Skip to main content
NCBI home page
Sleep logoLink to Sleep
. 2015 Mar 1;38(3):381–389. doi: 10.5665/sleep.4494

eAMI: A Qualitative Quantification of Periodic Breathing Based on Amplitude of Oscillations

Helio Fernandez Tellez 1,2,✉, Nathalie Pattyn 1,2,3, Olivier Mairesse 2,3,4, Leja Dolenc-Groselj 5, Ola Eiken 6, Igor B Mekjavic 7, P F Migeotte 8, Eoin Macdonald-Nethercott 9,10, Romain Meeusen 1, Xavier Neyt 2,11
PMCID: PMC4335521  PMID: 25581914

Abstract

Study Objectives:

Periodic breathing is sleep disordered breathing characterized by instability in the respiratory pattern that exhibits an oscillatory behavior. Periodic breathing is associated with increased mortality, and it is observed in a variety of situations, such as acute hypoxia, chronic heart failure, and damage to respiratory centers. The standard quantification for the diagnosis of sleep related breathing disorders is the apnea-hypopnea index (AHI), which measures the proportion of apneic/hypopneic events during polysomnography. Determining the AHI is labor-intensive and requires the simultaneous recording of airflow and oxygen saturation. In this paper, we propose an automated, simple, and novel methodology for the detection and qualification of periodic breathing: the estimated amplitude modulation index (eAMI).

Patients or Participants:

Antarctic cohort (3,800 meters): 13 normal individuals. Clinical cohort: 39 different patients suffering from diverse sleep-related pathologies.

Measurements and Results:

When tested in a population with high levels of periodic breathing (Antarctic cohort), eAMI was closely correlated with AHI (r = 0.95, P < 0.001). When tested in the clinical setting, the proposed method was able to detect portions of the signal in which subclinical periodic breathing was validated by an expert (n = 93; accuracy = 0.85). Average eAMI was also correlated with the loop gain for the combined clinical and Antarctica cohorts (r = 0.58, P < 0.001).

Conclusions:

In terms of quantification and temporal resolution, the eAMI is able to estimate the strength of periodic breathing and the underlying loop gain at any given time within a record. The impaired prognosis associated with periodic breathing makes its automated detection and early diagnosis of clinical relevance.

Citation:

Fernandez Tellez H, Pattyn N, Mairesse O, Dolenc-Groselj L, Eiken O, Mekjavic IB, Migeotte PF, Macdonald-Nethercott E, Meeusen R, Neyt X. eAMI: a qualitative quantification of periodic breathing based on amplitude of oscillations. SLEEP 2015;38(3):381–389.

Keywords: periodic breathing, quantification, Cheyne–Stokes respiration, loop gain, modulating index

INTRODUCTION

The most common types of sleep disordered breathing (SDB) are obstructive sleep apnea-hypopnea (OSA), followed by central sleep apnea-hypopnea (CSA) and periodic breathing. Obstructive sleep hypopnea is an event characterized by a transient reduction in breathing during sleep, while an apnea is a complete cessation caused by a partial or complete obstruction of the upper airway during sleep. Central sleep apnea-hypopnea is an absence or reduction of breathing and respiratory effort in the absence of any obstruction of the upper airway. Finally, periodic breathing is characterized by instability in the respiratory pattern that exhibits an oscillatory behavior with periods of hyperventilation followed by apneas or hypopneas. Periodic breathing may be observed in a variety of situations including damage to respiratory centers,1 acute exposure to high altitude,1–5 and in patients suffering from chronic heart failure.6–9 In patients suffering from chronic heart failure, periodic episodes of central and obstructive events are known to coexist due to feedback gains in addition to upper airway instability.10,11 Although frequently treated differently and considered as pathophysiologically unrelated events, there is growing evidence to suggest that in some patients with heart failure, both obstructive and central respiratory events might be part of a spectrum of periodic breathing.10,11

The standard measure for the diagnosis of periodic breathing is the apnea-hypopnea index (AHI): the total number of apneas and hypopneas per hour of sleep. Determining the AHI is labor-intensive and requires the simultaneous recording of airflow and oxygen saturation; thoracic and abdominal movements can be used to help distinguish central from obstructive events. Automated methods are available to determine the AHI, but due to the complexity and range of different respiratory events, computer-assisted manual scoring is not yet the clinical standard. Despite being a valuable clinical tool with proven prognostic value,12 when it comes to periodic breathing, the AHI provides no qualitative information regarding the amplitude of the oscillations in the respiratory signal, their distribution, or periodicity. Figure 1 displays a portion of a night record in which periodic breathing is present; it is characterized by the number of apneic/hypopneic events instead of being analyzed as a phenomenon in which the amplitude of the respiratory signal is being modulated. Literature regarding alternative tools for the quantification of periodic breathing is scarce. Of the 32 papers mentioning quantification of periodic breathing in Thomson Reuters's “Web of Science” database, only 11 applied or suggested a measure other than AHI.13–23 In 6 of the 11 papers, an indirect measure quantifying the grade of instability of the chemoreceptor feedback loop is used.14,15,17–19,21 In the remaining 5 papers, Small et al.16 estimate the fractal dimension of the respiratory flow, and both Millar20 and Pinna6,13,24 use a modulating index. The former is based on the relationship between lower and higher frequency components of the respiratory signal, and the latter on a measure of the changes in amplitude of the instant tidal volume during periodic breathing. Both methodologies do require a previous phase in which the existence of periodic breathing is validated to be later quantified by an index.

Figure 1.

Figure 1

Open in a new tab

Screenshot of the standard sleep disordered breathing scoring using a proprietary software. A portion of night record scored by a sleep expert. Airflow (nose) in arbitrary units. Thoraco-abdominal motion (thorax and abdomen) in arbitrary units. Oxygen saturation (SaO2) in percentage. In dark over the nose signal, each of the respiratory events. Each vertical line represents 15 seconds.

Our aim was to develop a simple and robust method to both detect the presence of periodic breathing and to quantify the amplitude of the oscillations in the breathing pattern at any given time. We wanted the method to be able to be applied to flow or thoraco-abdominal motion signals from a variety of different devices and be independent of calibration of the device used. As such, it was to be widely applicable beyond methodological considerations of hardware, without the need for initial validation of the existence of periodic breathing or its characterization. In this paper we describe development and validation of an index, which we term estimated amplitude modulating index (eAMI) that addresses this aim.

METHODOLOGY

The behavior of the respiratory signal during periodic breathing resembles an amplitude-modulated signal (AM). Therefore, by comparing the amplitudes of the modulation and that of the respiratory signal (flow or thoraco-abdominal motion), periodic breathing can be characterized, the key component being the characterization of the amplitude of the modulating signal. We propose a new quantification based on the amplitude of the oscillation exhibited during periodic breathing. This new index was applied to two different data-sets. The first was a large polysomnography dataset acquired at the Antarctic base Concordia. Because of the prevailing chronic hypobaric hypoxia (due to the corrected altitude of approximately 3,800 meters), Concordia provides a unique environment for the study of periodic breathing.2–5,25 The second dataset was acquired at sea level in a controlled clinical environment following the international guidelines of the American Academy of Sleep Medicine (AASM 201226; ICSD 201427).

Development of the Index

eAMI

We characterized the modulating wave exhibited during periodic breathing by using a standard AM demodulation scheme28 by low-pass filtering the absolute value of the respiratory signal, band-pass filtered in the range of quiet breathing in adults, i.e., 0.125 Hz to 0.4 Hz.29 Let the respiratory signal be RS[n], and RSresp[n] be the respiratory signal band-pass filtered at respiratory frequencies, and finally, RSam[n] the envelope of RSresp[n]. Both signals are characterized by the amplitude of the sine wave that would exhibit the same energy E[n]:

graphic file with name aasm.38.3.381a-e1.jpg

where Eresp[n] and Eam[n] denote the energies of RSresp[n] and RSam[n], respectively. Knowing that the square root of the energy of a sinusoidal wave is proportional to its amplitude, we can define an estimated amplitude signal eA[n] for a given respiratory signal RS[n] during an interval N as:

graphic file with name aasm.38.3.381a-e2.jpg

where eAam[n] and eAresp[n] denote the estimated amplitudes of both the modulating and the respiratory signals computed over a moving window of length N. We finally propose a modulating index, the estimated amplitude modulating index (eAMI) as:

graphic file with name aasm.38.3.381a-e3.jpg

It must be noted that comparing either the energies (Eresp[n] and Eam[n]) or the equivalent amplitudes (eAam[n] and eAresp[n]) is fundamentally the same approach, as it only implies a change in the scale. The eAMI takes values around one when both equivalent amplitudes are the same, i.e., around apneas, and takes negative values in the absence of periodic breathing. Figure 2 illustrates a possible scheme for the computation of eAMI[n].

Figure 2.

Figure 2

Open in a new tab

Scheme proposed for computing eAMI.

Detection of Periodic Breathing and Loop Gain Estimation

As eAMI[n] is a signal that qualifies periodic breathing in terms of the amplitude of the oscillations of the respiratory signal, we can use it to both quantify and detect periodic breathing. We consider the possible existence of a periodic breathing event if eAMI[n] remains above a certain threshold θC during at least a certain duration, L (expressed in number of samples). Since the gold standard AHI quantifies the number of apneic/hypopneic events per hour during one night's record, to be able to compare both the AHI and the eAMI, we defined the clinical periodic breathing index (cPBI) as the ratio of the total recording time and the sum of the length of the events longer than L during which eAMI[n] is above θC. This measure is analogous to the frequently used percent of time spent during periodic breathing but estimated from the eAMI. A second threshold θSC could also be defined as the level of oscillation at which the expert can observe an oscillatory behavior on the respiratory signal, but where it does not reach the reduction in thoraco-abdominal motion ≥ 50%, which is necessary for it to be considered as a central event.26

Finally, given that the amplitude of the oscillations during periodic breathing is thought to vary with the loop gain of the respiratory control, we tested whether the eAMI was a good estimator of the instability of the respiratory system. For that, the average loop gain for each of the files was calculated as proposed by Sands et al.19 and then compared against the average eAMI for each of the files.

Artifact Management

The energy from artifacts and spiky respiratory events such as sighs might spread through both the modulating and the respiratory frequency bands. This can cause the eAMI at some point to take values above zero despite the absence of periodic breathing. Short-time false positives should be easy to reject by the requirement of a minimal time during which the eAMI has to remain above the threshold θC. For simplicity, we will set L = 2N, i.e., require an event to be at least twice as long as the length of the window used to compute the signal energy. Longer duration artifacts related with signal loss or subject movements, or an accumulation of sighs during a period of time could still cause false positives. Although it is recommended to remove them from the signal, we kept them in our study to further support our claim of simplicity of the method in terms of requirements for preprocessing.

Validation of the Index

Datasets

Antarctica Cohort: To validate the proposed methodology, we performed an analysis on polysomnography recordings in which periodic breathing without additional comorbidities was expected. For this, a dataset of polysomnography recordings (around 518 h of data) were taken at the Antarctic base Concordia. Because of its location (at an approximate equivalent altitude of 3,800 m), Concordia provides a unique environment for the study of periodic breathing due to high altitude in otherwise healthy individuals.30 Participants arrived during the preceding Antarctic summer, and remained at the Concordia Station for the Antarctic winter, on average spending 13 months at altitude. Thirteen members of the 14-person all-male crew of one winter campaign participated in the experiment (age = 39 ± 9.8 years; BMI = 24.2 ± 2.2; 3 smokers). None of the subjects reported a history of significant medical conditions. Thoraco-abdominal motion was monitored by inductance plethysmography with transducers placed around the participant's chest and abdomen. We recorded the electroencephalogram, electrooculogram, electrocardiogram, and electromyogram from surface electrodes. Oxygen saturation and heart rate were also derived from an oximeter. Data collection was programmed throughout the entire overwintering campaign with a periodicity of 6 weeks. Each of the subjects had on average 7 scheduled repeated measures during the campaign, including one habituation night at the beginning of their campaign. From the total 91 scheduled measures, 74 files including habituations nights were finally acquired and used in this study. The remaining 17 measures were not successful due to device failure or malfunction of the respiratory inductance plethysmography bands (n = 8), and because occasionally some of the participants declined a scheduled measure (n = 9). A sleep expert analyzed all recordings following the AASM (2012) criteria for respiratory event scoring.26 Hypopnea was defined as a reduction in thoraco-abdominal motion > 50% compared to the preceding 2 minutes baseline; apnea was defined as a reduction > 90%. Only events lasting ≥ 10 s and accompanied by an oxygen desaturation > 3% were taken into account.26 Due to the lack of nasal airflow signal, discrimination between central and obstructive apneas was done by testing for the occurrence of paradoxical breathing: breathing movements caused by obstruction in which the rib cage moves in on inspiration and out on expiration, in reverse of the abdomen. Periodic breathing was present at a clinically significant level in most of the recordings (AHI = 65.4 ± 14.55).

Clinical Cohort: To validate the proposed method with signals that were acquired in a clinical setting, a set of 39 polysomnographic recordings from 2 different clinical environments was used. The dataset was used to test both the proposed method's ability to detect periodic breathing at subclinical levels, i.e., in records where no CSA events were identified by the sleep expert, and also at clinical levels in records with very high indices of CSA and/or CSA events. For this, a dataset of polysomnography recordings from 39 different patients suffering from diverse sleep related pathologies was used. Signals were collected in controlled clinical environments in accredited sleep laboratories in both Brussels (∼ 60 m altitude) and Ljubljana (∼ 300 m altitude). Full polysomnography recordings included nasal airflow, thoraco-abdominal motion, electroencephalogram, electrooculogram, electrocardiogram, electro-myogram, leg movements and oxygen saturation. A certified expert somnologist analyzed all recordings following international guidelines.26 Detailed PSG characteristics for both cohorts are reported in Table 1.

Table 1.

PSG characteristics.

graphic file with name aasm.38.3.381.t01.jpg

Open in a new tab

Validation Protocols

Optimum θC Threshold: To estimate a candidate value for θC, we made use of the fact that high levels of periodic breathing were expected in the Antarctic cohort. Indeed, once the dataset was scored, the sleep expert corroborated that high AHI levels were mainly due to periodic breathing. We thus looked for the value of θC that maximized the correlation between the estimated clinical periodic breathing index cPBI and the AHI for each of the records in the dataset. To estimate the optimum θC, we divided our complete set of 74 high altitude recordings in 10,000 randomized groups of 30. We used one of the randomized groups to find an optimum θC. We then chose θC to maximize the correlation between the AHI and the resulting cPBI. We computed the correlation between the AHI and the resulting cPBI for each of the randomized groups. The variance of the correlation among the 10,000 groups gave an idea of the sensitivity of θC as a function of particular signals. This optimum θC was then tested against the second cohort.

Subclinical Periodic Breathing Detection Performance

To test if the proposed method was able to unmask undetected apparent episode of periodic breathing, for each of the files from the clinical cohort a pruned signal including only periods in which the sleep expert did not score any events was obtained. All events with values of eAMI above a threshold θC and lasting for at least L were preselected. Each of these events was visually inspected by an expert and classified into 3 different groups: Non-PB events or false positives, periodic breathing events with erratic or more complex variations of the respiratory signal, and finally periodic breathing events with variations of the respiratory signal > 50%. It is important to note that this criterion to classify periodic breathing events was established solely on the amplitude of the oscillation of the respiratory signal without taking into account any of the other requirements suggested by the AASM guidelines for measuring sleep-related breathing disorders in adults.26

Implications of the Source of Respiratory Signal Used for Computing eAMI

To further demonstrate our claim that the eAMI works independently of the type of source signals, where possible, eAMI was calculated from both the nasal air flow and thoraco-abdominal motion signals and the results compared.

Statistics

Correlations were computed using two-tailed, bivariate correlations (SPSS v 17.0, Chicago, IL, USA) at P < 0.05 level of significance. Data are expressed as means ± standard deviations, with 95% CI for effects of interest.

RESULTS

Characterizing Periodic Breathing by the Amplitude of the Oscillation of the Respiratory Signal

In Figure 3, we show the behavior of the eAMI on a portion of a respiratory signal where periodic breathing was present. This figure illustrates both the temporal resolution and the descriptive value of the proposed method. The eAMI signal qualifies periodic breathing in terms of the amplitude of the oscillation of the respiratory signal, which changes over time. For an expert it was easy to observe oscillations on the respiratory signal with positive values of eAMI, whereas with negative values of eAMI, oscillations were not relevant enough or discernable. Therefore, we chose zero to be a good compromise value for θSC. The eAMI apneic threshold, i.e., the value of eAMI for which apneas start to be observed by an expert, was approximately 0.9.

Figure 3.

Figure 3

Open in a new tab

Fragment of respiration with different levels of PB and the value of eAMI for the given period. The solid line shows the respiratory signal LV. The dotted and dashed lines represent eAMI as defined in equation 3 (see methodology section of manuscript) for N = 100 and N = 50 respectively. In both we did not use any lower threshold θ'.

The effect of the length N of the window can also be observed in Figure 3 and is further discussed in the following section.

In non-preprocessed signals, we observed that the eAMI could produce false positives in the presence of some long-duration artifacts or in the presence of an accumulation of sighs during a period of time > L. While single events where rejected by the previously mentioned minimum event length requirement, an accumulation of them during a period of time could induce the eAMI to take higher values.

Optimum θC and Periodic Breathing Detection Performance

Results for the obtained average correlations between the AHI and the cPBI for a varying θC are depicted in Figure 4. We found θC = 0.65 to be the optimal value, achieving a maximal mean correlation coefficient of 0.89 ± 0.018 (P < 0.01) between the cPBI and the AHI. We can also observe that the correlation between the cPBI and the AHI does not depend on the combination of signals used, as the variance of that correlation is smaller than < 2%. When using this threshold, the correlation between the central apneic events (CAI) and the cPBI for the whole dataset was r = 0.95 (P < 0.001). A scatterplot of the data is shown in Figure 5. We also observed that the length of the window N in which the eAMI was computed was not a critical parameter. In Figure 6, we can observe the obtained average correlations between the AHI and the cPBI in the Antarctica cohort for a varying N and fixed θC = 0.65. For N > 40 s, the average correlation between the AHI and the cPBI remains practically unchanged.

Figure 4.

Figure 4

Open in a new tab

Average correlation coefficient between AHI from a dataset scored by a professional sleep expert and proposed cPBI. Dataset was divided in 10,000 randomized groups of 30 nights. For each of the group the correlation between the AHI gold standard and proposed method was calculated and averaged for different values of θC. On top, the average correlation factor with the standard deviation error. On the bottom, a detail view of the pessimistic correlation (ρ-sd) as a function of θC.

Figure 5.

Figure 5

Open in a new tab

Correlation between central apnea-hypopnea index (CAI) and clinical periodic breathing index (cPBI) computed using eAMI. Plot shows for each of the records both the scored CAI and the cPBI.

Figure 6.

Figure 6

Open in a new tab

Average correlation coefficient between AHI from a dataset scored by a sleep expert and proposed cPBI as a function of L. The dataset was divided in 10,000 randomized groups of 30 nights. For each of the group the correlation between the AHI gold standard and proposed method was calculated and averaged for different values of L.

Using the same optimum θC = 0.65, the correlation between the AHI and the cPBI in the clinical cohort was r = 0.89 (P < 0.001). When using this threshold, the correlation between the CAI and cPBI for the whole clinical dataset was r = 0.79 (P < 0.001).

eAMI as a Predictor of the Loop Gain

The correlation between the average eAMI and the loop gain for the combined clinical and Antarctica cohorts was r = 0.58 (P < 0.001). If only using the clinical cohort, the correlation was r = 0.8 (P < 0.001) when using the nasal airflow signal for the computation of eAMI or r = 0.76 (P < 0.001) when using the thoraco-abdominal motion instead. A scatter plot of the data is shown in Figure 7.

Figure 7.

Figure 7

Open in a new tab

Correlation between eAMI and loop gain (LG). On the left, plot shows for each of the records both the average LG and eAMI. On the right, just for the clinical cohort.

Subclinical Periodic Breathing Detection Performance

In the clinical cohort, there were 93 events identified as periodic breathing events by the proposed method, but not by the expert when using the nasal airflow signal for the computation of eAMI; 91 when using the thoraco-abdominal motion one. These 2 missing events were due to the fact that during their occurrence, the thoraco-abdominal motion signal was corrupted. Events were visually classified by the sleep expert using the proposed classification. Of these, 29 were periodic breathing events with erratic or more complex variations of the respiratory signal and 50 were periodic breathing events with variations of the respiratory signal > 50%. In the remaining 14 events (false positives), different types of respiratory instabilities, for example, an accumulation of sighs, were observed in a short period of time, but none of them resembled periodic breathing.

In the Antarctic data, the number of non-scored periodic breathing events which still showed above θC were numerous. From the approximately 518 h of data analyzed, a portion of 90 h had mild periodic breathing. These were events in which the amplitude of the oscillation of the respiratory signal was still > 50%, but was still not considered as periodic breathing by the expert.

DISCUSSION

The main goal of our study was to provide a new tool for the early diagnosis of nocturnal periodic breathing. The prevalence of undiagnosed SDB, a condition of repeated episodes of apnea and hypopnea during sleep, is high among adults.31 Several studies have demonstrated a direct link between SDB and mortality.32,33 SDB is likely to be a risk factor for hypertension, which might result in increased cardiovascular morbidity in the general population.34,35 SDB has also been associated with impaired daytime performance in children,36 while drowsiness due to SDB might be the cause of many accidents.37,38 In this paper, we propose an automated, simple and novel methodology to detect and quantify periodic breathing, a form of SDB, based on the estimated amplitudes of the respiratory signal and its envelope. Periodic breathing may be observed in a variety of situations including damage to respiratory centers,1 acute exposure to high altitude1–5 and in patients suffering from chronic heart failure.6–9 Prognosis of heart failure is uniformly poor when treatment of the underlying problem is not initiated as soon as possible; half of patients with a diagnosis of heart failure will die within 4 years, and in patients with severe heart failure the rate of mortality is > 50%.39 As pointed out by Gilmartin et al.,40 although central apneas and severe periodic breathing—including Cheyne-Stokes respiration—are readily recognizable, more subtle forms of periodic breathing are much more difficult to characterize. Understanding the mechanisms behind periodic breathing and developing the necessary screening tools might help to improve early diagnosis and treatment of underlying diseases, including chronic heart failure cases in which period breathing is present.

In terms of quantification and description, we have shown that the eAMI is a continuous signal that qualifies periodic breathing in terms of the amplitude of the oscillation of the respiratory signal. This treats the periodic breathing event as a single process describing its evolution through time instead of treating periodic breathing as an accumulation of single central events, as in the current clinical gold standard AHI. Furthermore, the average value of the eAMI was also correlated with the loop gain of the record (combined cohorts r = 0.58, P < 0.001; clinical cohort r = 0.8, P < 0.001). This descrip -tive capability might help us to further unravel how this phenomenon evolves over time assessing how loop gain changes with sleep state or under the effects of different interventions. Since an increased loop gain in the respiratory control feedback loop has been proposed as possible explanation for periodic breathing,14,41,42 a quantifying tool that is correlated with loop gain might facilitate our understanding of the processes involved in its development and help monitor certain patient categories in clinical settings.

In terms of clinical periodic breathing detection, we have proposed a candidate value for the clinical threshold θC by maximizing the correlation between the fraction of time spent above this threshold (cPBI) with the AHI scored by the sleep expert. We found an optimal value for the clinical threshold θC around 0.65. When using this threshold in the Antarctica cohort, the correlation between the cPBI and AHI was r = 0.89 (P < 0.01) and between the cPBI and CAI was r = 0.95 (P < 0.001). The reason for this is that in this cohort the majority of the events were indeed central. This clinical threshold θC was further tested in the clinical cohort, showing in that case a higher correlation with the AHI (r = 0.89, P < 0.001) than with the CAI (r = 0.79, P < 0.001). These results might at first glance appear contradictory. The explanation, however, is that in the clinical dataset we added 3 recordings from patients showing severe cyclic OSA (AHI 39.64, 92.87, and 97.35), in which the proposed method detected a strong periodicity resembling that of periodic breathing. Although this could be considered a limitation of the method, it should be noted that in these 3 patients, prior to the block of airflow one could easily observe the consecutive crescendo and decrescendo cycles characteristic from periodic breathing. This raises the question whether it is appropriate to treat the patients as exclusively exhibiting classical OSA, or whether other abnormalities, including CSA, might have contributed to their breathing patterns. This notion has recently been considered by several research groups, such as Tkacova and colleagues, who concluded that in some patients with heart failure, OSA and CSA are part of a spectrum of periodic breathing that can shift over time.10,43 In another study by Hoffman and Schulman on the appearance of CSA after treatment of OSA, the authors indicate that there is evidence that many laboratories diagnose patients shown to have mixed apneas with OSA, treating these events as if they were obstructive, when in fact they pathophysiologically may be closer related to CSA.11 The authors suggest that by eliminating the contributing obstructive component after CPAP, therapy could expose central hypopneas and centrally originating mixed apneas that were classified as obstructive from the initial PSG. We therefore suggest that the eAMI is capable of detecting this periodical instability beyond CSA, or OSA cases. Once periodic breathing has been identified, the estimated loop gain by the eAMI, or its periodicity, might open a new window of information. For instance, short cycles of periodic breathing are believed to be caused by peripheral chemoreceptors, while longer ones would have a more central origin.

In terms of subclinical periodic breathing, we have also validated the usefulness of the proposed algorithm in both detecting and in quantifying periodic breathing events that were not scored at first by the sleep expert. Therefore, not only can we assert that the proposed method is able to detect periodic breathing at clinical levels and estimate the AHI, but it appears that it may also help in detecting and quantifying periodic breathing at subclinical levels. This can be particularly useful in cases in which the sleep expert has to reject episodes of apparent periodic breathing in which the duration of the central event was not long enough to be considered an apnea or a hypopnea by conventional scoring rules. Given the impaired prognosis associated with periodic breathing, such ability to detect both clinical and subclinical levels holds promise as a valuable tool for early diagnosis.

Modulating indices of the respiratory signal or the relationship between lower and higher frequency components of the respiratory signal for detecting or qualifying periodic breathing have already been explored by other authors, e.g., Millar,20 Pinna,6,13,24 and Macey.22 We have shown that the eAMI has some additional benefits, mainly in that it can be used to both detect and quantify periodic breathing, but also in terms of simplicity. Finally, of all algorithms in the aforementioned publications, only that of Pinna6,13,24 and the eAMI compute the envelope of the signal instead of using the low frequency component. This might be an issue when comparing results from different respiratory signals (nasal airflow and thoraco-abdominal motion), the same respiratory signals but different sensors (e.g., thoraco-abdominal motion measured by inductance plethysmography or strain gauges), or finally, even the same sensors but different device manufacturers. To show that the proposed method could be computed in different respiratory signals, we used different ones in each of the datasets, and when possible, compared. The method of Pinna6,13,24 is perhaps the one that shows the most similarities with the proposed method. However, this method also requires a preliminary phase in which the existence of periodic breathing has to be validated before quantifying it. The reason is that, by definition, the lung volume modulating index (LVMI) proposed by Pinna is the normalized difference between the maximum and the minimum values of the instantaneous tidal volume at each of the periods of periodic breathing. Therefore, LVMI cannot be computed in absence of these periods. Finally, we have shown that it is possible to obtain similar results with a much simpler envelope recovering technique: an AM demodulator. This also avoids the need of a previous phase in which the existence of periodic breathing has to be validated.

The eAMI is a very simple tool for detection and quantification of periodic breathing. It can reduce costs and thus resources needed to screen periodic breathing and it is a good candidate for low-cost monitoring devices. The eAMI is inexpensive in terms of computing resources: routines for the method were programmed in Matlab and the analysis of an entire 518 h of dataset takes around 20 s using a commercial laptop running an Intel Core i7 (2627M@2.7GHz) with 8 GB of RAM. The eAMI quantifies periodic breathing as a single event, where the amplitude of the oscillations of the respiratory signal is the evolving variable. This measure, beyond simply giving a quantitative measure like the AHI, offers qualitative information about periodic breathing and the loop gain associated. The use of the eAMI could thus help achieve a better understanding, and an earlier diagnosis of periodic breathing in both clinical and research environments.

DISCLOSURE STATEMENT

This was not an industry supported study. This work was supported in part by the PRODEX program of the European Space Agency (ESA) managed with the help of the Belgian Science Policy Office (BELSPO). The authors have indicated no financial conflicts of interest.

ACKNOWLEDGMENT

We would like to express our gratitude to the Adult Unit for Sleep of the UZ Brussel and especially Sonia De Weerdt and Aisha Cortoos for sharing polysomnographic data.

REFERENCES

  • 1.Brown HW, Plum F. Neurologic basis of Cheyne-Stokes respiration. Am J Med. 1961;30:849–60. [Google Scholar]
  • 2.Waggener TB, Brusil PJ, Kronauer RE, Gabel RA, Inbar GF. Strength and cycle time of high-altitude ventilatory patterns in unacclimatized humans. J Appl Physiol. 1984:56. doi: 10.1152/jappl.1984.56.3.576. [DOI] [PubMed] [Google Scholar]
  • 3.Khoo MC, Anholm JD, Ko SW, et al. Dynamics of periodic breathing and arousal during sleep at extreme altitude. Respir Physiol. 1996;103:33–43. doi: 10.1016/0034-5687(95)00057-7. [DOI] [PubMed] [Google Scholar]
  • 4.Bloch KE, Latshang TD, Turk AJ, et al. Nocturnal periodic breathing during acclimatization at very high altitude at Mount Muztagh Ata (7,546 m) Am J Respir Crit Care Med. 2010;182:562–8. doi: 10.1164/rccm.200911-1694OC. [DOI] [PubMed] [Google Scholar]
  • 5.Dempsey JA, Veasey SC, Morgan BJ, O'Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90:47–112. doi: 10.1152/physrev.00043.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Pinna GD, La Rovere MT, Robbi E, Maestri R. Assessing the severity and improving the understanding of sleep-related breathing disorders in heart failure patients. Conf Proc IEEE Eng Med Biol Soc. 2010;2010:3571–4. doi: 10.1109/IEMBS.2010.5627463. [DOI] [PubMed] [Google Scholar]
  • 7.Javaheri S. A mechanism of central sleep apnea in patients with heart failure. New Engl J Med. 1999;341:949–54. doi: 10.1056/NEJM199909233411304. [DOI] [PubMed] [Google Scholar]
  • 8.Francis DP, Willson K, Davies LC, Coats AJ, Piepoli M. Quantitative general theory for periodic breathing in chronic heart failure and its clinical implications. Circulation. 2000;102:2214–21. doi: 10.1161/01.cir.102.18.2214. [DOI] [PubMed] [Google Scholar]
  • 9.Bradley TD, Floras JS. Sleep apnea and heart failure - Part II: Central sleep apnea. Circulation. 2003;107:1822–6. doi: 10.1161/01.CIR.0000061758.05044.64. [DOI] [PubMed] [Google Scholar]
  • 10.Tkacova R, Wang H, Bradley TD. Night-to-night alterations in sleep apnea type in patients with heart failure. J Sleep Res. 2006;15:321–8. doi: 10.1111/j.1365-2869.2006.00528.x. [DOI] [PubMed] [Google Scholar]
  • 11.Hoffman M, Schulman DA. The appearance of central sleep apnea after treatment of obstructive sleep apnea. Chest. 2012;142:517–22. doi: 10.1378/chest.11-2562. [DOI] [PubMed] [Google Scholar]
  • 12.Lanfranchi PA, Braghiroli A, Bosimini E, et al. Prognostic value of nocturnal Cheyne-Stokes respiration in chronic heart failure. Circulation. 1999;99:1435–40. doi: 10.1161/01.cir.99.11.1435. [DOI] [PubMed] [Google Scholar]
  • 13.Maestri R, Pinna GD, Robbi E, et al. RESP-24: a computer program for the investigation of 24-h breathing abnormalities in heart failure patients. Comput Methods Programs Biomed. 2002;68:147–59. doi: 10.1016/s0169-2607(01)00180-8. [DOI] [PubMed] [Google Scholar]
  • 14.Younes M, Ostrowski M, Thompson W, Leslie C, Shewchuk W. Chemical control stability in patients with obstructive sleep apnea. Am J Respir Crit Care Med. 2001;163:1181–90. doi: 10.1164/ajrccm.163.5.2007013. [DOI] [PubMed] [Google Scholar]
  • 15.Wellman A, Malhotra A, Fogel RB, Edwards JK, Schory K, White DP. Respiratory system loop gain in normal men and women measured with proportional-assist ventilation. J Appl Physiol. 2003;94:205–12. doi: 10.1152/japplphysiol.00585.2002. [DOI] [PubMed] [Google Scholar]
  • 16.Small M, Judd K, Lowe M, Stick S. Is breathing in infants chaotic? Dimension estimates for respiratory patterns during quiet sleep. J Appl Physiol. 1999;86:359–76. doi: 10.1152/jappl.1999.86.1.359. [DOI] [PubMed] [Google Scholar]
  • 17.Maayan C, Carley DW, Axelrod FB, Grimes J, Shannon DC. Respiratory system stability and abnormal carbon-dioxide homeostasis. J Appl Physiol. 1992;72:1186–93. doi: 10.1152/jappl.1992.72.3.1186. [DOI] [PubMed] [Google Scholar]
  • 18.Khoo MC, Yang F, Shin JJ, Westbrook PR. Estimation of dynamic chemoresponsiveness in wakefulness and non-rapid-eye-movement sleep. J Appl Physiol. 1995;78:1052–64. doi: 10.1152/jappl.1995.78.3.1052. [DOI] [PubMed] [Google Scholar]
  • 19.Sands SA, Edwards BA, Kee K, et al. Loop gain as a means to predict a positive airway pressure suppression of cheyne-stokes respiration in patients with heart failure. Am J Respir Crit Care Med. 2011;184:1067–75. doi: 10.1164/rccm.201103-0577OC. [DOI] [PubMed] [Google Scholar]
  • 20.Millar TW, Hanly P, Kryger MH. Short technical note: quantification of periodic breathing: preliminary studies. Sleep. 1992;15:364–70. doi: 10.1093/sleep/15.4.364. [DOI] [PubMed] [Google Scholar]
  • 21.Nemati S, Edwards BA, Sands SA, et al. Model-based characterization of ventilatory stability using spontaneous breathing. J Appl Physiol. 2011;111:55–67. doi: 10.1152/japplphysiol.01358.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Macey KE, Harper RM, Page WH, Macey PM. Quantifying amplitude modulation of breathing in infants using the wavelet transform. Conf Proc IEEE Eng Med Biol Soc. 2000;1:61–3. [Google Scholar]
  • 23.Varanini M, Mortara A, Raciti M, et al. Detection of apneas during 24-hour ambulatory monitoring of ECG and respiration in chronic heart failure patients with Cheyne-Stokes breathing. Comput Cardiol. 1999;26:459–62. [Google Scholar]
  • 24.Pinna GD, Maestri R, Mortara A, La Rovere MT, Fanfulla F, Sleight P. Periodic breathing in heart failure patients: testing the hypothesis of instability of the chemoreflex loop. J Appl Physiol. 2000;89:2147–57. doi: 10.1152/jappl.2000.89.6.2147. [DOI] [PubMed] [Google Scholar]
  • 25.Eichenberger U, Weiss E, Riemann D, Oelz O, Bartsch P. Nocturnal periodic breathing and the development of acute high altitude illness. Am J Respir Crit Care Med. 1996;154:1748–54. doi: 10.1164/ajrccm.154.6.8970365. [DOI] [PubMed] [Google Scholar]
  • 26.Berry RB, Budhiraja R, Gottlieb DJ, et al. Rules for scoring respiratory events in sleep: update of the 2007 AASM manual for the scoring of sleep and associated events. J Clin Sleep Med. 2012;8:597–619. doi: 10.5664/jcsm.2172. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.American Academy of Sleep Medicine. International classification of sleep disorders. 3rd ed. Darien, IL: American Academy of Sleep Medicine; 2014. [Google Scholar]
  • 28.Oppenheim AV, Schafer RW. Digital signal processing. Prentice-Hall; 1975. [Google Scholar]
  • 29.Phillipson EA. Control of breathing during sleep. Am Rev Respir Dis. 1978;118:909–39. doi: 10.1164/arrd.1978.118.5.909. [DOI] [PubMed] [Google Scholar]
  • 30.Fernandez Tellez H, Mairesse O, Macdonald-Nethercott E, Neyt X, Meeusen R, Pattyn N. Sleep-related periodic breathing does not acclimatize to chronic hypobaric hypoxia: a 1-year study at high altitude in Antarctica. Am J Respir Crit Care Med. 2014;190:114–6. doi: 10.1164/rccm.201403-0598LE. [DOI] [PubMed] [Google Scholar]
  • 31.Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occurrence of sleep-disordered breathing among middle-aged adults. New Engl J Med. 1993;328:1230–5. doi: 10.1056/NEJM199304293281704. [DOI] [PubMed] [Google Scholar]
  • 32.Young T, Finn L, Peppard PE, et al. Sleep disordered breathing and mortality: eighteen-year follow-up of the Wisconsin Sleep Cohort. Sleep. 2008;31:1071–8. [PMC free article] [PubMed] [Google Scholar]
  • 33.Punjabi NM, Caffo BS, Goodwin JL, et al. Sleep-disordered breathing and mortality: a prospective cohort study. Plos Med. 2009;6:9. doi: 10.1371/journal.pmed.1000132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Peppard PE, Young T, Palta M, Skatrud J. Prospective study of the association between sleep-disordered breathing and hypertension. New Engl J Med. 2000;342:1378–84. doi: 10.1056/NEJM200005113421901. [DOI] [PubMed] [Google Scholar]
  • 35.Nieto FJ, Young TB, Lind BK, et al. Association of sleep-disordered breathing, sleep apnea, and hypertension in a large community-based study. JAMA. 2000;283:1829–36. doi: 10.1001/jama.283.14.1829. [DOI] [PubMed] [Google Scholar]
  • 36.Gozal D. Sleep-disordered breathing and school performance in children. Pediatrics. 1998;102:616–20. doi: 10.1542/peds.102.3.616. [DOI] [PubMed] [Google Scholar]
  • 37.Dinges DF. An overview of sleepiness and accidents. J Sleep Res. 1995;4:4–14. doi: 10.1111/j.1365-2869.1995.tb00220.x. [DOI] [PubMed] [Google Scholar]
  • 38.Leger D. The cost of sleep-related accidents - a report for the National Commission On Sleep Disorders Research. Sleep. 1994;17:84–93. doi: 10.1093/sleep/17.1.84. [DOI] [PubMed] [Google Scholar]
  • 39.McMurray J, McDonagh T, Morrison CE, Dargie HJ. Trends in hospitalization for heart-failure in Scotland 1980-1990. Eur Heart J. 1993;14:1158–62. doi: 10.1093/eurheartj/14.9.1158. [DOI] [PubMed] [Google Scholar]
  • 40.Gilmartin GS, Daly RW, Thomas RJ. Recognition and management of complex sleep-disordered breathing. Curr Opin Pulm Med. 2005;11:485–93. doi: 10.1097/01.mcp.0000183061.98665.b0. [DOI] [PubMed] [Google Scholar]
  • 41.Khoo MC. Determinants of ventilatory instability and variability. Respir Physiol. 2000;122:167–82. doi: 10.1016/s0034-5687(00)00157-2. [DOI] [PubMed] [Google Scholar]
  • 42.Edwards BA, Sands SA, Skuza EM, et al. Increased peripheral chemosensitivity via dopaminergic manipulation promotes respiratory instability in lambs. Respir Physiol Neurobiol. 2008;164:419–28. doi: 10.1016/j.resp.2008.09.003. [DOI] [PubMed] [Google Scholar]
  • 43.Tkacova R, Niroumand M, Lorenzi G, Bradley TD. Overnight shift from obstructive to central apneas in patients with heart failure - Role of Pco(2) and circulatory delay. Circulation. 2001;103:238–43. doi: 10.1161/01.cir.103.2.238. [DOI] [PubMed] [Google Scholar]

RESOURCES