ADVERTISEMENT

Issue Navigator

Volume 14 No. 10
Earn CME
Accepted Papers





Scientific Investigations

Effects of Increased Pharyngeal Tissue Mass Due to Fluid Accumulation in the Neck on the Acoustic Features of Snoring Sounds in Men

Shumit Saha, MSc1,2,5; Zahra Moussavi, PhD1; Peyman Hadi, MSc3; T. Douglas Bradley, MD2,4; Azadeh Yadollahi, PhD2,5
1Department of Biomedical Engineering, University of Manitoba, Winnipeg, Manitoba, Canada; 2Toronto Rehabilitation Institute-University Health Network, Toronto, Ontario, Canada; 3BresoTec Inc., Toronto, Ontario, Canada; 4Department of Medicine, University of Toronto, Toronto, Ontario, Canada; 5Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada

ABSTRACT

Study Objectives:

Snoring sounds are generated by the vibration of pharyngeal tissue due to the upper airway narrowing. While recorded by a microphone placed over the neck, snoring can pass through the pharyngeal tissue surrounding the upper airway. Thus, changes in the pharyngeal tissue content may change the acoustic properties of the snoring sounds. Rostral fluid shift and the consequent increases in neck fluid volume (NFV) and neck circumference (NC) can increase pharyngeal tissue mass. Therefore, the goal of this study was to investigate the relationship between increases in pharyngeal tissue mass, as assessed by increased NFV and NC, and snoring sounds features.

Methods:

We obtained data from a previous study where 20 males who were not obese participated in a daytime polysomnography and their NC and NFV were measured before and after sleep. During sleep, snoring sounds were recorded with a microphone placed over the neck. Spectral centroid of the snoring sounds was estimated. Then, the first five snoring segments were selected from the first and last 30 minutes of stage N2 sleep.

Results:

We found a significant decrease in the snoring spectral centroid from the beginning to end of sleep. We also found that spectral centroid from the end of sleep in frequency ranges below 200 Hz was inversely correlated with the increases in NFV and NC from before to after sleep.

Conclusions:

These results suggest that snoring spectral centroid can be used as a noninvasive and convenient method to assess variations in the pharyngeal tissue mass.

Citation:

Saha S, Moussavi Z, Hadi P, Bradley TD, Yadollahi A. Effects of increased pharyngeal tissue mass due to fluid accumulation in the neck on the acoustic features of snoring sounds in men. J Clin Sleep Med. 2018;14(10):1653–1660.


BRIEF SUMMARY

Current Knowledge/Study Rationale: Increased pharyngeal tissue mass, either by increased pharyngeal fat or fluid, can narrow the upper airway and worsen the severity of obstructive sleep apnea. Narrowing of the upper airway can cause vibrations in the pharyngeal tissue and induce snoring. We aim to investigate the effects of increased pharyngeal tissue mass on the acoustic features of snoring sounds.

Study Impact: Increased pharyngeal tissue mass caused by increased fluid was inversely correlated with the spectral centroid of snoring sounds. Our results have demonstrated the potential application of the spectral centroid of snoring sounds to access the variations of the pharyngeal tissue mass.

INTRODUCTION

Obstructive sleep apnea (OSA), a common respiratory disorder in approximately 10% of adults,1 is associated with an increased risk of the development of hypertension,2 stroke,3 and heart failure,4,5 as well as vehicle and work-related accidents.68 OSA is the result of the partial (hypopnea) or complete (apnea) cessation of airflow due to either partial or complete collapse of the upper airway.9 However, the mechanisms of upper airway collapse during sleep are not completely understood.10 We have shown that increased tissue content in the neck, either by the increased pharyngeal fat or edema, contributes to the upper airway collapsibility and worsening OSA severity.11,12

Previous studies have demonstrated that due to gravity and sedentary living, fluid accumulates in the legs while a patient is upright during the daytime.13 When lying supine at night, part of the fluid accumulated in the legs will shift rostrally into the neck.14 As a result, neck fluid volume (NFV) increases in association with an increase in neck circumference (NC) and a decrease in the upper airway cross-sectional area (UA-XSA),15 which consequently increases the upper airway resistance16 and collapsibility.17 Furthermore, a previous study showed that in men who were at least 40 years old with no or mild to moderate OSA, intravenous infusion of physiological saline (0.9% of NaCl) led to inducement or worsening of OSA severity as evidenced by a threefold increase in the occurrence of apneas and hypopneas per hour of sleep (as assessed by apnea-hypopnea index, AHI).18 Conversely, fluid removal by ultrafiltration in hemodialyzed patients with renal failure, exercise, and wearing compression stockings during the day were associated with reductions in overnight rostral fluid shift and 25% to 40% reductions in the AHI of patients with OSA.1921 These results provide strong evidence that increased NFV during sleep can increase the upper airway narrowing and increase the risk of the development or worsening of OSA. Therefore, development of noninvasive, convenient, and portable techniques to measure pharyngeal fluid content and monitoring its effects on the upper airway size could improve our understanding of the mechanisms of upper airway collapse in OSA.

The common methods to measure NFV or edema in the neck are based on magnetic resonance imaging (MRI) or bioelectrical impedance analysis,18 which are inconvenient and difficult to perform continuously during sleep. However, respiratory or snoring sound analysis is a simple and noninvasive method to investigate variations in the upper airway size and physiology during sleep.22,23 Turbulent air flow through a narrow upper airway causes vibration of the pharyngeal tissue and induces snoring.24 These vibratory sounds can be recorded by a microphone placed on the neck over the trachea. Thus, acoustic features of snoring sounds can change not only with changing upper airway size but also with changes in the composition of tissues surrounding the upper airway. Although snoring sounds analysis has been used to investigate the relationship with the upper airway narrowing23,25,26 or sites of the upper airway obstruction,2729 its relationship to changes in pharyngeal tissue content surrounding the upper airway during sleep has not been investigated before.

This study aims to investigate the effects of increased pharyngeal tissue content during sleep, as assessed by increases in NC and NFV, on the acoustic features of snoring. Based on the physics of sound propagation through a tissue, we hypothesize that increases in the volume of neck fluid will increase the tissue mass surrounding the upper airway, which consequently will decrease the spectral frequencies of snoring sounds. To investigate this hypothesis, we have proposed a spring and mass model for the pharyngeal tissue surrounding the upper airway, and investigated the effects of increased tissue mass due to increased NFV on snoring sounds features.

METHODS

Data Measurement

Participants

This study is a retrospective study and we used data that was recorded as part of a randomized, double crossover study to investigate the effects of fluid overloading by saline infusion on sleep apnea severity in men.18 Data from this study was extracted from the control study arm of the original study, during which the amount of saline infusion was minimal (less than 100 mL). The study was approved by the research ethics board of the Toronto Rehabilitation Institute and all participants gave written consent before participation. The inclusion criteria of the original study were men who had no issues with high body weight with a body mass index (BMI) < 30 kg/m2, a Berlin questionnaire score ≤ 1, and blood pressure ≤ 140/90 mmHg. The exclusion criteria were individuals with a history of cardiovascular, renal, neurological, or respiratory diseases or taking any medication for these diseases, and dominant central sleep apnea.18,23

Polysomnography

Daytime polysomnography was performed, and participants slept in the supine position only on a single pillow. Polysomnography was used to detect sleep stages and sleep apnea severity assessed by AHI. Apneas and hypopneas were defined and classified in accordance with the American Academy of Sleep Medicine standard.30 Elaboration of details regarding the experimental protocol and sleep studies can be found in our previous work.18,23

NC, UA-XSA, and NFV Measurement

NC and UA-XSA were measured before and after sleep using a measuring tape and acoustic pharyngometry,31 respectively. NC was measured at the level of cricothyroid cartilage before and after sleep and the same level was ensured by drawing a line. In this study, NFV was estimated based on the bioelectrical impedance measurement of the neck, as follows:

jcsm.14.10.1653a-e1.jpg
where L is the neck length, C is the NC, R is the neck resistance estimated from the bioimpedance measurement (using MP150 Biopac System and EBI100C module), and ρ is the fluid resistivity. To measure bioimpedance, four electrodes were attached to the skin. The voltage drop across the length of the neck was measured by two electrodes attached on the right side of the neck under the right ear and at the base of the neck. Two additional electrodes were fixed 1 inch apart from the voltage measuring electrodes. These two electrodes were used to inject a low amplitude (400 μA) current at 50 kHz. At the beginning of the study, the neck length (L in equation 1) was measured with a tape as the distance between the voltage measuring electrodes, while the participants were standing and their head was in the neutral position. Details of this method are discussed in our previous work.32

Breath Sound Recordings

During sleep, breath sounds were recorded with a microphone (Sony ECM-44B) placed over the suprasternal notch using double-sided tape. The sounds were low-pass filtered (cutoff frequency of 5 kHz) and digitized (sampling frequency of 12.5 kHz) by Biopac DA100C and MP150 Biopac System, respectively.

Snoring Sounds Annotation

Although there is no universally accepted definition of snoring, we manually detected snoring sounds by listening to the signals and observing them in the time-frequency domains to ensure vibratory characteristics of the snoring sounds using a digital audio software (Praat, Version 5.4.08)33 (Figure 1). Because snoring during the inspiratory phase is more common than snoring during the expiratory phase, we only investigated the inspiratory snoring segments. After manual segmentation of snoring sounds, different features in the time and frequency domains were extracted.

Segmentation of snoring sounds.

(A) 13 seconds of breathing segment with normal breathing and snoring. (B) Spectrogram of the segment presented in (A). The color codes represent the sound intensity at different frequencies, when red and blue are associated with higher and lower intensities.

jcsm.14.10.1653a.jpg

jcsm.14.10.1653a.jpg
Figure 1

Segmentation of snoring sounds.

(more ...)

Feature Extraction

In time domain, we calculated two features: snoring percentage and snoring time index. Snoring percentage was calculated as the number of snoring segments in each sleep stage divided by the total number of snoring segments in the entire sleep period. Snoring time index was calculated for the total sleep period and every sleep stage. Snoring time index for total sleep time was calculated as the total snoring time duration during the entire sleep period divided by the total sleep time. Furthermore, snoring time index in each sleep stage was calculated as the total snoring time in each sleep stage divided by the time spent in that sleep stage.34

To calculate the spectral features, snoring segments were preprocessed by a band-pass filter in the frequency range of 30–4000 Hz to eliminate the effects of low- and high-frequency noises.35 We estimated power spectral density (PSD) of each snoring segment based on the Welch method using a Hamming window of 100 ms and 50% overlap between adjacent windows. From the PSD, we calculated the average power and spectral centroid of all snoring segments. Average power represents the intensity of the snoring sounds. Spectral centroid is the frequency that represents the weighted average of the PSD frequencies within a frequency band.36 The equations of average power and spectral centroid are provided as follows:

jcsm.14.10.1653a-e2.jpg
jcsm.14.10.1653a-e3.jpg
where, Pavg is average power, f1 is lower band frequency, fu is higher band frequency, and P(f) is the estimated PSD.

Both the average power and spectral centroid were calculated for the entire frequency band (30–4000 Hz). In addition, spectral centroid was calculated for several subbands including 60–140 Hz, 60–170 Hz, 60–200 Hz, 150–450 Hz, 450–600 Hz, 600–1200 Hz, 1200–1800 Hz, 1800–2500 Hz, and 2500–4000 Hz.23,37 Previous studies have shown that the main vibratory frequency of the snoring sounds lies in 64– 135 Hz.38 Snoring sounds pass through upper airway and the surrounding tissue of the neck. Therefore, not only the upper airway geometry but also the tissue properties of the neck can affect the characteristics of the snoring sounds. Additionally, previous studies have shown that low-frequency components of lung sounds (< 300 Hz) tend to couple with the tissue compartment and propagate through the tissue.39 However, several studies reported that the first formant frequency of the snoring sounds, which is mainly affected by the upper airway shape, can be approximately 300 Hz.28,40,41 Therefore, to incorporate the effects of sound propagation through tissue with minimum interference from the effects because of the shape of the upper airway, we chose to use spectral centroid for our frequency analysis with subbands more densely in lower frequency ranges.

We investigated the changes in spectral centroid of snoring sounds from the beginning to the end of sleep for each study participant. To accomplish this task, we calculated the average spectral centroid of the first 5 snoring segments during the first 30 minutes of stage N2 sleep (first N2) void of hypopneas. Similarly, we calculated the average spectral centroid of the first 5 snoring segments in the last 30 minutes of stage N2 sleep (last N2) void of hypopneas. Afterward, we compared the differences between the average spectral centroids of these two periods of sleep.

For each subject, we also calculated the average of spectral centroid and Pavg for all the snoring segments during the entire sleep and also for each sleep stage separately.

Spring and Mass Model for the Surrounding Tissues of the Upper Airway

If a force is applied to a system that consists of a mass (m) attached to a spring with elasticity (K), the mass exhibits a simple harmonic motion, with the natural frequency of42:

jcsm.14.10.1653a-e4.jpg
Therefore, the natural frequency of a system depends on its elasticity and mass. In this study, pharyngeal tissue around the upper airway was modeled as a mass and spring system (Figure 2), where K and m were the elasticity and mass of the pharyngeal tissue, respectively. The mass of the pharyngeal tissue was calculated as m = d × V where d is the density and V is the volume of the surrounding tissues of the upper airway including fat, fibers, muscles, and fluids.

Spring and mass model.

The spring and mass system of surrounding tissues of the upper airway. Here, K is the elasticity and m is the mass of the tissue.

jcsm.14.10.1653b.jpg

jcsm.14.10.1653b.jpg
Figure 2

Spring and mass model.

(more ...)

Therefore, the frequency of the tissue vibration due to snoring sounds propagation was derived as:

jcsm.14.10.1653a-e5.jpg
We assumed that from the beginning to the end of sleep, the only factor that would change the tissue volume surrounding the upper airway is the changes in NFV. Therefore, equation 4 can be rewritten as:
jcsm.14.10.1653a-e6.jpg
where, Δω and ΔV are the changes in snoring frequency and tissue volume due to NFV from beginning to the end of sleep. The other factors that may change due to increased NFV are K and d. However, it is known that the density of water is changed with the temperature and the amount of salt in the water.43 In a typical individual, temperature falls up to 2°F during sleep and rises upon waking.44 However, these changes in temperature are not significant enough to change the density of the water. Also, the amount of salt content cannot change due to increased water content in the neck. Therefore, we may assume that density of the tissue content (d) due to increased fluid would remain constant. Furthermore, previous studies have shown that the elasticity of collagen tissue at low stretch does not change with the increased water content in the tissue.45 During sleep, the pharyngeal tissue is under no stretch, and therefore, we assumed that increased fluid in the neck during sleep may not change the elasticity of the pharyngeal tissue (K) in equation 5.

Statistical Analysis

The normality of the data was assessed by Kolmogorov-Smirnov test. The changes in NC, UA-XSA, and NFV from before to after sleep along with the changes in spectral centroid from the beginning to the end of sleep were assessed by paired t-tests or Wilcoxon rank-sum test for normally and non-normally distributed data, respectively. The changes in measured snoring sounds features between different sleep stages were investigated by the analysis of variance and the post hoc Tukey test. Correlations between the snoring sounds features and changes in the NFV and NC were investigated by Pearson or Spearman rank coefficient, for normally and non-normally distributed data, respectively. Correlations between changes in spectral centroid from the beginning to the end of sleep and changes in NFV or NC from before to after sleep were also performed by Pearson or Spearman coefficient based on their normality. Statistical analyses were performed by R (version 3.3.2) and two-tailed P < .05 was considered significant. Data are presented as mean ± standard deviation.

RESULTS

Characteristics of the Participants

Twenty men (age: 45.95 ± 11.09 years; BMI: 25.57 ± 2.93 kg/m2) involved in our previous study23 met all inclusion criteria and were included in this study. Participants slept for an average of 148.5 ± 45.1 minutes. Participants spent most of the time in stage N2 sleep (Table 1, 59.82 ± 13.00%). The average AHI of the participants was 26.62 ± 24.79 events/h. Details of the baseline characteristics and sleep structure of the participants are given in our previous work.23

Percentage of sleep and time domain features of snoring.

jcsm.14.10.1653.t01.jpg

table icon
Table 1

Percentage of sleep and time domain features of snoring.

(more ...)

NC, UA-XSA, and NFV Measurement

From before to after sleep, there were significant increases in NC (ΔNC: 0.58 ± 0.32 cm, P < .001; 41.83 ± 2.72 cm to 42.42 ± 2.72 cm) and NFV (ΔNFV: 18.56 ± 6.72 mL, P < .001; 266.37 ± 47.93 mL to 284.93 ± 49.43 mL) and decreases in UAXSA (ΔUA-XSA: −0.37 ± 0.23 cm2, P < .001; 2.63 ± 0.61 cm2 to 2.26 ± 0.55 cm2).

Snoring Features

An average of 342 ± 223 inspiratory snoring segments were manually extracted from each participant (133.02 ± 93.76 snoring segments per hour of sleep). The number of snoring segments in stage N2 sleep was significantly higher than that in other sleep stages (P < .001). However, the snoring time index was similar for different sleep stages (Table 1, P > .10). Similarly, snoring time index was similar in both non-rapid eye movement (NREM) sleep and rapid eye movement (REM) sleep (P > .10). Furthermore, no significant correlations were found between the time domain features of snoring sounds and either baseline or changes in UA-XSA or NFV.

The spectral centroid of snoring sounds decreased significantly from the first N2 to the last N2 period at 60–140 Hz (Δ = −5.73 ± 9.50 Hz, P = .016), 60–170 Hz (Figure 3A, Δ = −5.97 ± 11.77 Hz, P = .039), and 60–200 Hz (Figure 3B, Δ = −9.18 ± 14.14 Hz, P = .011).

Decreases in spectral centroid.

There were significant decreases in the spectral centroid of snoring sounds from first N2 to last N2 in (A) 60–170 Hz frequency range; and (B) 60–200 Hz frequency range (assessed by paired t test).

jcsm.14.10.1653c.jpg

jcsm.14.10.1653c.jpg
Figure 3

Decreases in spectral centroid.

(more ...)

We also found a significant and negative correlation between the increases in NFV during sleep and the spectral centroid of snoring in the Last N2 between 60–140 Hz (r = −.53, P = .020), 60–170 Hz (Figure 4A, r = −.55, P = .012) and 60–200 Hz (Figure 4B, r = −.58, P = .010). Furthermore, we found that increases in NC during sleep were significantly related to the decreases in spectral centroid of snoring in the last N2 between 60–140 Hz (r = −.51, P = .014), 60–170 Hz (Figure 4C, r = −.58, P = .008) and 60–200 Hz (Figure 4D, r = −.57, P = .009).

Correlation between spectral centroid and neck fluid volume as well as neck circumference.

Increases in neck fluid volume (NFV) were inversely correlated with the spectral centroid of snoring sounds in last N2 in the (A) 60–170 Hz frequency range; and (B) 60–200 Hz frequency range. Increases in neck circumference (NC) were inversely correlated with the spectral centroid of snoring sounds in the last N2 in the (C) 60–170 Hz frequency range; and (D) 60–200 Hz frequency range.

jcsm.14.10.1653d.jpg

jcsm.14.10.1653d.jpg
Figure 4

Correlation between spectral centroid and neck fluid volume as well as neck circumference.

(more ...)

Furthermore, there was a borderline significant inverse correlation between decreases in the spectral-centroid (60–170 Hz) from the beginning to the end of sleep and the percentage increases in the NC from before to after sleep (Figure 5, r = −.44, P = .053). Further analysis of this revealed that with the increases in NC, 8 out of 20 participants did not have a change or increase in the centroid frequencies. The average AHI of these 8 participants was significantly higher than the rest of participants (42.9 ± 30.3 versus 15.7 ± 10.6 events/h, P < .05). Furthermore, before sleep NC was significantly larger in these 8 participants than others (43.5 ± 1.9 versus 40.5 ± 2.5 cm, P < .05). Moreover, before sleep NFV was smaller in these 8 participants than others (246.7 ± 30.6 versus 280.1 ± 51.3 mL, P < .05).

Correlation between changes in the spectral centroid and neck circumference.

A trend of an inverse relationship was observed between Δspectral-centroid (difference between spectral centroid in last and first part of the stage 2 of sleep) and ΔNC in 60–170 Hz frequency band.

jcsm.14.10.1653e.jpg

jcsm.14.10.1653e.jpg
Figure 5

Correlation between changes in the spectral centroid and neck circumference.

(more ...)

Additionally, there was a trend for an inverse relationship between changes in the spectral-centroid in first N2 between 30–4000 Hz and AHI (r = −.42, P = .064).

There was no significant correlation between ΔNC, ΔNFV and spectral centroid at frequency ranges higher than 200 Hz.

DISCUSSION

Our study has given rise to several novel observations on the relationships between snoring sounds features and increases in the pharyngeal tissue content during sleep. The main findings of this study were (1) from beginning to the end of sleep, spectral centroid of snoring sounds decreased significantly in the frequency ranges below 200 Hz; and (2) increases in NFV and NC during sleep were associated with decreases in the spectral centroid frequency of snoring sounds in last 30 minutes of stage N2 sleep in the frequency ranges below 200 Hz.

Although we found significant relationships between changes in NFV and snoring features in the lower frequency ranges, no such relationships were found in the higher frequency ranges. Previous studies have shown that the low-frequency components (< 300 Hz) of breath sounds tend to couple with the tissue-filled compartment and propagate through the tissue.39,46 However, the high-frequency components of snoring sounds tend to propagate through the airway path rather than tissue because the airway walls behave as rigid structures at higher frequencies.46 Therefore, it may be justified that lower frequency components of snoring sounds and not the higher frequency components will be affected by the fluid accumulation in the neck.

Fluid accumulation in the neck is a continuous process and NFV increases continuously while supine.32 Therefore, it is expected that centroid frequencies of snoring would change from the beginning to the end of sleep. Our results showed that from the first N2 to last N2, spectral centroid of snoring sounds decreased significantly. Based on this result, it can be assumed that increases in the NFV were associated with the decreases in the spectral centroid of snoring sounds. Similarly, we found significant correlations between increases in NFV or NC during sleep and reduction of the spectral centroid of the snoring sounds in the last 30 minutes of sleep. Our findings suggest that increases in NFV could increase the mass of the pharyngeal tissue surrounding the upper airway and consequently decrease the centroid frequency. We acknowledge the fact that this analysis is based on the assumption that increases in NFV do not change the pharyngeal tissue elasticity. Future studies are required to measure the stiffness of the pharyngeal tissue and its association with NFV.

Furthermore, our results comply with previous studies that modeled the generation of wheezing in collapsible tubes.47,48 Although wheeze and snoring sounds are different, both are influenced by the collapsible tubes. Grotberg et al. have shown that increases in the thickness of surrounding walls of a collapsible tube will reduce the frequency of vibratory sounds.48 In this study, increases in the NFV can increase the thickness of the surrounding tissue of the upper airway, as assessed by the increases in the NC. Accordingly, results of our study and previous modeling studies would suggest that increases in the pharyngeal mass and tissue thickness because of the increased NFV may decrease the centroid frequency of the snoring sounds. It should be noted that soft-palate vibration is a dominant site of vibration in snoring sound generation, which is missing in the wheezing generation models. Upper airway narrowing because of increased tissue content during sleep increases the airflow speed and based on Bernoulli's theorem, increased airflow speed will cause a pressure drop across the upper airway. Pressure drop across the upper airway will further narrow the upper airway and increase airflow speed. This consequence of events will lead to vibration of soft palate or pharyngeal wall and induce snoring sounds. Although there is no established relationship between soft-palate vibration and increased pharyngeal tissue content, upper airway narrowing because of increased tissue content during sleep may influence the soft-palate vibration. This should be validated in future studies.

Although spectral centroid in the last 30 minutes of stage N2 sleep was correlated with the changes in NC and NFV, spectral centroid in the first 30 minutes of stage N2 sleep was not. Our previous study has shown that fluid starts to shift rostrally as soon as the participants lay supine, and approximately 80% of fluid shifts within the first 2 hours of lying supine.32 In our participants, snoring sounds in first N2 occurred after approximately 36 ± 32 minutes of lying supine (10 of 20 patients have first snoring after 30 minutes of lying supine). Therefore, it is expected that at the time of snoring within first N2, approximately 30% of fluid may have already been accumulated in the neck. Thus, the snoring sounds features in first N2 may not be correlated to the baseline or changes in NC or NFV.

Severity of OSA is characterized by number of episodes of upper airway collapse per hour of sleep. However, upper airway collapse is a multifactorial phenomenon and increased pharyngeal tissue content surrounding the upper airway due to increased NFV during sleep is one of the risk factors for OSA.12 Because upper airway and pharyngeal tissue content may be represented by high- and low-frequency features of snoring sounds, respectively, AHI may hold a correlation with both the high- and low-frequency components.46 In compliance with this, our results showed a trend for a negative correlation between the AHI and spectral centroid of snoring sounds between the frequency ranges of 30 to 4000 Hz. This negative association between AHI and frequency features of snoring sounds complies with previous studies showing that formant frequencies of snoring sounds are smaller in patients with more severe OSA.23,40 Although spectral centroid is not the same as the formant frequencies, other studies have shown that spectral centroid in wider frequency ranges (300 to 2000 Hz) may have similar properties as those of formant frequencies.49 Moreover, we found a borderline significant correlation between decreases in spectral centroid and increases in NC. Further inspection of this correlation revealed that reduction in spectral centroid of snoring sounds because of increased NFV during sleep was more prominent in participants with the low AHI and small baseline NC, which may indicate less pharyngeal fat and tissue. These results may be used to implement spectral centroid of snoring sounds for assessing the increased pharyngeal tissue content due to increased fluid shift during sleep. Because this method is more convenient and dynamic than other anatomical measurement such as measuring fat or NC, our proposed method based on spectral centroid of snoring sounds can be used to access the risk of OSA due to increased pharyngeal tissue content.

This study has some limitations. Because the study was performed during daytime sleep, the results may not be fully applicable to overnight sleep. Future studies should validate these results in overnight sleep studies. Furthermore, participants slept only in the supine position to minimize the effects of posture on OSA severity. Future studies should address the effects of posture on snoring sounds features. In this study, we only considered men, as previous studies from our group have shown that men are more susceptible to the adverse effects of rostral fluid shift than women.50 Moreover, this study was a retrospective study and our current dataset is limited in terms of the sample size. Future studies should investigate the relationship between snoring sounds features and pharyngeal tissue content in women, children, and patients with fluid-retaining condition such as heart failure and renal failure in the larger population.

In conclusion, this study has demonstrated the concept of developing a noninvasive and convenient measure by spectral centroid of snoring sounds to estimate the increases in pharyngeal tissue mass, as assessed by the increases in NFV and NC, during sleep. Once established, snore-driven monitoring of pharyngeal tissue content could be used to investigate its effect on the upper airway collapsibility and OSA severity.

DISCLOSURE STATEMENT

Work for this study was performed at the Toronto Rehabilitation Institute, Canada. All authors have reviewed and approved the manuscript. The research leading to these results has received partial funding from operating grants by the Canadian Institute of Health Research (CIHR) MOP-82731, and Ontario Centers of Excellence-VIPI. S.S was supported by the Connaught International Scholarships for Doctoral Students, University of Toronto and University of Manitoba Graduate Fellowship (UMGF), University of Manitoba. T.D.B. was supported by the Clifford Nordal Chair in Sleep Apnea and Rehabilitation Research and the Godfrey S. Pettit Chair in Respiratory Medicine. The authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

BMI

body mass index

First N2

first 30 minutes of stage N2 sleep

Last N2

last 30 minutes of stage N2 sleep

NC

neck circumference

NFV

neck fluid volume

OSA

obstructive sleep apnea

UA-XSA

upper airway cross sectional area

REFERENCES

1 

Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol. 2013;177(9):1006–1014. [PubMed Central][PubMed]

2 

Brooks D, Horner RL, Kozar LF, Render-Teixeira CL, Phillipson EA. Obstructive sleep apnea as a cause of systemic hypertension. Evidence from a canine model. J Clin Invest. 1997;99(1):106[PubMed Central][PubMed]

3 

Djonlagic I, Malhotra A. Risk of stroke from sleep apnea in men and women. Expert Rev Neurother. 2010;10(8):1267–1271. [PubMed]

4 

Floras JS. Sleep apnea and cardiovascular risk. J Cardiol. 2014;63(1):3–8. [PubMed]

5 

Kasai T, Floras JS, Bradley TD. Sleep apnea and cardiovascular disease: a bidirectional relationship. Circulation. 2012;126(12):1495–1510. [PubMed]

6 

Sanna A. Obstructive sleep apnoea, motor vehicle accidents, and work performance. Chron Respir Dis. 2013;10(1):29–33. [PubMed]

7 

Teran-Santos J, Jimenez-Gomez A, Cordero-Guevara J. The association between sleep apnea and the risk of traffic accidents. N Engl J Med. 1999;340(11):847–851. [PubMed]

8 

Akerstedt T, Philip P, Capelli A, Kecklund G. Sleep loss and accidents--work hours, life style, and sleep pathology. Prog Brain Res. 2011;190:169–188. [PubMed]

9 

Dempsey JA, Veasey SC, Morgan BJ, O'Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47–112. [PubMed Central][PubMed]

10 

Verbraecken JA, De Backer WA. Upper airway mechanics. Respiration. 2009;78(2):121–133. [PubMed Central][PubMed]

11 

Pahkala R, Seppa J, Ikonen A, Smirnov G, Tuomilehto H. The impact of pharyngeal fat tissue on the pathogenesis of obstructive sleep apnea. Sleep Breath. 2014;18(2):275–282. [PubMed]

12 

Katz I, Stradling J, Slutsky AS, Zamel N, Hoffstein V. Do patients with obstructive sleep apnea have thick necks? Am Rev Respir Dis. 1990;141(5 Pt 1):1228–1231. [PubMed]

13 

White LH, Bradley TD. Role of nocturnal rostral fluid shift in the pathogenesis of obstructive and central sleep apnoea. J Physiol. 2013;591(5):1179–1193. [PubMed]

14 

White LH, Motwani S, Kasai T, Yumino D, Amirthalingam V, Bradley TD. Effect of rostral fluid shift on pharyngeal resistance in men with and without obstructive sleep apnea. Respir Physiol Neurobiol. 2014;192:17–22. [PubMed]

15 

Shiota S, Ryan CM, Chiu KL, et al. Alterations in upper airway cross-sectional area in response to lower body positive pressure in healthy subjects. Thorax. 2007;62(10):868–872. [PubMed Central][PubMed]

16 

Kasai T, Motwani SS, Yumino D, et al. Contrasting effects of lower body positive pressure on upper airways resistance and partial pressure of carbon dioxide in men with heart failure and obstructive or central sleep apnea. J Am Coll Cardiol. 2013;61(11):1157–1166. [PubMed]

17 

Su MC, Chiu KL, Ruttanaumpawan P, et al. Lower body positive pressure increases upper airway collapsibility in healthy subjects. Respir Physiol Neurobiol. 2008;161(3):306–312. [PubMed]

18 

Yadollahi A, Gabriel JM, White LH, Taranto Montemurro L, Kasai T, Bradley TD. A randomized, double cross-over study to investigate the influence of saline infusion on sleep apnea severity in men. Sleep. 2014;37(10):1699–1705. [PubMed Central][PubMed]

19 

White LH, Lyons OD, Yadollahi A, Ryan CM, Bradley TD. Effect of below-the-knee compression stockings on severity of obstructive sleep apnea. Sleep Med. 2015;16(2):258–264. [PubMed]

20 

Redolfi S, Arnulf I, Pottier M, et al. Attenuation of obstructive sleep apnea by compression stockings in subjects with venous insufficiency. Am J Respir Crit Care Med. 2011;184(9):1062–1066. [PubMed]

21 

Lyons OD, Chan CT, Yadollahi A, Bradley TD. Effect of ultrafiltration on sleep apnea and sleep structure in patients with end stage renal disease. Am J Respir Crit Care Med. 2015;191(11):1287–1294. [PubMed]

22 

Ng AK, Koh TS, Baey E, Puvanendran K. Role of upper airway dimensions in snore production: acoustical and perceptual findings. Ann Biomed Eng. 2009;37(9):1807–1817. [PubMed]

23 

Saha S, Bradley TD, Taheri M, Moussavi Z, Yadollahi A. A subject-specific acoustic model of the upper airway for snoring sounds generation. Sci Rep. 2016;6:25730[PubMed Central][PubMed]

24 

Huang L, Quinn SJ, Ellis PD, Williams JE. Biomechanics of snoring. Endeavour. 1995;19(3):96–100. [PubMed]

25 

Rembold CM, Suratt PM. Airway turbulence and changes in upper airway hydraulic diameter can be estimated from the intensity of high frequency inspiratory sounds in sleeping adults. J Physiol. 2014;592(17):3831–3839. [PubMed]

26 

Xi J, Si X, Kim J, Su G, Dong H. Modeling the pharyngeal anatomical effects on breathing resistance and aerodynamically generated sound. Med Biol Eng Comput. 2014;52(7):567–577. [PubMed]

27 

Xu H, Huang W, Yu L, Chen L. Sound spectral analysis of snoring sound and site of obstruction in obstructive sleep apnea syndrome. Acta Otolaryngol. 2010;130(10):1175–1179. [PubMed]

28 

Lee LA, Lo YL, Yu JF, et al. Snoring sounds predict obstruction sites and surgical response in patients with obstructive sleep apnea hypopnea syndrome. Sci Rep. 2016;6:30629[PubMed Central][PubMed]

29 

Koo SK, Choi JW, Myung NS, Lee HJ, Kim YJ, Kim YJ. Analysis of obstruction site in obstructive sleep apnea syndrome patients by drug induced sleep endoscopy. Am J Otolaryngol. 2013;34(6):626–630. [PubMed]

30 

Berry RB, Brooks R, Gamaldo CE, et al; for the American Academy of Sleep Medicine. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Version 2.0. Darien, IL: American Academy of Sleep Medicine; 2012.

31 

Fredberg JJ, Wohl ME, Glass GM, Dorkin HL. Airway area by acoustic reflections measured at the mouth. J Appl Physiol Respir Environ Exerc Physiol. 1980;48(5):749–758. [PubMed]

32 

Yadollahi A, Singh B, Bradley TD. Investigating the dynamics of supine fluid redistribution within multiple body segments between men and women. Ann Biomed Eng. 2015;43(9):2131–2142. [PubMed]

33 

Boersma P, Weenink D. Praat: a system for doing phonetics by computer. Glot Int. 2001;5:341–345

34 

Saha S, Taheri M, Mossuavi Z, Yadollahi A. Effects of changing in the neck circumference during sleep on snoring sound characteristics. Conf Proc IEEE Eng Med Biol Soc. 2015;2015:2235–2238. [PubMed]

35 

Yadollahi A, Moussavi ZM. A robust method for heart sounds localization using lung sounds entropy. IEEE Trans Biomed Eng. 2006;53(3):497–502. [PubMed]

36 

FitzGerald D, Paulus J. Unpitched Percussion Transcription. In: Klapuri A, Davy M. Signal Processing Methods for Music Transcription. Boston, MA: Springer; 2006:131–162.

37 

Yadollahi A, Moussavi ZM. A robust method for estimating respiratory flow using tracheal sounds entropy. IEEE Trans Biomed Eng. 2006;53(4):662–668. [PubMed]

38 

Beck R, Odeh M, Oliven A, Gavriely N. The acoustic properties of snores. Eur Respir J. 1995;8(12):2120–2128. [PubMed]

39 

Pasterkamp H, Kraman SS, Wodicka GR. Respiratory sounds. Advances beyond the stethoscope. Am J Respir Crit Care Med. 1997;156(3 Pt 1):974–987. [PubMed]

40 

Fiz JA, Abad J, Jané R, et al. Acoustic analysis of snoring sound in patients with simple snoring and obstructive sleep apnoea. Eur Respir J. 1996;9(11):2365–2370. [PubMed]

41 

Ng AK, Koh TS, Baey E, Lee TH, Abeyratne UR, Puvanendran K. Could formant frequencies of snore signals be an alternative means for the diagnosis of obstructive sleep apnea? Sleep Med. 2008;9(8):894–898. [PubMed]

42 

Halliday D, Resnick R, Walker J. Fundamentals of Physics Extended. 9th ed. New York, NY: Wiley; 2010.

43 

Jones FE, Harris GL. ITS-90 density of water formulation for volumetric standards calibration. J Res Nat Inst Stand Technol. 1992;97(3):335–340

44 

Horner RL, Peever JH. Brain circuitry controlling sleep and wakefulness. Continuum (Minneap Minn). 2017;23(4, Sleep Neurology):955–972.

45 

Haut TL, Haut RC. The state of tissue hydration determines the strain-rate-sensitive stiffness of human patellar tendon. J Biomech. 1997;30(1):79–81. [PubMed]

46 

Kaniusas E. Biomedical Signals and Sensors I: Linking Physiological Phenomena and Biosignals. Berlin, Germany: Springer-Verlag Berlin Heidelberg; 2012.

47 

Gavriely N, Shee TR, Cugell DW, Grotberg JB. Flutter in flow-limited collapsible tubes: a mechanism for generation of wheezes. J Appl Physiol (1985). 1989;66(5):2251–2261

48 

Grotberg JB, Gavriely N. Flutter in collapsible tubes: a theoretical model of wheezes. J Appl Physiol (1985). 1989;66(5):2262–2273

49 

Paliwal KK. Spectral subband centroid features for speech recognition. Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181); New York, NY: IEEE; 1998.

50 

Kasai T, Motwani SS, Yumino D, Mak S, Newton GE, Bradley TD. Differing relationship of nocturnal fluid shifts to sleep apnea in men and women with heart failure. Circ Heart Fail. 2012;5(4):467–474. [PubMed]