Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Nov 1.
Published in final edited form as: J Biomech. 2022 Sep 15;144:111304. doi: 10.1016/j.jbiomech.2022.111304

Computational Assessment of Upper Airway Muscular Activity in Obstructive Sleep Apnea – In Vitro Validation

Daniel Ignatiuk 1,2,3, Qiwei Xiao 1,3, Keith McConnell 1, Robert Fleck 3,4, Christine Schuler 2,5, Andreas Schuh 6, Raouf Amin 1,2, Alister Bates 1,2,3,4,7
PMCID: PMC9664483  NIHMSID: NIHMS1844734  PMID: 36170766

Abstract

Neuromuscular control of the upper airway contributes to obstructive sleep apnea (OSA). An accurate, non-invasive method to assess neuromuscular function is needed to improve surgical treatment outcomes. Currently, surgical approaches for OSA are based on airway anatomy and are often not curative. When the airway surface moves, the power transferred between air in the airway lumen and the structures of the upper airway may be a measure of airway neuromuscular activity. The aim of this study was to validate power transfer as a measure of externally applied forces, representing neuromuscular activity, through cine computed tomography (CT) imaging and computational fluid dynamics (CFD) analysis in a 3D-printed airway model. A hollow elastic airway model was manufactured. An insufflation/exsufflation device generated airflow within the model lumen. The model was contained in an airtight chamber that could be positively or negatively pressurized to represent muscular forces. These forces were systematically applied to dilate and collapse the model. Cine CT imaging captured airway wall movement during respiratory cycles with and without externally applied forces. Power transfer was calculated from the product of wall movement and internal aerodynamic pressure forces using CFD simulations. Cross-correlation peaks between power transfer and changes in externally applied pressure during exhalation and inhalation were −0.79 and 0.95, respectively. Power transfer calculated via cine CT imaging and CFD was an accurate surrogate measure of externally applied forces representing airway muscular activity. In the future, power transfer may be used in clinical practice to phenotype patients with OSA and select personalized therapies.

Keywords: Airway, obstructive sleep apnea, muscle activity, computational fluid dynamics, power transfer

Introduction

Obstructive sleep apnea (OSA) is a medical disorder characterized by recurrent upper airway obstruction during sleep resulting in impaired respiratory gas exchange. Upper airway patency is a function of airway anatomy (Dempsey et al., 2002; Schwab et al., 2003; Sforza et al., 2000), body position with respect to gravity (Bafkar et al., 2020; Jokic et al., 1999; Ong et al., 2011), air pressure on the airway wall due to respiratory airflow (Remmers et al., 1978), and neuromuscular activation of structures surrounding the upper airway (Tangel et al., 1991; Worsnop et al., 1998). Neuromuscular activation is related to the complex relationship between muscle tone and ventilatory drive and contributes to respiratory disturbance and arousal during sleep (Dempsey et al., 2010; Warner et al., 1987; Wheatley et al., 1993).

If untreated, OSA is associated with serious health consequences including cardiovascular disease and neuropsychological impairment (Bradley and Floras, 2009; Wallace and Bucks, 2013; Wheaton et al., 2012). Therefore, many factors that undermine airway patency are targets of OSA treatment. Treatment options include hypoglossal nerve stimulation of upper airway musculature (Strollo et al., 2014), surgical changes to address abnormal airway anatomy (Epstein et al., 2009; Marcus et al., 2012), and application of continuous positive airway pressure (CPAP) to pneumatically stent the upper airway open (Jordan et al., 2014; Marcus et al., 1995; Patil et al., 2019). Upper airway surgery is often a promising option, offering the possibility of OSA resolution postoperatively. However, wide variation in surgical success rates is problematic (Bhattacharjee et al., 2010; Camacho et al., 2016a; Camacho et al., 2016b; Kezirian and Goldberg, 2006; Manickam et al., 2016; Marcus et al., 2013; Stuck et al., 2018; Ulualp, 2014).

Variable surgical treatment outcomes for OSA are likely due, at least in part, to the inherent difficulty in quantifying all factors that contribute to upper airway obstruction. Upper airway anatomical structure may be characterized through endoscopy and/or radiologic imaging (Cavaliere et al., 2013; Chen et al., 2016; Truong et al., 2012). However, quantifying airway muscle tone and neuromuscular activity is more complex. Invasive testing, such as electromyography, would be informative in many cases but is not often performed outside of research settings (Tangel et al., 1991; Worsnop et al., 1998). Consequently, surgical intervention for OSA often occurs without fully measuring the role of airway muscle tone and neuromuscular activity.

Quantification of neuromuscular activity in the upper airway with non-invasive means is theoretically possible. Specifically, cine imaging of the airway has been combined with airflow modeling via computational fluid dynamics (CFD) simulation for this purpose (Bates et al., 2019; Bates et al., 2018). Cine imaging captures airway surface motion, while CFD calculates air pressure forces acting on the airway. This combined approach is valuable because it may estimate power transfer between air in the airway lumen and structures of the upper airway. However, this method has not been validated with direct measurements of neuromuscular activity.

Quantifying power transfer between the airway wall and the air inside the lumen is vital to understanding why airways dilate or collapse in relation to air pressure and how neuromuscular forces contribute to airway behavior. Airway movement in the same direction as applied air pressure is considered synchronous and may indicate airway muscle tone is inadequate to maintain airway patency against applied air pressure; power is transferred from the air to the airway wall. For example, the upper airway may collapse in response to negative pressure in the airway during inhalation. Conversely, the airway may move in the opposite direction to the air pressure; an asynchronous response. In this case, power is transferred from the airway to the air, and this asynchronous motion may be related to neuromuscular activity. In synchronous or asynchronous power transfer, airway motion may dilate or collapse the airway, yielding four types of possible power transfer: synchronous dilation, synchronous collapse, asynchronous dilation, and asynchronous collapse.

This study aimed to validate power transfer as a measure of neuromuscular activity through cine computed tomography (CT) imaging and CFD analysis in a 3D-printed airway model.

Methods

Experimental Design

A pliable upper airway model, and a closed system that allowed for model manipulations and measurements, were manufactured for this study. This model was designed such that the forces dilating and collapsing the airway could be altered, and resultant airway motion could be measured by cine CT imaging. Both intraluminal respiratory airflow and externally applied pressure forces, simulating muscular activity, could be applied to the model. Three respiratory cycles were modeled in which both synchronous and asynchronous power transfer were expected. Power transfer was calculated via CFD simulation. The relationship between power transfer and applied external forces was then tested through cross-correlation to determine if power transfer may be used as a measure of neuromuscular control.

Airway Model, Chamber, and Materials

Using an existing CT scan of a pediatric subject without OSA or airway abnormalities, airway lumen geometry was segmented to create a virtual airway surface to which a thickness of 0.95 to 1.00 mm was added. The airway model was 3D-printed from elastic resin (Elastic 50A resin. Formlabs, Somerville, MA, USA) and was coated with silicone (Dragon Skin 30 silicone, Smooth-On, Inc., Macungie, PA, USA) to maintain integrity during deformation. The final surface thickness was 2 to 4 mm after silicone application. The model was hollow with isolated openings at the nostrils and distal trachea (Figure 1). The entire model was free to deform except for the nasal and tracheal ends, which were attached to tubing that allowed airflow through the model. The nasal tubing was connected to an opening at atmospheric pressure. The tracheal tubing was attached to a mechanical insufflation/exsufflation device, which acted as the lungs, providing bidirectional flow through the model to represent the respiratory cycle.

Figure 1.

Figure 1.

Open in a new tab

3D-printed flexible airway model with openings at nostrils and trachea.

The model was contained in an airtight chamber which could be positively or negatively pressurized (with respect to atmospheric pressure) to create external collapsing or dilating forces, respectively. These external pressures were achieved by connecting the chamber to central medical air and vacuum lines. The external forces were designed to move the airway model similarly to how the in vivo airway moves in response to neuromuscular activity. However, in this simplified model, the external forces acted globally across the airway surface, rather than having forces affecting specific regions of the airway surface. The magnitude of external dilating forces was smaller than the magnitude of external collapsing forces because the magnitude of the external forces was determined as the force necessary to provide visible movement of the airway wall. Figure 2 shows how internal airflow and external pressure (representing artificial muscle force) were applied to the model.

Figure 2.

Figure 2.

Open in a new tab

(a) Picture of airway model mounted within pressure chamber with connection points for airflow and pressurization at base, (b) schematic demonstrating connections for internal airflow through model and pressurization of chamber to simulate artificial muscle activity.

The resulting airway model was idealized and did not depict a single patient or a subject with OSA. The model had uniform elasticity and was not intended to simulate the underlying tissues or specific muscles. Instead, the model was designed to display surface geometry and airway wall motion similar to an in vivo airway moving in response to internal airflow and neuromuscular forces.

Airway Model Imaging

Cine CT imaging of the model was performed using an Aquilion ONE (Canon Medical Systems, Tochigi, Japan) 320-detector CT scanner. Dynamic CT images at 200 ms intervals were acquired using volumetric mode with low-dose technique (80 kVp, 20–30 mA, gantry rotation time of 0.35 s, voxel resolution of 0.39 × 0.39 × 0.25 mm, and scan field of view of 20 × 20 × 160 cm). Three breathing scenarios were imaged, including one breath with only internal airflow, one breath with internal airflow and an external dilating force applied during inhalation, and one breath with internal airflow and an external collapsing force applied during exhalation. Each respiratory cycle consisted of two seconds of exhalation followed by two seconds of inhalation, driven by 10 cm H2O of positive or negative pressure from the insufflation/exsufflation device. This pressure could visibly collapse and dilate the model (in the absence of other external forces), without risking model integrity with higher pressures. Figure 3 shows example CT images of the airway model under different flow and external force conditions.

Figure 3.

Figure 3.

Open in a new tab

CT image of airway model (a) without internal airflow, (b) during peak exhalation pressures, (c) during peak inhalation pressures, (d) with external dilating force, and (e) with external collapsing force. Airway motion is apparent at the superior portion of the tongue.

Throughout imaging, volumetric airflow through the model and pressure within the chamber but external to the model were continuously measured using a spirometer (ADInstruments FE141, Colorado Springs, CO, USA) connected to a respiratory flow head (MLT300L) and data recorder (PowerLab 8/35). Data were synchronized with the CT scanner via a trigger signal, such that the precise flows and pressures occurring during each image were recorded. The volumetric flow rate at the distal trachea was recorded for use as the inlet boundary condition for computational airflow simulations.

CFD Simulation

A CT image of the airway was first generated without chamber pressurization or airflow through the model. This was done to capture the initial condition for each breath, which was a non-pressurized state, and created a virtual airway surface that matched the 3D-printed model. This image was segmented using a contouring technique (ITK-Snap 3.8.0, Penn Image Computing and Science Laboratory, USA, www.itksnap.org) (Yushkevich et al., 2006). A surface mesh was produced and smoothed (Meshlab 2021.07, Visual Computing Laboratory, ISTI-CNR, Pisa, Italy) (Cignoni et al., 2008).

Next, motion maps were generated using non-rigid cine CT image sequences. Motion maps were applied to the virtual model surface to create a moving airway surface that followed the motion of the printed model through each breath. A motion-tracking algorithm was applied (MIRTK 1.1 (Department of Computing, Imperial College London, UK, https://mirtk.github.io/). These moving surfaces provided the geometry boundary conditions for the CFD simulations. The approach used to produce a map of model motion through each breath was described previously by Bates et al. (Bates et al., 2019; Bates et al., 2018).

CFD simulations and mesh generation were performed using Simcenter STAR-CCM+ 2019.2 (Siemens PLM Software, Plano, TX, USA). The initial airway surface was meshed using 9 polyhedral layers and polyhedral elements in the bulk of the mesh. The mean volume of cell elements was 0.03 mm3 and a total of 2.4 million mesh elements were used. These parameters were determined by previous mesh convergence studies performed in pediatric airways within the same anatomical regions which showed these mesh settings produce comparable results to those of a quasi-direct numerical simulation (Bates et al., 2019; Bates et al., 2018).

CFD simulations were unsteady and simulated the duration of each breath. The virtual airway surface moved through each breath following the prescribed motion obtained from image registration. As the airway surface moved, the volume mesh also moved using Simcenter Star-CCM+’s morphing function, and the Navier-Stokes equations were modified to incorporate mesh motion. These techniques have previously been described in detail by Bates et al. (Bates et al., 2019; Bates et al., 2018).

A temporally varying mass-flow boundary condition was provided by the flow meter attached to the physical model and applied to the distal tracheal boundary. The nasal end of the model was held at atmospheric pressure because the nose in the physical model was connected to room air. The large eddy simulation (LES) turbulence model was used with a wall-adapting local eddy-viscosity (WALE) subgrid-scale model. Temporal discretization was set at 1 ms and was determined by the convergence study described above.

Power Transfer Calculation

Power transfer from the airway wall to the air, or vice versa, was calculated from the scalar product of the pressure force acting on the airway wall (the vector sum of air pressure and wall shear stress) and the velocity of airway wall movement. This product was calculated at each face of the airway surface and integrated across the airway surface at each time-step. When power transfer had a positive value, the airway wall was moving in the same direction as the applied pressure force and was categorized as “synchronous.” When there was a >90-degree discrepancy between the airway wall velocity and pressure force vectors, power transfer had a negative value and was categorized as “asynchronous”. For ease of reporting in this study, all power transfer values are described as absolute values with positive values considered synchronous and negative values considered asynchronous. Motion and power transfer were further categorized by whether motion dilated or collapsed the airway, resulting in four possible categories: synchronous dilation, synchronous collapse, asynchronous dilation, and asynchronous collapse.

Statistical Analysis

Cross-correlation analyses assessed the relationship between calculated power transfer and the rate of change of external pressure forces acting on the airway throughout the entire 4-second simulated respiratory cycle. This resulted in two cross-correlation analyses for the two imaging scenarios during which either external collapsing or dilating pressure forces were applied. The rate of change of applied external pressure was used to better represent in vivo muscle activation as a change from baseline muscle tone. This external pressure rate of change was defined by the following equation:

ΔP=Pt+Δt−PtΔt

For this analysis, Δt was defined as a 1 ms time step. Therefore, cross-correlation analyses compared data series values at 1 ms intervals throughout each 4-second simulated respiratory cycle. Cross-correlation analysis allowed for quantitative measure of the relationship between two time-dependent series and generated correlation coefficients which varied based on the time shift between these series (Nelson-Wong et al., 2009; Smith, 2003). This analysis method has been applied to human movement research, functional MRI, signal processing, and OSA-related research (Berry et al., 1998; Hyde and Jesmanowicz, 2012; Nelson-Wong et al., 2009). Coefficient ranges were interpreted as 0–0.39 representing weak correlation, 0.4–0.69 representing moderate correlation, and 0.7 to 1.0 representing strong correlation, consistent with generally recommended correlation coefficient interpretation (Mukaka, 2012; Schober et al., 2018).

Results

Data was collected from three separate CT imaging acquisitions occurring over the course of a 4-second simulated respiratory cycle. The first acquisition occurred without applied external force, the second included an external dilating force starting during inhalation, and the third included an external collapsing force starting during exhalation. Figure 4a shows the flow rate within the model throughout the respiratory cycle. This profile remained similar throughout all imaging acquisitions. Figure 4b demonstrates the associated external pressure profiles in the two cases with applied external pressure. The external force was timed to start in the phase of breathing during which external and internal forces would be in opposing directions. This external force continued throughout the remainder of the breath.

Figure 4:

Figure 4:

Open in a new tab

Comparison of (a) internal volumetric flow rate and (b) magnitude of externally applied pressure throughout the simulated respiratory cycle during relevant imaging acquisitions. The initiation of the internal flow cycle ranged up to ± 0.1 seconds between imaging acquisitions.

Figure 5a shows variation in total pressure difference between the distal trachea and nostrils throughout each breath calculated from the CFD model. The red curve shows the case without external force applied to the model. As expected, pressure at the distal trachea was higher than at the nose during exhalation, and lower during inhalation. Figure 5b shows the effect of internal pressure changes on the volume of the model. High pressure expanded the model during exhalation, whereas negative pressure caused model collapse during inhalation.

Figure 5:

Figure 5:

Open in a new tab

Plot of (a) airway model pressure difference between distal trachea and nostrils and (b) total airway model volume throughout all imaging acquisitions (1during asynchronous collapse imaging only, 2during asynchronous dilation imaging only).

The yellow and blue curves represent cases with external forces applied. Both curves deviate from the red curve as the model deformed differently in these cases. Specifically, the yellow curve (applied external collapsing force) shows significant collapse beginning around 1.6 s into the breath. The blue curve (applied external dilating force) shows dilation at around 3.8 s. These changes in airway volume affected the pressure difference along the airway. The collapse shown in the yellow curve (Figure 5b) is accompanied by an increase in pressure difference along the airway, as the narrowing model created increased resistance. During inhalation, this increased resistance manifested as a negative pressure difference of greater magnitude.

Figure 6 (d, e) shows the change in externally applied artificial muscle force with time. In the case with an external collapsing force, the positive pressure occurred shortly before 1.6 s, correlating with changes in airway volume and pressure difference described above. In the case of an external dilating force, a negative pressure was applied shortly before 3.6 s, corresponding to the increase in airway volume shown in Figure 5b at that time.

Figure 6:

Figure 6:

Open in a new tab

Plot of (a, b, c) power transfer data for all imaging acquisitions and (d, e) associated external pressure rate of change for acquisitions with externally applied force (1external dilating force, 2external collapsing force).

Power transfer for each of the three cases (without externally applied force, external dilating force, and external collapsing force) is also depicted in Figure 6 (a–c). In the absence of externally applied force (Figure 6a), peaks of synchronous collapse and dilation occurred at the beginning of exhalation and inhalation; these peaks indicate power transfer from the air to the model and the model passively responding to changes in internal pressure. When an external dilating force was applied, there was a peak in asynchronous power transfer coincident with the application of this force, shortly before 3.6 s. Similarly, when an external collapsing force was applied, there was also a peak in asynchronous power transfer. Asynchronous power transfer peaks also occurred outside of external force application and were usually present at end-exhalation or end-inhalation. However, these peaks were smaller in magnitude and ranged from 3.2 to 3.4 times smaller in comparison. Synchronous power transfer peaks were most prominent at the beginning of exhalation and inhalation.

To determine whether power transfer was a valid alternate measure of neuromuscular activation, power transfer was compared to the change in externally applied pressure throughout the two scenarios with externally applied dilating or collapsing forces. This included one cross-correlation of the data series in Figure 6b and 6d, for imaging with external dilating force, and a second cross-correlation of the data series in Figure 6c and 6e, for imaging with external collapsing force. Power transfer significantly correlated with the change in external, artificial muscular force throughout the scenario with external dilating force [normalized cross-correlation peak = 0.95] and the scenario with external collapsing force [normalized cross-correlation peak = −0.79]. In human subjects, whose upper airways are moved by muscles acting on specific sections of the airway, rather than globally, it may be useful to measure the power transfer on specific regions of the airway surface. To demonstrate that this technique allows calculation of power transfer over any section of interest, power transfer was compared to the change in externally applied pressure for a subsection of the model located along the anterior surface, inferior to the nasal cavity and superior to the trachea. The resulting cross-correlation peak was −0.79 in the imaging scenario with externally applied collapsing force, similar to the value for the entire model surface.

Discussion

This study is the first to examine power transfer as a surrogate for airway neuromuscular control using a 3D-airway model; neuromuscular forces are a key factor contributing to OSA that are often not quantified before treatment. Overall, the model performed as intended and replicated real physiology. Our results indicate that power transfer is a valid measure of neuromuscular activity in vitro, with significant associations between power transfer and externally applied artificial muscle forces. Asynchronous movement peaks matched peaks in externally applied forces, while synchronous peaks appeared at periods in the respiratory cycle when the direction or magnitude of airflow changed. Overall, our findings support that in this model, power transfer differentiates motion that occurs due to changes in air pressure from motion due to externally applied forces.

Our findings have potentially important clinical care implications. Although current diagnostic techniques such as drug-induced sleep endoscopy or cine imaging can identify airway motion, details about airflow and the forces driving airway motion cannot be derived from these methods. Our in vitro model successfully quantified the association between controlled externally applied forces and power transfer. If power transfer were validated as a surrogate for neuromuscular control in vivo, and correlated to clinical metrics such as polysomnography, it would provide a new, non-invasive measure of neuromuscular forces contributing to OSA. Characterizing the location and degree of upper airway muscular activity may play an important role in identifying the treatment(s) most likely to be successful among the options typically available: CPAP therapy (Jordan et al., 2014; Marcus et al., 1995; Patil et al., 2019), upper airway surgery (Camacho et al., 2016a; Epstein et al., 2009; Manickam et al., 2016; Marcus et al., 2013; Ulualp, 2014), and hypoglossal nerve stimulation to activate extrinsic tongue musculature and prevent upper airway obstruction (Strollo et al., 2014). Although hypoglossal nerve stimulation is still being evaluated in children, patients with low muscle tone in this nerve distribution may benefit more substantially from this therapy.

Detailed knowledge of neuromuscular activity may be particularly helpful in guiding operative treatments for OSA. In surgical planning, clinicians target anatomic regions that demonstrate upper airway obstruction. This approach often yields high rates of residual OSA following surgical intervention (Camacho et al., 2016a; Camacho et al., 2016b; Kezirian and Goldberg, 2006; Manickam et al., 2016; Marcus et al., 2013; Stuck et al., 2018; Ulualp, 2014). Knowledge of neuromuscular function may identify alternative sites for surgical intervention or pinpoint the optimal location for surgical modification in patients with multiple sites of obstruction.

This study incorporated cine CT imaging of an in vitro model with intervals of 200 ms. This resolution effectively measures obstructive events that occur during a typical adult breath. Although younger patients may have a shorter respiratory cycle duration, a 200 ms imaging frequency would be adequate even among infants with respiratory rates up to 60–70 breaths per minute (Fleming et al., 2011). MR imaging is often preferred in vivo to avoid exposure to ionizing radiation. Cine MRI could be used in lieu of CT imaging to calculate power transfer without exposing patients to unnecessary radiation.

There are several limitations to this study. First, artificial isotropic forces were applied over the entire exterior of the airway model in a direction normal to the model surface. This is different from anisotropic forces with direction of application dependent on the location of muscular attachment, which would be the case in a human airway. Our model could not replicate isolated motion of single muscles. This limitation is inherent to the printed model however, and not the power transfer metric. Additionally, power transfer does not differentiate neuromuscular expansion from passive expansion when neuromuscular forces and airway pressure move the airway in the same direction. In vivo neuromuscular control would typically be used to counteract air pressure and to dilate the airway, for example, during inspiration in response to pressure driven collapse. Thus, inability to differentiate neuromuscular from passive motion would not be a major concern in vivo.

Asynchronous power transfer that occurred without artificial muscle activation, and the magnitude of asynchronous contraction and dilation peaks, also merit further discussion. Relatively small magnitude asynchronous power transfers occurred near end-expiration or end-inspiration, not corresponding with external pressure forces. These misclassified asynchronous peaks were likely due to passive forces in the walls of the 3D-printed airway model, causing the model to return to its original size when internal air pressure forces were not of sufficient magnitude to counteract these passive wall forces. The passive restoring forces of the model airway wall were not included in the power transfer analysis and were therefore recorded as asynchronous power transfer peaks. The difference in magnitude of the cross-correlation peak during asynchronous collapse (−0.79) and dilation (0.95) is also important to highlight. This may be due to the sustained nature of external collapsing forces throughout inhalation and the end of exhalation.

Conclusion

Power transfer calculated via cine imaging and CFD was an accurate surrogate measure of externally applied forces in our 3D-printed airway model. In humans, a force externally applied to the airway would be due to neuromuscular activity except for observable shifts in body position causing airway wall movement due to gravity. Therefore, power transfer is validated as a quantitative measure of muscular activity resulting in airway wall motion in a direction opposing the expected direction of movement due to internal aerodynamic forces. This novel metric should be evaluated in vivo to determine if power transfer accurately captures directly measured neuromuscular activity throughout the airway in patients. Future correlation of power transfer to measures used in clinical practice may accelerate OSA phenotyping and selection of personalized therapies.

Acknowledgments

The authors thank the National Institutes of Health (NIH) and the American Thoracic Society ASPIRE Fellowship Program for help in funding this research (NIH grant K99 HL144822). We also thank the Cincinnati Children’s Hospital Medical Center CT imaging technicians, who performed the imaging for this study, and Matthew Nelson, medical artist, who helped create the elastic airway model.

References

  1. Bafkar O, Cajas JC, Calmet H, Houzeaux G, Rosengarten G, Lester D, Nguyen V, Gulizia S, Cole IS, 2020. Impact of sleeping position, gravitational force & effective tissue stiffness on obstructive sleep apnoea. J Biomech 104, 109715. [DOI] [PubMed] [Google Scholar]
  2. Bates AJ, Schuh A, Amine-Eddine G, McConnell K, Loew W, Fleck RJ, Woods JC, Dumoulin CL, Amin RS, 2019. Assessing the relationship between movement and airflow in the upper airway using computational fluid dynamics with motion determined from magnetic resonance imaging. Clin Biomech (Bristol, Avon) 66, 88–96. [DOI] [PubMed] [Google Scholar]
  3. Bates AJ, Schuh A, McConnell K, Williams BM, Lanier JM, Willmering MM, Woods JC, Fleck RJ, Dumoulin CL, Amin RS, 2018. A novel method to generate dynamic boundary conditions for airway CFD by mapping upper airway movement with non-rigid registration of dynamic and static MRI. Int J Numer Method Biomed Eng 34, e3144. [DOI] [PubMed] [Google Scholar]
  4. Berry RB, Asyali MA, McNellis MI, Khoo MC, 1998. Within-night variation in respiratory effort preceding apnea termination and EEG delta power in sleep apnea. J Appl Physiol (1985) 85, 1434–1441. [DOI] [PubMed] [Google Scholar]
  5. Bhattacharjee R, Kheirandish-Gozal L, Spruyt K, Mitchell RB, Promchiarak J, Simakajornboon N, Kaditis AG, Splaingard D, Splaingard M, Brooks LJ, Marcus CL, Sin S, Arens R, Verhulst SL, Gozal D, 2010. Adenotonsillectomy outcomes in treatment of obstructive sleep apnea in children: a multicenter retrospective study. Am J Respir Crit Care Med 182, 676–683. [DOI] [PubMed] [Google Scholar]
  6. Bradley TD, Floras JS, 2009. Obstructive sleep apnoea and its cardiovascular consequences. Lancet 373, 82–93. [DOI] [PubMed] [Google Scholar]
  7. Camacho M, Dunn B, Torre C, Sasaki J, Gonzales R, Liu SY, Chan DK, Certal V, Cable BB, 2016a. Supraglottoplasty for laryngomalacia with obstructive sleep apnea: A systematic review and meta-analysis. Laryngoscope 126, 1246–1255. [DOI] [PubMed] [Google Scholar]
  8. Camacho M, Li D, Kawai M, Zaghi S, Teixeira J, Senchak AJ, Brietzke SE, Frasier S, Certal V, 2016b. Tonsillectomy for adult obstructive sleep apnea: A systematic review and meta-analysis. Laryngoscope 126, 2176–2186. [DOI] [PubMed] [Google Scholar]
  9. Cavaliere M, Russo F, Iemma M, 2013. Awake versus drug-induced sleep endoscopy: evaluation of airway obstruction in obstructive sleep apnea/hypopnoea syndrome. Laryngoscope 123, 2315–2318. [DOI] [PubMed] [Google Scholar]
  10. Chen H, Aarab G, de Ruiter MH, de Lange J, Lobbezoo F, van der Stelt PF, 2016. Three-dimensional imaging of the upper airway anatomy in obstructive sleep apnea: a systematic review. Sleep Med 21, 19–27. [DOI] [PubMed] [Google Scholar]
  11. Cignoni P, Callieri M, Corsini M, Dellepiane M, Ganovelli F, Ranzuglia G, Scarano V, De Chiara R, Erra U, 2008. Sixth Eurographics Italian Chapter Conference. Salerno, Italy, 129–136. [Google Scholar]
  12. Dempsey JA, Skatrud JB, Jacques AJ, Ewanowski SJ, Woodson BT, Hanson PR, Goodman B, 2002. Anatomic determinants of sleep-disordered breathing across the spectrum of clinical and nonclinical male subjects. Chest 122, 840–851. [DOI] [PubMed] [Google Scholar]
  13. Dempsey JA, Veasey SC, Morgan BJ, O’Donnell CP, 2010. Pathophysiology of sleep apnea. Physiol Rev 90, 47–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Epstein LJ, Kristo D, Strollo PJ Jr., Friedman N, Malhotra A, Patil SP, Ramar K, Rogers R, Schwab RJ, Weaver EM, Weinstein MD, 2009. Clinical guideline for the evaluation, management and long-term care of obstructive sleep apnea in adults. J Clin Sleep Med 5, 263–276. [PMC free article] [PubMed] [Google Scholar]
  15. Fleming S, Thompson M, Stevens R, Heneghan C, Plüddemann A, Maconochie I, Tarassenko L, Mant D, 2011. Normal ranges of heart rate and respiratory rate in children from birth to 18 years of age: a systematic review of observational studies. Lancet 377, 1011–1018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hyde JS, Jesmanowicz A, 2012. Cross-correlation: an fMRI signal-processing strategy. Neuroimage 62, 848–851. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Jokic R, Klimaszewski A, Crossley M, Sridhar G, Fitzpatrick MF, 1999. Positional treatment vs continuous positive airway pressure in patients with positional obstructive sleep apnea syndrome. Chest 115, 771–781. [DOI] [PubMed] [Google Scholar]
  18. Jordan AS, McSharry DG, Malhotra A, 2014. Adult obstructive sleep apnoea. Lancet 383, 736–747. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kezirian EJ, Goldberg AN, 2006. Hypopharyngeal surgery in obstructive sleep apnea: an evidence-based medicine review. Arch Otolaryngol Head Neck Surg 132, 206–213. [DOI] [PubMed] [Google Scholar]
  20. Manickam PV, Shott SR, Boss EF, Cohen AP, Meinzen-Derr JK, Amin RS, Ishman SL, 2016. Systematic review of site of obstruction identification and non-CPAP treatment options for children with persistent pediatric obstructive sleep apnea. Laryngoscope 126, 491–500. [DOI] [PubMed] [Google Scholar]
  21. Marcus CL, Brooks LJ, Draper KA, Gozal D, Halbower AC, Jones J, Schechter MS, Sheldon SH, Spruyt K, Ward SD, Lehmann C, Shiffman RN, 2012. Diagnosis and management of childhood obstructive sleep apnea syndrome. Pediatrics 130, 576–584. [DOI] [PubMed] [Google Scholar]
  22. Marcus CL, Moore RH, Rosen CL, Giordani B, Garetz SL, Taylor HG, Mitchell RB, Amin R, Katz ES, Arens R, Paruthi S, Muzumdar H, Gozal D, Thomas NH, Ware J, Beebe D, Snyder K, Elden L, Sprecher RC, Willging P, Jones D, Bent JP, Hoban T, Chervin RD, Ellenberg SS, Redline S, 2013. A randomized trial of adenotonsillectomy for childhood sleep apnea. N Engl J Med 368, 2366–2376. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Marcus CL, Ward SL, Mallory GB, Rosen CL, Beckerman RC, Weese-Mayer DE, Brouillette RT, Trang HT, Brooks LJ, 1995. Use of nasal continuous positive airway pressure as treatment of childhood obstructive sleep apnea. J Pediatr 127, 88–94. [DOI] [PubMed] [Google Scholar]
  24. Mukaka MM, 2012. Statistics corner: A guide to appropriate use of correlation coefficient in medical research. Malawi Med J 24, 69–71. [PMC free article] [PubMed] [Google Scholar]
  25. Nelson-Wong E, Howarth S, Winter DA, Callaghan JP, 2009. Application of autocorrelation and cross-correlation analyses in human movement and rehabilitation research. J Orthop Sports Phys Ther 39, 287–295. [DOI] [PubMed] [Google Scholar]
  26. Ong JS, Touyz G, Tanner S, Hillman DR, Eastwood PR, Walsh JH, 2011. Variability of human upper airway collapsibility during sleep and the influence of body posture and sleep stage. J Sleep Res 20, 533–537. [DOI] [PubMed] [Google Scholar]
  27. Patil SP, Ayappa IA, Caples SM, Kimoff RJ, Patel SR, Harrod CG, 2019. Treatment of Adult Obstructive Sleep Apnea With Positive Airway Pressure: An American Academy of Sleep Medicine Systematic Review, Meta-Analysis, and GRADE Assessment. J Clin Sleep Med 15, 301–334. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Remmers JE, deGroot WJ, Sauerland EK, Anch AM, 1978. Pathogenesis of upper airway occlusion during sleep. J Appl Physiol Respir Environ Exerc Physiol 44, 931–938. [DOI] [PubMed] [Google Scholar]
  29. Schober P, Boer C, Schwarte LA, 2018. Correlation Coefficients: Appropriate Use and Interpretation. Anesth Analg 126, 1763–1768. [DOI] [PubMed] [Google Scholar]
  30. Schwab RJ, Pasirstein M, Pierson R, Mackley A, Hachadoorian R, Arens R, Maislin G, Pack AI, 2003. Identification of upper airway anatomic risk factors for obstructive sleep apnea with volumetric magnetic resonance imaging. Am J Respir Crit Care Med 168, 522–530. [DOI] [PubMed] [Google Scholar]
  31. Sforza E, Bacon W, Weiss T, Thibault A, Petiau C, Krieger J, 2000. Upper airway collapsibility and cephalometric variables in patients with obstructive sleep apnea. Am J Respir Crit Care Med 161, 347–352. [DOI] [PubMed] [Google Scholar]
  32. Smith SW, 2003. Digital signal processing : a practical guide for engineers and scientists. Newnes, Amsterdam; Boston. [Google Scholar]
  33. Strollo PJ Jr., Soose RJ, Maurer JT, de Vries N, Cornelius J, Froymovich O, Hanson RD, Padhya TA, Steward DL, Gillespie MB, Woodson BT, Van de Heyning PH, Goetting MG, Vanderveken OM, Feldman N, Knaack L, Strohl KP, 2014. Upper-airway stimulation for obstructive sleep apnea. N Engl J Med 370, 139–149. [DOI] [PubMed] [Google Scholar]
  34. Stuck BA, Ravesloot MJL, Eschenhagen T, de Vet HCW, Sommer JU, 2018. Uvulopalatopharyngoplasty with or without tonsillectomy in the treatment of adult obstructive sleep apnea - A systematic review. Sleep Med 50, 152–165. [DOI] [PubMed] [Google Scholar]
  35. Tangel DJ, Mezzanotte WS, White DP, 1991. Influence of sleep on tensor palatini EMG and upper airway resistance in normal men. J Appl Physiol (1985) 70, 2574–2581. [DOI] [PubMed] [Google Scholar]
  36. Truong MT, Woo VG, Koltai PJ, 2012. Sleep endoscopy as a diagnostic tool in pediatric obstructive sleep apnea. Int J Pediatr Otorhinolaryngol 76, 722–727. [DOI] [PubMed] [Google Scholar]
  37. Ulualp SO, 2014. Modified expansion sphincter pharyngoplasty for treatment of children with obstructive sleep apnea. JAMA Otolaryngol Head Neck Surg 140, 817–822. [DOI] [PubMed] [Google Scholar]
  38. Wallace A, Bucks RS, 2013. Memory and obstructive sleep apnea: a meta-analysis. Sleep 36, 203–220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Warner G, Skatrud JB, Dempsey JA, 1987. Effect of hypoxia-induced periodic breathing on upper airway obstruction during sleep. J Appl Physiol (1985) 62, 2201–2211. [DOI] [PubMed] [Google Scholar]
  40. Wheatley JR, Mezzanotte WS, Tangel DJ, White DP, 1993. Influence of sleep on genioglossus muscle activation by negative pressure in normal men. Am Rev Respir Dis 148, 597–605. [DOI] [PubMed] [Google Scholar]
  41. Wheaton AG, Perry GS, Chapman DP, Croft JB, 2012. Sleep disordered breathing and depression among U.S. adults: National Health and Nutrition Examination Survey, 2005–2008. Sleep 35, 461–467. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Worsnop C, Kay A, Pierce R, Kim Y, Trinder J, 1998. Activity of respiratory pump and upper airway muscles during sleep onset. J Appl Physiol (1985) 85, 908–920. [DOI] [PubMed] [Google Scholar]
  43. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, Gerig G, 2006. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage 31, 1116–1128. [DOI] [PubMed] [Google Scholar]

RESOURCES