Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/268413402

CombiningMIMICSandComputationalFluidDynamics(CFD)toassesstheefficiencyofaMandibularAdvancementDevice(MAD)totreatObstructiveSleepApnea(OSA)

ARTICLE

READS

32

7AUTHORS,INCLUDING:

OlivierMVanderveken

UniversitairZiekenhuisAntwerpen

89PUBLICATIONS780CITATIONS

SEEPROFILE

MarcBraem

UniversityofAntwerp

115PUBLICATIONS5,313CITATIONS

SEEPROFILE

Dirkvandyck

UniversityofAntwerp

128PUBLICATIONS2,409CITATIONS

SEEPROFILE

Availablefrom:OlivierMVanderveken

Retrievedon:05February2016

Combining MIMICS and Computational Fluid Dynamics (CFD) to assess the efficiency of a Mandibular Advancement Device (MAD) to

treat Obstructive Sleep Apnea (OSA)

Ir. J. De Backer1,2, Ir. W. Vos1,2, Dr. O. Vanderveken2, Dr. A. Devolder2, Prof. M. Braem3, Prof. D. Van Dyck 1, Prof. W. De Backer2

1University of Antwerp, Department of Physics, Antwerp, Belgium

2University Hospital of Antwerp, Department of Pulmonology, Antwerp, Belgium 3University Hospital of Antwerp, Department of Dentistry, Antwerp, Belgium

Abstract Obstructive Sleep Apnea (OSA) is a condition in which the patient repeatedly stops breathing during sleep due to the collapse of the upper airway. The upper airway collapse is caused by a relaxation of the upper airway muscles, combined with a decrease in intraluminal pressure and hence an increase in external pressure from the surrounding tissue. OSA can occur in both male and female patients from all age categories and is becoming more and more frequent. It is estimated that today approximately 4% of the global male and 2% of the global female populations suffers from OSA. A variety of treatments exists to treat OSA, ranging from external appliances like Continuous Positive Airway Pressure (CPAP) to surgical procedures such uvulopalatopharyngoplasty (UPPP). Another treatment that is becoming more and more popular, due to its reversibility, is the oral intervention like a Mandibular Advancement Device (MAD). This device brings the mandibula forward in order to increase the upper airway volume and prevent total upper airway collapse during sleep. The patient only uses the device during the night. However the efficiency of the MAD can vary significantly from patient to patient. Combining MIMICS software and Computational Fluid Dynamics (Fluent) allows for a prediction of the outcome of a treatment with a MAD. By making a CT scan of a patient with and without the MAD in place, it is possible to compare the change in upper airway volume and the anatomical characteristics through the conversion to CAD models with MIMICS. Often this approach can give a good first indication of the efficiency of the device. When this is not conclusive, the change in upper airway resistance can be calculated through flow simulations. Thereby the CAD model (.stl) is used to create a finite volume mesh to solve the numerical flow equations on. Boundary conditions for the model (inlet and outlet pressures, mass flow, etc) are obtained from the patient during a sleep study. Therefore the flow modelling is based on patient specific geometry and patient specific boundary conditions. Whenever the upper airway resistance decreases, also a clinical improvement can be observed. Clinical improvements are measured by looking at the apnea-hypopnea index (AHI) which indicates the number of events (upper airway closures or near closures) per hour and the snoring index quantifying the degree of snoring. Table 1 gives an overview of the results. In conclusion, one can say that the combination of advanced imaging and functional analysis shows a large potential for future medical treatments.

1 Introduction Obstructive Sleep Apnea Syndrome (OSAS) is a sleep related breathing disorder characterized by repeated collapse of the upper airway. The closure of the upper airway causes the patient to stop breathing. This leads to oxygen desaturation, followed by an awakening of the patient. The result of having OSAS is sleep fragmentation. OSAS can be considered as a potentially very dangerous disease with possible severe consequences such as hypertension, atherosclerosis or stroke and even heart failure(1). Symptoms of OSAS include amongst others daytime sleepiness, morning headaches and snoring. In OSAS the airway collapses at the end of the expiration(2). This is caused by an increased pressure of the surrounding tissue on the airway due to relaxation of the muscles and a decrease in intraluminal pressure at the end of the expiration. Several treatments exist for reducing or eliminating the effects of OSAS, the most effective one is Continuous Positive Airway Pressure (CPAP). Patients using CPAP, sleep with a mask that raises the intraluminal pressure in order to prevent airway collapse. Although efficient, CPAP is not always

tolerated well by the patients due to the relatively high discomfort of sleeping with a mask. Therefore other treatments are used such as uvulopalatopharyngyoplasty (UPPP) where parts of the upper airway are removed surgically to increase upper airway volume. The success rate of this kind of treatment does however vary quite substantially. An alternative that has become more and more popular due to its reversibility is the oral device such as the Mandibular Advancement Device (MAD). The MAD consists of a mouth piece that is used only during the night. It brings the lower jaw or mandibula forward, thereby increasing the upper airway volume. .

Figure 1 Marklund Mandibular Advancement Device

Figure 1 presents such a MAD as developed by Marklund(3). The success rate of the MAD can be estimated between 60% and 70%. The aim of the current study is to see whether the efficiency of the MAD treatment can be correlated with anatomical characteristics and Computational Fluid Dynamics (CFD) calculations based on advanced imaging techniques. This correlation can be used to predict the outcome of the treatment in advance thereby increasing the success rate of the technique and at the same time optimizing patient care and reducing cost. Section 2 describes the study outline and the methods that were used; this includes the imaging techniques and the clinical settings. Section 3 presents the findings. Further discussion in these findings can be found in section 4 and finally section 5 formulates the conclusions.

2 Description of the study and the methods used Two methods were used: one method described the functional imaging process (Figure 3); the other method outlined the clinical settings that were used to validate the functional imaging modelling (Figure 2).

Figure 2 Workflow clinical setting

For this project, 8 patients in total (7 male /1 female) were studied. All the patients were diagnosed with a mild to moderate form of sleep apnea and/or symptomatic snoring. All patients underwent a split night polysomnography. Polysomnography is a diagnostic test during which a number of physiologic variables are measured and recorded during sleep(4). For this study the flow was measured using a pneumotach and the intraluminal pressures were measured using a multi-sensor pressure catheter during the night. The flow measurements allowed for an assessment of the number of apneas, the location of collapse could be determined using the multi sensor catheter. The first half of the night the patient slept without MAD, halfway during the night the patient was asked to use the MAD. This test allowed for a clinical assessment of the efficiency of the MAD. The morning after the polysomnography, two CT scans per patient were taken; one with and one without MAD. The CT scanner type was Siemens Sensation 16 with a slice thickness of 0.5 mm. The patient was scanned from the nasopharynx down to the larynx. The scanning area was chosen such that the region of collapse was included but was kept to a minimum to limit the amount of radiation that the patient was exposed to. These CT scans provided images of both the change in airway volume as well as the position of the mandibula (lower jaw) with respect to the maxilla (upper jaw).

Figure 3 Workflow functional imaging

Figure 4 shows the sagital reconstructions of the scan taken from a patient suffering from moderate sleep apnea. The right image shows the patient without the MAD, in the left image the patient is using the mouthpiece. The slices are taken at the same position as indicated by the slice number (120.00). Based on these CT images the maxilla, mandibula and upper airway can be reconstructed into three-dimensional Computer Aided Design (CAD) models. These CAD models are based on the gradient in Hounsfield units (HU). Every part of the human body is represented in a CT scan through different Hounsfield units.

Figure 4 Sagital reconstruction of patient with (left) and without (right) MAD Table 1 presents the typical value of Hounsfield units for parts of the human body. For our project we are primarily interested in air and bone. So the opposites of the Hounsfield scale (-1000 and 1000) are of importance to us.

Table 1 Hounsfield Units for human body

Hounsfield Units for human body Bone 1000 Liver 40 to 60

White Matter 46 Grey Matter 43

Blood 40 Muscle 10 to 40 Kidney 30

Cerebrospinal Fluid 15 Water 0

Fat -50 to -100 Air -1000

The program that constructs three-dimensional models based on Hounsfield units is called Mimics and is developed by Materialise. The CT data in DICOM format is read into the program using a CT compression. In order to reconstruct 3D images based on Hounsfield units, the appropriate voxels must be grouped accordingly. To this end masks are created that contain voxels with predefined Hounsfield units. Since we are interested in constructing the maxilla, mandibula and upper airway three masks are created. To capture all the air, the upper airway masks ranges from -700HU to -1024HU. The maxilla and mandibula are defined through masks with HU between 226 and 3071.

Figure 5 Illustration of masks containing mandibula (red) maxilla (blue) and upper airway (yellow)

Figure 5 depicts the masks that were used for patient with moderate sleep apnea. The red mask contains the mandibula, the maxilla is defined by the blue mask and the collapsible part of the upper airway can be found in the yellow mask. The airway region was chosen such that the effect of the mandibular advancement was included. Once the masks were defined accurately the three dimensional models could be constructed. Figure 6 presents the CAD models of the mandibula, maxilla and upper airway from a patient suffering from sleep apnea reconstructed based on the HU in the masks. On the left the patient is depicted without the MAD, the right image shows the effect of the MAD on the mandibular position and the upper airway volume. The three dimensional CAD model can be exported in a universal format such as binary stereo lithography (.stl) files or Drawing eXchange Format (.dxf) files. Subsequently these files can be post-processed with CAD packages such as Autodesks’AutoCAD or Robert McNeels’ Rhinoceros. These programs allow for a full three dimensional manipulation of the model (rotating, translating,…), they also allow for an accurate measurement of distance and angles. This feature will be used to analyse the models in more depth (mandibular angle, upper airway volume etc).

Figure 6 Three dimensional reconstruction of mandibula, maxilla (upper part) and upper airway

(lower part) for patient suffering from sleep apnea Now that the upper airway volume has been converted into a three dimensional image, it can be used to analyse the flow behaviour inside the airway using Computational Fluid Dynamics (CFD) for both cases. CFD is a method that allows for a mathematical description of flow in a domain based on pre-defined boundary conditions(5). To this end the Navier Stokes equations are solved numerically. This means that these equations are solved in discrete volume elements, also referred to as a computational grid. The results of a CFD calculation is that the local flow characteristics are know in every computational cell. The CFD technique can be used to make an assessment of the change in airway resistance due to the MAD.

Figure 7 Upper Airway divided into discrete cells

Figure 7 presents an upper airway model divided into discrete cells. A typical upper airway model contains between 600,000 and 800,000 cells, depending on the complexity of the model. This amount of cells limits the cell Reynolds number and thereby minimises the numerical diffusion (errors due to numerical schemes). Due to the narrowness of the upper airway region, the flow could be considered turbulent(6). Therefore a turbulence model is used in the calculations. For these types of problem the k-ω has been used. The methods as described above have been used to generate the models of all the patients in the study. The next section describes the results of this modelling.

3 Results The method from the previous section has been applied to all 8 patients of the project. The treatment was unsuccessful for 3 out of 8 patients (37,5%). The primary aim of the study was to see whether or not functional upper airway modeling could predict the efficiency of the device. If this was the case then future patients could be analysed with a less expensive wax bite registration to see if they would benefit from a MAD or not. This would lead to better patient care at a lower cost and could set an example for future research projects. An overview of the patient data and the outcome of the clinical evaluation are given in Table 2. For every patient the age, sex and body mass index (BMI) is given. The clinical evaluation has been done through the assessment of the apnea-hypopnea index. This index lists the number of events per hour. An event is defined as a closure or near closure of the upper airway and can be determined from the polysomnography that registered the flow during sleep. An apnea or hypopnea is characteristed by a limitation in flow (hypopnea) or a complete cessation of flow (apnea). Also a decrease in snoring index is part of a clinical improvement especially in the non-apneic snorers (AHI<5)

Table 2 Overview of patient data and clinical evaluation Patient Age Sex BMI AHI -MAD AHI +MAD clinical improvementVDBRA 48 M 27 27 9 Y

JT 60 M 34 31 9 YSC 57 M 25 14 2 Y

VDBRO 54 M 25 16 1 YLL 45 F 26 1 0 Y

VVM 44 M 31 1 4 NCG 52 M 29 12 14 NSM 51 M 24 22 21 N

The age of the patients ranged from 44 to 60 with BMI’s between 24 and 34. Figure 9 shows the 3D reconstruction of maxilla, mandibula and upper airway for a successfully treated patient on the left and for a patient who didn’t clinically improve on the right. The unsuccessful patient even showed an obstruction of the upper airway when the MAD was used. This explains the interruption in the upper airway volume. For all the patients in the study, an anatomical assessment was made where special attention was paid to the angle formed by the mandibula or Angulus Mandibulae. The base of the tongue determines the shape of the upper airway. The tongue base is connected to the front of the mandibula (spina mentalis) via the musculus genioglossus(7), this can be seen in Figure 8. It is therefore important to consider what happens to the spina mentalis when the mandibula is advanced since this will determine the movement of the tongue base. The Angulus Mandibulae determines the position of the spina mentalis after mandibular advancement.

Figure 8 Anatomical drawing of Musculus Genioglossus

The Angulus Mandibulae was assessed for all patients using the angle measurement tool in the CAD program. To determine this angle two construction lines were created. One line connected the midpoint of the arc formed by the Angulus Mandibulae with the center of the Condylus (black line). The other line contained the same midpoint and acted as a tangent line to the lower part of the mandibula (blue line). The angle between the blue and the black line is considered to be the Angulus Mandibulae.

The actual mandibular advancement is indicated in Figure 9 through the vertical white lines. This illustrates the effect of the Angulus Mandibulae on the horizontal mandibular advancement. An in-depth analysis can be found in the discussion in the next section. It can be observed that the model on the left has more detail in the dental region while the model on the right shows a more crude shape. This is a consequence of the fact that the patient on the right has more artifacts or fillings in his teeth compared to the patient on the left. The material used in teeth fillings is radiopaque and distorts the CT image. The distortion can be reduced by manually adjusting the masks but it will never be as accurate since the original shape of the teeth cannot be seen in the CT images anymore. This can be avoided by scanning the plaster dental models and placing them inside the rest of the 3D model. Since the actual shape of the teeth is not of principal importance, this is beyond the scope of this project.

Figure 9 Sagital images of a succesfully treated patient (left) and an unsuccesfully treated patient (right)

Besides the mandibular angle, also the change in upper airway volume was calculated. Enlarging the upper airway volume was the main goal of a MAD. The change was determined by comparing the upper airway volumes extracted from the CT scans. To ensure that the comparison was accurate the upper airway model limits were exactly the same for both cases. This could be done by considering the slice numbers. Table 3 gives an overview of the mandibular angles and the change in upper airway volume as determined for all patients.

Table 3 Overview of Angulus Mandibulae and change in Upper Airway Volume for all patients

PatientAngulus

Mandibulae (degrees)

Change in Upper Airway Volume (%)

Clinical Impvrovement

VDBRA 110 27.6 YJT 118 320.0 YSC 110 54.5 Y

VDBRO 109 40.4 YLL 113 31.7 Y

VVM 138 0.4 NNN

CG 130 -54.1SM 135 -6.2

From Table 3 it can be seen that a positive change or enlargement in upper airway volume has a clinical improvement as a consequence. Apparently enlarging the airway enhances the airflow in the upper airway presumably by redcuing the critical closing pressure thus preventing the upper airway from collapsing during expiration. However patients were the upper airway volume didn’t change or even decreased also did not improve clinically. When comparing the change in upper airway volume with magnitude of the Angulus Mandibulae it is obvious that a larger angle of the mandibula as defined previously leads to a negative or non-existing change in upper airway volume.

Effect of Mandibular Angle on Upper Airway Volume Change for MAD patients

-100.0

-50.0

0.0

50.0

100.0

150.0

200.0

250.0

300.0

350.0

100 110 120 130 140Angulus Mandibulae (degrees)

Upp

er A

irway

Vol

ume

Cha

nge

MA

D+

(%)

Figure 10 The change in upper airway volume as a function of mandibular angle

Figure 10 presents the change in upper airway volume as a function of the mandibular angle. The red line shows that there is a clear trend indicating that the larger the mandibular angle is, the smaller the change in upper airway volume that can be achieved through mandibular advancement. To quantify the effect of the change in airway volume, the resistance of the upper airway to the flow is calculated using CFD. The resistance can be defined as:

FpR ∆

=

Where pressure difference over the airway and p∆ F is the flow. The boundary conditions (flow and pressures) are taken from the in-patient measurements. Figure 11 presents vectors couloured by velocity and pressure for the cases with and without MAD.

Figure 11 Vectors coloured by velocity (top) and pressure (bottom) for situation without (left) and

with (right) MAD Based on these flow characteristics throughout the domain an assessment of the change in upper airway resistance can be made. Table 4 gives an overview of the changes in upper airway resistance for all the patients in the study. For the successfully treated patients, the change in upper airway resistance varied between -39% and -72%. A decrease in AHI corresponds to a decrease in airway resistance.

Table 4 Overview of change in Upper Airway Resistance for all patients

PatientChange in Upper

Airway Resistance (%)

Clinical Impvrovement

VDBRA -56 YJT n/a YSC -53 Y

VDBRO -72 YLL -39 Y

VVM 5 NNN

CG n/aSM 4

The next section discusses the results in more detail.

4 Discussion This section discusses the results in more detail; we will try to link the anatomical physics and flow characteristics with the clinical outcome of the treatment. In Figure 9 it can be seen that the vertical white stripes give an indication of the horizontal mandibular advancement. The lower teeth are used as a reference. This is also the reference point used by dentist to set and adjust the protrusion of the device. As a golden standard the device is set to approximately 65% of the maximal protrusion of the patient. The maximal protrusion is defined as the maximum forward position of the lower teeth with respect to the upper teeth. Although the maximum protrusion would most likely give the best results it cannot be applied since it would not be tolerated well by the patient. The vertical lines indicate that for both patients the lower teeth do show a horizontal advancement. However, when looking at the successfully treated patient on the left it can be seen that also the Spina mentalis is advanced horizontally. Although the lower teeth of the patient on the right are also advanced, to a lesser extent nevertheless, the Spina mentalis stays virtually at the same location. This would mean that the muscle connected to the tongue base does not pull the tongue forward and therefore does not increase the upper airway volume. Indeed when considering the upper airway volume it can be seen that this volume is not enlarged; on the contrary the rotation of the lower jaw even causes an occlusion of the upper airway at velopharyngeal level. Anatomical features do play an important role in these phenomena. The magnitude of the mandibular angle as defined in the previous section determines to an extent the horizontal advancement and the change in upper airway volume. A larger mandibular angle leads to a smaller mandibular advancement. A larger mandibular angle also seems to limit the horizontal advancement of the lower teeth when considering the 65% rule. For patients with a mandibular angle larger than 130° no change or even a negative change in upper airway volume was determined. The successfully treated patients all had a mandibular angle below 120°. The Computational Fluid Dynamics analysis is a mean to quantify the change in upper airway resistance when using the MAD. The results show, as expected, that an overall increase in upper airway volume reduces the resistance to the flow. However, local changes in flow characteristics do occur and can be examined using CFD. Especially the pressure at the smallest cross sectional area is of importance since this is usually the place of collapse. An efficiently treated patient will show a local increase in pressure at this location, indicating that the local cross sectional area also has increased. It is possible that a patient shows a global increase in upper airway volume but that locally, at the smallest cross sectional area, the change is insufficient to prevent the upper airway collapse. These nuances are analysed using CFD. For two patients, the flow analysis could not be performed, since their upper airway already showed a collapse in one of the two situations (JT without MAD, CG with MAD). Due to a lack of a continuous domain, flow could not pass through the model, hence the resistance for this situation could be considered infinite. For these patients, however, the change in upper airway volume already clearly indicated whether or not the treatment would be successful.

5 Conclusions The aim of the project was to see whether the efficiency of a Mandibular Advancement Device could be predicted based on anatomical features and Computational Fluid Dynamics analyses. If this turned out to be possible then the success rate of this treatment could be significantly increased. This would mean that patients would be better treated at lower cost. During the course of the project, it became clear that indeed certain anatomical and flow characteristics could be considered as markers for success. Three aspects in particular were examined: change in upper airway volume, change in upper airway resistance and the mandibular angle. By combining advanced imaging techniques such as high resolution CT scans and MIMICS with flow simulation tools (CFD) it becomes possible to attain functional imaging methods. These functional imaging tools made it possible to relate the three aspects to the clinical outcome and to each other. The study showed that it was possible to predict the outcome of the MAD treatment based on the change in upper airway volume, the change in airway resistance and the mandibular angle. For all patients a correct assessment was made. The method as described will be part of the protocol for MAD treatment in the future. It is the authors’ hope and believe that these high-tech methods will find their way into clinical medicine to prove their value.

6 References Reference List

(1) Wolk R, Somers VK. Cardiovascular consequences of obstructive sleep apnea. Clin

Chest Med 2003 June;24(2):195-205.

(2) Vanderveken OM, Oostveen E, Boudewyns AN, Verbraecken JA, Van de Heyning PH, De Backer WA. Quantification of pharyngeal patency in patients with sleep-disordered breathing. ORL J Otorhinolaryngol Relat Spec 2005;67(3):168-79.

(3) Marklund M, Stenlund H, Franklin KA. Mandibular advancement devices in 630 men and women with obstructive sleep apnea and snoring: tolerability and predictors of treatment success. Chest 2004 April;125(4):1270-8.

(4) Goldstein NA, Sculerati N, Walsleben JA, Bhatia N, Friedman DM, Rapoport DM. Clinical diagnosis of pediatric obstructive sleep apnea validated by polysomnography. Otolaryngol Head Neck Surg 1994 November;111(5):611-7.

(5) Anderson J. Fundamentals of Aerodynamics. second ed. McGraw-Hill; 1991.

(6) Nowak N, Kakade PP, Annapragada AV. Computational fluid dynamics simulation of airflow and aerosol deposition in human lungs. Ann Biomed Eng 2003 April;31(4):374-90.

(7) Putz R. Sobotta. Houtem: Bohn Stafleu Van Loghum; 1994.