capability in identifying short duration changes in The vagal-sympathetic unbalance would be in the func- a synchronized way, analyzing simultaneously all the signs tion of the intracardiac neuronal network with vagal-sym- in the polisomnography, which could not have been possible pathetic co-activation, in which the brachicardia represents using the Fourier transform(1) because it is insensitive to the an adaptive reaction to protect the heart of hypoxia by re- fast changes and cannot be used in non-steady signals. ducing oxygen consumption. The concomitant sympathetic It is rather frequent to find bradi-tachycardia episodes, activation would improve contractility, optimizing systolic but only in patients with severe apnea/hypopnea sleep syn- volume(25) . drome, relating this to the unbalances between the sympa- The referential signal for the detection of autonomic thetic-parasympathetic activity relation. In previous evalua- phenomena was developed by Pitson and Stradling(26) and tions using CAP, it could be observed that the sympathetic consists in the analysis of the pulse transit time between the activation is significantly higher during CAP events than heart and the peripheral sensor placed in a finger, as it varies non-CAP ones, and it is clearly related with the increase in following the changes of arterial tension. This analysis has blood pressure. Parasympathetic activity is predominant in the objective of determining the presence of an autonomic NREM, unlike REM which has sympathetic predominance. response, called autonomic micro-awakenings, no detectable The disappearance of the vagal predominance in NREM was on the EEG. This signal not evaluated in our work repre- observed in patients with myocardial infarction and it was sents a limitation to consider the results as found directly proposed as a significant factor in the occurrence of noctur- related to autonomic activation. However, the high correla- nal fatal events. Moreover, the obstruction of the baroreflex tion found between the EEG micro-awakenings and ECG increment in CAP would lead to an increase in the possi- signal characterized by the increment in power of the QRS Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 27----------------------- 141 Blanco S, Smurra M, Sala H, Di Risio C band is probably an expression of increment on the auto- cardiologic activity, represented by the changes in the QRS nomic activity not described in the previous reading of the power in the routinely polysomnographic tracing. Wavelet polisomnography. coherence established a mathematical relationship between Recently, the analysis of the cardiac frequency was used fast cortical and autonomic responses, showing more sever- during specific intervals (e.g. micro-awakenings) using the ity when both reactions are observed. average cardiac frequency prior to and during the event, In our knowledge, there are not enough data using wave- considering likelihood ratios over 10 levels of cardiac fre- let coherence to analyze autonomical signal. Further studies quency variations through an algorithm of autonomic activa- in our group try to get the link among PTT, ECG and corti- tion identification based only in cardiac frequency changes. cal activity. The model showed a lack of correlation between the visual reading of the EEG and the ECG events detected through REFERENCES the algorithm(27) . It is not clear whether this discrepancy in 1. Dingli K, Assimakopoulos T, Fietze I, Witt C, Wraith P, Doug- signals responded to autonomic events, as the signal was not las NJ. Electroencephalographic spectral analysis: detection of referred to phenomena detected through pulse transit time. cortical activity changes in sleep apnoea patients. Eur Respir J. 2002;20(5):1246-53. Our interpretation is that the increase in the power of 2. Pardey J, Roberts S, Tarassenko L, Stradling J. A new approach to the QRS slow band in detriment of the fast band represents the analysis of the human sleep/wakefulness continuum. J Sleep a more interesting signal in the appreciation of cardiac par- Res. 1996;5(4):201-10. ticipation than frequency, as the latter is influenced by ex- 3. Svanborg E, Guilleminault C. EEG frequency changes during tra cardiac phenomena, although it would be interesting in sleep apneas. Sleep. 1996;19(3): 248-54. further analysis to determine which of these two variations 4. Argod J, Pépin JL, Smith RP, Lévy P. Comparison of esophageal is more influenced by autonomic control, correlating both pressure with pulse transit time as a measure of respiratory effort for scoring obstructive nonapneic respiratory events. Am J Respir with PTT, which was a limitation of our work. Crit Care Med. 2000;162(1):87-93. It does not seem possible that these results would be 5. Lavie P, Schnall RP, Sheffy J, Shlitner A. Peripheral vasoconstric- due to an intrinsic illness of the conduction system, but to tion during REM sleep detected by a new plethysmografic meth- a delay in the intraventricular conduction, modulated by a od. Nat Med. 2000;6(6):606. sympathetic-parasympathetic unbalance, as analyzed pre- 6. Mokrane A, Nadeau R. Dynamics of heart rate response to sympa- viously. A similar behavior is considered a predictor of the thetic nerve stimulation. Am J Physiol. 1998;275(3 Pt 2):H995- 1001. appearance of arrhythmic events and mortality in cases of 7. Elsenbruch S, Harnish MJ, Orr WC. Heart rate variability myocardial infarction or congestive heart failure(28) . This cir- during waking and sleep in healthy males and females. Sleep. cumstance might be related to the observation made in this 1999;22(8):1067-71. work that the mentioned behavior is more frequently found 8. Narkiewicz K, Montano N, Cogliati C, van de Borne PJ, Dyken in the most severe patients, which could be related with the ME, Somers VK. Altered cardiovascular variability in obstruc- clinical evolution of SDB complicated with the appearance tive sleep apnea. Circulation. 1998;98(11):1071-7. Comment in: of fatal and non fatal cardiovascular events(29) . Circulation. 1999;100(25):e136-7. 9. Davies R, Bennet L, Barbour C, Tarassenko L, Stradling J. Sec- Another possible explanation for our observation is the ond by second patterns in cortical electroencephalograph and amplitude increase of QRS, which, even in individuals with- systolic blood pressure during Cheyne-Stokes. Eur Respir J. out SDB, implies an increased risk in arrhythmia produc- 1999;14(4):940-5. tion when compared to narrower QRS(30) . Polysomnographic 10. De Carli F, Nobili L, Gelcich P, Ferrillo F. A method for the auto- tracings were related to sleep apneas and hypopneas of dif- matic detection of arousals during sleep. Sleep. 1999;22(5):561- ferent severities with the purpose of correlating cortical ac- 72. 11. Rechstschaffen A, Kales A. A manual of standardized terminol- tivity and cardiologic response. The presence of a threshold ogy: techniques and scoring systems for sleep stages of human in 40 micro-awakenings/hour shows that the correlation subjects. Bethesda, Md: National Institutes of Health;1968. was significant when the presence of micro-awakenings was (Publication No.204) higher, which could mean that the sympathetic-parasym- 12. EEG arousals: scoring rules and examples: a preliminary report pathetic unbalance requires a higher level of severity in the from the Sleep Disorders Atlas Task Force of the American Sleep breathing event frequency in order to obtain higher levels of Disorders Association. Sleep. 1992;15(2):173-84. cortical activation. 13. Schaltenbrand N, Lengelle R, Toussaint M, Luthringer R, Carelli G, Jacqmin A, et al. Sleep stage scoring using the neural network model: comparison between visual and automatic analysis in nor- CONCLUSION mal subjects and patients. Sleep. 1996;19(1):26-35. This work, as far as we know, for the first time makes clear 14. Baumgart-Schmitt R, Herrmann WM, Eilers R. On the use of the possibility of detecting modifications of the intrinsic neural network techniques to analyze sleep EEG data. Third com- Sleep Sci. 2010;3(4):136–142136–142–142 ----------------------- Page 28----------------------- 142 EEG and ECG synchrony by wavelet method munication: robustification of the classificator by applying an 23. Blasi A, Jo J, Valladares E, Morgan BJ, Skatrud JB, Khoo MC. algorithm obtained from 9 different networks. Neuropsychobiol- Cardiovascular variability after arousal from sleep: time-varyng ogy. 1998;37(1):49-58. spectral analysis. J Appl Physiol. 2003;95(4):1394-404. 15. Rosa AC, Parrino L, Terzano MG. Automatic detection of cyclic 24. Togo F, Yamamoto Y. Decreased fractal component of human alternating pattern (CAP) sequences in sleep: preliminary results. heart rate variability during non-REM sleep. Am J Physiol Heart Clin Neurophysiol. 1999;110(4):585-92. Circ Physiol. 2001;280(1):H17-21. 16. Iellamo F, Placidi F, Marciani MG, Romigi A, Tombini M, Aqui- 25. Paton JF, Boscan P, Pickering A, Nalivaiko E. The yin and yan of lani S, et al. Baroreflex buffering of sympathetic activation during cardiac autonomic control : vago-sympathetic interactions revis- sleep: evidence from autonomic assessment of sleep macroarchi- ited. Brain Res Brain Res Rev. 2005;49(3):555-65. tecture and microarchitecture. Hypertension. 2004;43(4):814-9. 26. Pitson DJ, Stradling JR. Value of beat-to-beat blood pressure 17. Sforza E, Chapotot F, Lavoie S, Roche F, Pigeau R, Buguet A. changes, detected by pulse transit time, in the management of Heart rate activation during spontaneous arousals from sleep: the obstructive sleep apnoea/hypopnoea syndrome. Eur Respir J. effect of sleep deprivation. Clin Neurophysiol. 2004;115(11): 1998;12(3):685-92. 2442-51. 27. Basner M, Griefahn B, Müller U, Plath G, Samel A. An ECG- 18. Smith JH, Baumert M, Nalivaiko E, McEvoy RD, Catcheside PG. based algorrithm for the automatic identification of autonomic Arousal in obstructive sleep apnoea patients is associated with activations associated with cortical arousal. Sleep. 2007;30(10): ECG RR and QT interval shortening and PR interval lengthen- 1349-61. ing. J Sleep Res. 2009;18(2):188-95. 28. Javaheri S, Parker T, Liming J, Cobert WS, Nishiyama H, Wexler 19. Horner RL, Brooks D, Kozar LF, Tse S, Phillipson EA. Immedi- L, et al. Sleep apnea I 81 ambulatory male patients with stable ate effects of arousal from sleep on cardiac autonomic outflow in heart failure. Types and their prevalences. consequences and pre- absence of breathing in dogs. J Appl Physiol. 1995;79(1):151-62. sentations. Circulation. 1998;97(21):2154-9. Comment in: Cir- 20. Morgan BJ, Crabtree DC, Puleo DS, Badr MS, Toiber F, Skatrud culation. 1999;99(20):2709-12. JB. Neurocirculatory consequences of abrupt changes in sleep 29. Marín J, Carrizo S, Vicente E, Agustí A. Long-term cardiovascular state in humans. J Appl Physiol. 1996;80(5):1627-36. outcomes in men with obstructive sleep apnoea-hypopnoea with 21. Sforza E, Jouny C, Ibanez V. Cardiac activation during arousal in or without treatment with continuous positive airways pressure: humans: further evidence for hierarchy in the arousal response. an observational study. Lancet. 2005;365(9464):1046-53. Clin Neurophysiol. 2000;111(9):1611-9. 30. Stecker EC, Zargarian M, Dogra V, John BT, Kron J, McAnulty 22. Halász P, Terzano M, Parrino L, Bódizs R. The nature of arousal JH, et al. Native QRS duration predicts the occurrence of ar- in sleep. J Sleep Res. 2004;13(1):1-23. rhythmic events in ICD recipients. Europace. 2006;8(10):859-62. Sleep Sci. 2010;3(4):136–142136–142–142