D

Sa

b

a

A

KNPHT3

1

tnmtsodmoioisi1tttta2

1d

Respiratory Physiology & Neurobiology 163 (2008) 111–120

Contents lists available at ScienceDirect

Respiratory Physiology & Neurobiology

journa l homepage: www.e lsev ier .com/ locate / resphys io l

igital particle image velocimetry studies of nasal airflow

eung-Kyu Chunga,∗, Sung Kyun Kimb

Department of Otorhinolaryngology, Head & Neck Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul 135-710, South KoreaDepartment of Mechanical Engineering, Konkuk University, Seoul 143-701, South Korea

r t i c l e i n f o

rticle history:Accepted 21 July 2008

a b s t r a c t

Understanding the properties of airflow in the nasal cavity is essential to understanding physiologic and

eywords:asal airflowarticle image velocimetryuman nasal cavity

pathologic aspects of nasal breathing. Many attempts have been made to evaluate nasal airflow patternsusing the best possible analytical methods available at the time. Recently, digital particle image velocime-try (DPIV) and computational fluid dynamic methods have been applied to this area. Digital PIV is anexperimental method used to evaluate airflow in an accurately reproduced transparent model of thenasal cavity. In this review, use of the DPIV procedure in the study of nasal airflow, airflow patterns inquiet respiration, and changes to airflow after modification of the nasal turbinates are reviewed, along

echni

1

wtdtcfoeasalgeitta

m

urbinectomyD model with aspects of the DPIV t

. Introduction

To understand nasal functions such as humidification, filtration,emperature control of inspired air, and olfaction, knowledge ofasal airflow properties is important. Many attempts have beenade to evaluate nasal airflow patterns using the most up-to-date

echniques available at a particular time in question. Using theame principle that the movement of dust particles can be rec-gnized in a sliver of sunlight passing through a window, so toooes the technique of particle image velocimetry (PIV) enable theeasurement fluid velocity in a flow passage constructed using

ptically transparent materials (Adrian, 1991). In this way, the flows seeded with tracer particles and is illuminated with a sheetf laser light. Two- or three-dimensional velocities within thelluminated plane are calculated with post processing after twoubsequent positions of the particle are recorded. Digital PIV (DPIV)s the digital counterpart of the PIV technique (Willert and Gharib,991), where digitally recorded video images are analyzed compu-ationally. While there was one difficulty in applying this techniqueo complex flow passages such as the nasal cavity and therebyo replicate these structures with optically transparent materials,his problem was overcome using the rapid prototyping methodnd water-soluble materials for negative models (Hopkins et al.,

000).

∗ Corresponding author. Tel.: +82 2 3410 3572; fax: +82 2 3410 3879.E-mail address: rhinochung@gmail.com (S.-K. Chung).

i1tt(sno

569-9048/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.resp.2008.07.023

que and the future role of DPIV in this field of research.© 2008 Elsevier B.V. All rights reserved.

.1. Studies of nasal airflow

In 1882 Paulsen first described inspiratory airflow in the nose,ith air rising fairly vertically along the bridge of the nose

oward the anterior end of the middle turbinate, where it iseflected and passes between the middle turbinate and the sep-um towards the posterior end of the inferior turbinate and thehoana (Uddströmer, 1940). Thereafter, many investigators per-ormed airflow experiments using cadavers as models or usingther materials (Uddströmer, 1940). From the late 1930s onwards,xperimental models were created from the nasal casts of cadaversnd nasal airflow was examined with various kinds of methods,uch as cigar smoke (Tonndorf, 1939), smoke (Proetz, 1951), anngle meter (Proctor, 1966), dye injection (Swift and Proctor, 1977),aser-Doppler velocimetry (Girardin et al., 1983), radioactive Xenonas (Hornung et al., 1987) or aerosolized water particles (Simment al., 1999). The models made from the cadavers have inherent lim-tations associated with the postmortem shrinkage of soft tissue orhe use of flat transparent nasal septa for visualization, meaninghat more accurate and/or enhanced size models that reproducectual conditions are required.

In the early 1990s, nasal cavity models were created usingagnetic resonance imaging (MRI) or computed tomography (CT)

mages. Three times (Schreck et al., 1993) or 20 times (Hahn et al.,993) scale models of the nasal cavity were made using clear plas-ic or Styrofoam slabs. Enlarged coronal images of the nasal airway,

aken from normal subjects after decongestion of the nasal cavitiesor not), were used as templates to cut clear plastic or Styrofoamlabs. The slabs were glued together serially to produce enlargedasal cavity models, and flow measured by using a hot anemometerr a hot film anemometer.

1 siology

aifdtsaesnp2tphnpmiweta

tott(a22lettt(2

i(aaae

2

2

ismtfltotfiem

pmiepitauiiccoptid

s(iTvvpPLbarwafig

2a

iohtppappfcapi2

i1ttu

12 S.-K. Chung, S.K. Kim / Respiratory Phy

Since the traditional forms of flow visualization, such as hot wirenemometer and laser-Doppler velocimetry, could no longer sat-sfy the needs of modern research on flow dynamics, where floweatures must be quantified instantaneously, the PIV method waseveloped to measure velocity fields non-intrusively and instan-aneously (Huang, 1994). The DPIV method was applied to thechematic anatomic model of the nasal cavity (Park et al., 1997), annatomically accurate nasal cavity model based on CT images (Kellyt al., 2000). Hopkins et al. recently established a procedure to con-truct a transparent rectangular box containing a X2 model of theasal cavity for PIV measurement using a combination of the rapidrototyping method and the curing of clear silicone (Hopkins et al.,000). Creating the accurate transparent flow passages is essentialo analyzing by PIV flow inside the complex flow passage. Theserocedures were improved to produce the nasal cavity model usingigh-resolution CT scan data in the “digital imaging and commu-ications in medicine” (DICOM) file format and surface renderingrocedures (Kim and Son, 2002; Kim and Haw, 2004b). An amendedethod was applied to evaluate the change of airflow after var-

ous kinds of middle turbinectomies (Kim and Chung, 2004a), asell as patterns of nasal flow during the respiratory cycle (Chung

t al., 2006). However, difficulties exist to apply DPIV methods tohe many kinds of experimental models representing normal andbnormal characteristics of the nasal cavities.

The numerical simulation method, also known as the computa-ional fluid dynamics (CFD) method, was first applied to an analysisf airflow in the nasal valve area (Tarabichi and Fanous, 1993) and tohe study of airflow in the nasal cavities (Keyhani et al., 1995). Sincehen, the CFD method has been applied to the study of nasal airflowHörschler et al., 2003; Weinhold and Mlynski, 2004; Ishikawa etl., 2006), the distribution of temperature (Lindemann et al., 2004,006), odorant solubility or transport (Kurtz et al., 2004; Zhao et al.,004, 2006a), deposition of nanoparticles (Shi et al., 2006), patho-

ogic conditions of the nasal cavity such as septal perforation (Grantt al., 2004; Pless et al., 2004; Ozlugedik et al., 2008), atrophic rhini-is (Garcia et al., 2007), and surgical conditions such as inferiorurbinate reduction (Wexler et al., 2005), various kinds of turbinec-omy (Lindemann et al., 2005b; Hörschler et al., 2006), polypectomyZhao et al., 2006b), and radical sinus surgery (Lindemann et al.,005a).

Given that a comparison of the results of experimental methodsncluding DPIV methods and CFD methods showed good agreementKeyhani et al., 1995; Weinhold and Mlynski, 2004; Hörschler etl., 2006), nowadays CFD is more commonly applied in studies ofirflow in the nasal cavity, particularly where flow instability mayrise, though CFD requires validation by the DPIV technique or otherxperimental methods.

. Digital particle image velocimetry

.1. What is digital particle image velocimetry?

One of the oldest techniques for studying velocities in a fluids seeding particles into the fluid. When the tracer particles aremall and their specific weight is close to that of the fluid, they willove locally with almost the same velocity as the fluid. Tracking

he tracer particles will therefore give information about the localuid movement. PIV is a basic method in fluid dynamics researchhat provides high spatial resolution in a two-dimensional slice

f the flow (Adrian, 1991). It is usually a planar laser light sheetechnique in which the light sheet is pulsed twice, and images ofne particles lying in the light sheet are recorded on a video cam-ra or a photograph. The displacement of the particle images iseasured in the plane of the image and used to determine the dis-

vsidh

& Neurobiology 163 (2008) 111–120

lacement of the particles in the flow. The most common way ofeasuring displacement is to divide the image plane into small

nterrogation spots and cross-correlate the images from two timexposures. Essentially a matrix of grey level values, comprising theixel image intensity values in an interrogation region of the first

mage, is compared with the image matrix of the second image,o determine whether similar features are found in both images,nd if so, their relative spatial displacement. Cross-correlation issed to determine whether the pattern of imaged particles in the

nterrogation region corresponds to a similar pattern in the secondmage. The magnitude of the correlation peak corresponds to howlosely the patterns are associated, and the displacement of theorrelation peak from the origin provides a best statistical estimatef the mean displacement of the associated patterns. Modern PIVrocessing techniques employ a range of techniques to maximizehe accuracy and robustness of the correlation estimate, includ-ng correlation scaling and fitting, adaptive region size and imageistortion.

Digital PIV is the digital counterpart of conventional laserpeckle velocimetry (LSV) and the photographic PIV techniqueWillert and Gharib, 1991). Because PIV and LSV rely on singlemages for flow velocity measurements, phase information is lost.his causes a directional ambiguity for each recovered velocityector in the field. With the DPIV technique, digitally recordedideo images are analyzed computationally, removing both thehotographic and opto-mechanical processing steps inherent toIV and LSV. The directional ambiguity associated with PIV orSV is resolved by implementing local spatial cross-correlationsetween two sequential single-exposed particle images (Willertnd Gharib, 1991). Experimental investigation of turbulent flowsequires techniques that allow three-dimensional measurementsith high spatial and temporal resolutions. There exist several vari-

nts of PIV to measure the three-dimensional (3D) flow velocityelds: Stereoscopic PIV (SPIV), Defocusing DPIV (DDPIV), Tomo-raphic PIV (TPIV), and Holographic PIV (HPIV).

.1.1. Recent developments in particle image velocimetrylgorithms

Stereoscopic PIV enables the measurement of all three veloc-ty components inside a relatively thicker light sheet. It is basedn the principle of stereoscopic imaging that is well known fromuman eyesight—two cameras image the illuminated flow par-icles from different angles. The combination of both camerarojections allows the reconstruction of the “real” particle dis-lacement inside the measurement area and the evaluation ofll three velocity components (Arroyo and Greated, 1991). Dual-lane SPIV is an extension of SPIV to cover a multi-plane or aanoramic view of a widened domain. The measurement systemor this method is composed of two double pulse lasers, fourharge-coupled device (CCD) cameras, a combination of optics,nd a sophisticated synchronization unit to cover two planes. Thisrocedure has been utilized to achieve three component veloc-

ty fields with SPIV temporal capability (Kähler and Kompenhans,000).

Defocusing DPIV is a method that uses the defocus principle todentify the 3D locations of seeded particles (Willert and Gharib,992). This technique was successfully applied to measuring therajectories of bubbles in two-phase flow (Pereira et al., 2000). Par-icle Tracking Velocimetry (PTV) is the optimal technique to besed when the particles are sparsely distributed in the observation

olume, while PIV is the method of choice when the particle den-ity is high and the velocity gradients across the cross-correlationnterrogation window are small. PIV is the preferred method in two-imensional (2D) measurements. For 3D measurements, 3D PTVas also been presented (Pereira et al., 2006).

siology

acokicuu

igpHl(

iaans

2

2t

mAiMMs

oaeatwtaa

amtotspoiroatitciFoo

fp

cpati

2

cmra2aoOuoStsPcic

toaflfliab

2

tav(i6ial

2

2

ait

Q

S.-K. Chung, S.K. Kim / Respiratory Phy

Tomographic PIV is an extension of the stereoscopic 3D conceptnd enables the instantaneous measurement of all three velocityomponents in a complete 3D measurement volume. It is basedn the principle of tomographic volume reconstruction that is wellnown from MRI. Recordings of the 3D light intensity field of thelluminated particles are reconstructed from the images of fourameras viewing from different angles. Velocity vectors in the vol-me are computed by an iterative 3D cross-correlation techniquesing deformed interrogation volumes (Elsinga et al., 2006).

The last technique to be introduced is HPIV, which uses volumellumination and holographic recording procedures. The holo-raphic concept is superior to the other methods since it canrovide an instantaneous 3D field with high spatial resolution.owever, current holography does not allow for the statistical col-

ection of data and is limited to relatively simple flow configurationsHinsch, 1995, 2002).

As the nasal cavity is composed of thin passages, nasal airflows nearly 2D, with the exception of regions near the nasal valvend nasopharynx. Tomographic measurement by 2D DPIV is quitedequate in most regions of the nasal cavity; however in the regionsear the nasal valve and pharynx, where flow is 3D, new techniquesuch as SPIV, TPIV, and DDPIV are recommended.

.2. Creation of nasal cavity model

.2.1. Three-dimensional computer model from computedomography data and rendering

Before the creation of actual PIV models or CFD analysis, experi-ental models of the nasal airways have to be computer generated.s mentioned in other reviews, the nasal cavity model can be an

maginary one or generated from cadaver characteristics or CT orR imaging. Although many authors described models from CT orR imaging as being anatomically accurate, not all of these repre-

ent the true nasal condition.When we compare the patency of both nasal cavities, usually

ne side is wider than the other side due to the nasal cycle. Gener-lly the wider side is selected because the manual segmentation isasier to perform for the wider side. When the width of the nasalirway is checked in coronal imaging without spraying deconges-ant, the width is similar from one side to the other and usuallyithin 1–1.5 mm. If the nasal mucosa is decongested before taking

he CT scan, the nasal airways become wider. However, for the cre-tion of an experimental model to evaluate the physiology of nasalirflow, images taken without decongestion should be used.

When a computer-generated model of the nasal cavity is made,decision of the margin between the airway lumen and the nasalucosa is very difficult to make, especially when the thickness of

he image slice is over 2.5 mm. Thus, for the basis of a model, it is rec-mmended to use the data of axial image reconstructed to have slicehickness less than 1.25 mm. The margins of the nasal cavities varyignificantly when the window width and levels are adjusted. Thishenomenon may come from the presence of mucus on the surfacef the nasal mucosa. Therefore, the trend is to make the nasal cav-ty wider such that the cavities take on the appearance of atrophichinitis. Compared to the automatic 3D reconstruction of the boner the contrast media-enhanced vessels with clear-cut margins,utomatic segmentation of the nasal cavity does not meet expec-ation as of now due to partial volume effect since the nasal cavitys relatively small and the margins are unclear. While the resolu-ion of the axial image is high, resolution of images in reformatted

oronal or sagittal planes may not be as high because the voxels not a regular hexahedron, height being longer than the length.or this reason, after automatic segmentation with threshold levelf brightness, the manual segmentation should be performed notnly in axial images but also in coronal and sagittal images. There-

wT2Ts

& Neurobiology 163 (2008) 111–120 113

ore, the anatomic and physiologic expertise of the researcher whoerforms the manual segmentation is important.

Another point to be considered concerns the extent of the nasalavity model. Because the primary mechanical source of the res-iration comes from the lung, the presence of the nasopharyngealirway is important. It is recommended that enough space aroundhe nasal vestibule is made so as to not change the direction of thenspiratory flow (Bailie et al., 2006).

.2.2. Prototyping of the experimental modelThe creation of a transparent rectangular box containing a nasal

avity model is essential for DPIV measurements to be made. Toake a transparent nasal cavity flow model, a computer-modeled

eplica prototype is possible with the rapid prototyping techniquend water-soluble material for a negative model (Hopkins et al.,000). Rapid prototyping is a generic term for rapid production ofphysical model replica. Usually the prototypes are filled with waxr cyanoacrylate to increase their strength and become insoluble.n the other hand, replicate prototypes of nasal cavities createdsing an RP machine from computer-generated models are madef water-soluble consolidated cornstarch and are brittle and porous.uch prototypes are then painted with water-soluble glue to sealhe pores at the surface, making the surface smooth and improvingtrength. The treated negative replica is suspended in a rectangularlexiglas box and is encased in transparent silicone. After the sili-one is cured in an oven, the negative replica made from cornstarchs dissolved with cold water and a transparent experimental modelontaining the nasal cavity flow passage is created.

To ameliorate the distortion of light when light passes throughhe flow passage and the silicone model, the index of refractionf the clear silicone and the working fluid should be matched; tochieve this, a mixture of water and glycerin is used as the workinguid. When the refraction indices of the silicone and the workinguid are matched accurately the margin of the nasal cavity model

n the silicone disappears completely, as does the distortion associ-ted with a grid placed behind the model. Usually the optimal ratioetween glycerol and water is 59% glycerol to 41% water by volume.

.2.3. Experimental setupWorking fluids containing particles are pumped through the

ransparent model from the nasopharyngeal portion, maintainingn input angle of 90◦. The vestibular portion is connected to a reser-oir at an angle of about 60◦ to the horizontal floor of the noseSwift and Proctor, 1977). The flow in the nasal airway is visual-zed with buoyant particles. As a tracer, glass spheres of diameter0 �m (Hopkins et al., 2000) or polyvinyl globular particles (80 �m

n diameter) (Kim and Chung, 2004a) are used. The tracer particlesre illuminated twice in rapid succession by a pulsed sheet of laseright and the image recorded by CCD camera.

.3. Validity of experimental setup

.3.1. Constant flow-rate conditionAchieving dynamic similarity with real breathing conditions in

n experimental setting using fluids requires that the created models scaled by a factor n, so that Reynolds number for the two condi-ions is matched. The relationship is described as:

mixture = n�mixture

�airQair

here Q is the volumetric flow rate and � is the kinematic viscosity.he kinematic viscosity of the working fluid is 6.55 × 10−6 m2/s at7.5 ◦C (Hopkins et al., 2000) and that of the air is 16 × 10−6 m2/s.herefore, a flow rate of 102 ml/s was used for a working fluid thatimulated the volumetric flow rate of air (125 ml/s) at a typical rest-

114 S.-K. Chung, S.K. Kim / Respiratory Physiology & Neurobiology 163 (2008) 111–120

F locityD dels. (Ap ty. Mo

iC

2

nWmaaflvttaot

2

oatdipiw

skinds of display methods. Therefore, the measured data are pre-sented in the plane showing the area where most of the flow passes.According to Keyhani’s report, major flow is in the axial directionand secondary velocities were small in comparison with the axial

ig. 1. Various display modes of average velocity and root mean square (RMS) veisplayed data is obtained from a 1 mm thickness sagittal plane in experimental molot of magnitude of velocity, (E) serial velocities and (F) Distribution of RMS veloci

ng breathing rate (Keyhani et al., 1995; Kelly et al., 2000; Kim andhung, 2004a; Chung et al., 2006).

.3.2. Dynamic similarity for respiratory cycleIn an experiment evaluating the nasal airflow pattern during

ormal respiration (Chung et al., 2006), Reynolds number and theomersley Number (frequency parameters) in the experimentalodel were matched with those numbers in a real situation for

chieving dynamic similarity with real breathing, while taking intoccount the size of the model and the characteristics of the workinguid. The period of the inspiratory–expiratory cycle and the tidalolume of the piston pump that simulate the respiration shouldhen be decided. For a twice-life-size nasal cavity model used forhe DPIV technique, where one inspiratory–expiratory cycle wasssumed to be 3 s, that cycle was 30 s in the model. The tidal volumef the nasal cavity model was calculated to 2060 ml, while that ofhe normal adult in one nasal cavity was assumed to be 250 ml.

.4. How to present DPIV results

The speed, direction and the root mean square of the flow arebtained. The average velocity can be displayed in the form ofvector plot, mean streamline, velocity in selected lines, ampli-

ude of velocity, isoline plot, and serial velocities, as well as the

istribution of root mean square (RMS) velocity in the measur-

ng plane (Fig. 1) and of the reconstructed image in the selectedlane (Fig. 2). Pictures of these presentations, except reconstructed

mages in selected planes, usually show only a few typical planes,hich are parallel to the septum. As such, it is difficult to under-

Fpv

in measured plane during inspiration under constant flow condition (120 ml/s).) Vector plot, (B) mean streamline, (C) average velocity in selected lines, (D) isoline

dified from Kim and Chung (2004a).

tand the global form of the flow in the nasal cavity with these

ig. 2. Average velocity distributions of airflow in left nasal cavity on seven coronallanes at the quiet inspiration with flow rate of 125 ml/s. Distribution of airflow isisualized three dimensionally. Modified from Chung et al. (2006).

siology & Neurobiology 163 (2008) 111–120 115

v1pctrA2oipa

2

saqtnmoarp

3

3

uttwo

mbePaafewmiwoimnlr

pcMcpes

FcaZ

accotm

toltpbd

ua1ltcl(a

sios

3

S.-K. Chung, S.K. Kim / Respiratory Phy

elocity through most of the main nasal passage (Keyhani et al.,995). Therefore, data are collected and redisplayed in the coronallane (Fig. 2). By these methods, a 3D pattern of flow in the nasalavity can be visualized from 2D images. Coronal reconstruction ofhe mean velocity and RMS value were thus performed to show 3Desults from 2D data (Kim and Chung, 2004a; Chung et al., 2006).lthough the reconstructed image looks 3D, it only contains theD vector in the measuring planes. This limitation in display meth-ds can also be applied to the 3 components of velocity data. If themages of the coronal plane are shown serially on video or com-uter, then the shape of the flow can be visualized as a movie or innimated form.

.5. Limitations

As for other solid models, this method cannot imitate changes inoft tissue form during respiration given that experimental modelsre not as flexible as living tissue. Until now, qualitative rather thanuantitative data have been obtained, making it difficult to evaluatehe results statistically. Also, evaluation of the normal pattern ofasal airflow in different populations is difficult because makingany DPIV models and taking measurements requires a great deal

f time. Given that most studies to date have been performed withsmall number of individuals, it remains to be seen how these

esults can be extrapolated to describe general properties of wholeopulations.

. Normal nasal airflow

.1. Nasal airflow under conditions of constant flow rate

Most studies to date on nasal airflow have been conductednder the condition of steady flow during inspiration or expira-ion, with the assumption that nasal airflow is quasi-steady, andhat flow patterns observed during the constant flow rate phaseould not change considerably with acceleration or deceleration

f flow (Bailie et al., 2006).The highest linear velocity is noted in the valve area, while the

ain nasal airflow during quiet inspiration was found in the spaceetween the middle meatus and the nasal septum in most earlyxperiments (Tonndorf, 1939; Proetz, 1951; Proctor, 1966; Swift androctor, 1977; Simmen et al., 1999). In a DPIV-based study of nasalirflow applied to a nasal cavity model based on CT scans (Kelly etl., 2000; Kim and Chung, 2004a), similar results were found exceptor the main pathway of airflow in the nasal cavity proper. The high-st velocity was noted in the nasal valve and relatively low flowsere noted in the olfactory region and both meatus. Although theajor portion of the airway flow was found in the inferior airway

n Kelly’s experiment, the main airflow was noted in the middle air-ay in our experiments (Fig. 2) (Kim and Chung, 2004a). Knowledgef the main airflow path in the nasal cavity is important when decid-ng on surgical methods for patients with nasal obstructions. The

ajor disagreements remain on the main pathways of airflow in theasal cavities. Connection of the flow horizontally to the vestibu-

ar portion of the experimental model in Kelley’s study might beesponsible for that difference.

Similar disagreements can be noted from early experimentserformed in this field (Uddströmer, 1940), as well as in studiesonducted using the CFD method in nasal cavity models based on

RI or CT scans. Different main airflows were noted in each nasal

avity of the same person (Fig. 3) (Zhao et al., 2004) and the princi-al pathways were highly variable within many samples (Churchillt al., 2004). The main pathways are influenced by many factorsuch as ethnic origin and shape of the external nose, especially the

rmsi

ig. 3. Contour plot of velocity magnitude in the coronal plane obtained using theomputational fluid dynamics method. The highest velocity was noted in the middleirway on the right side and in the inferior airway on the left side. Modified fromhao et al. (2004).

xis of the vestibule. The main flow tract seems to be decided by theurved line which connects the inlet and outlet axes. Therefore, itan be assumed that, if the axis of the inlet in the vestibule portionr the axis of the outlet portion in the nasopharynx is horizontal tohe nasal floor, the main flow will probably be noted in the inferior

eatus (Grützenmacher et al., 2005).The nasal cavity is composed of the nasal cavity proper and

hree meatus under the corresponding turbinates. The boundaryf this area is not clearly defined in the real nasal cavity, but isocated roughly around the inferior and middle airway withouthere being any definite margin. Therefore, how much of the flowasses through any specific area is difficult to calculate. This maye the reason that the inferior and middle airways are used in theescription of nasal airflow.

At an airflow rate of 12.5 l/min, airflow in the nasal vestibule issually laminar and changes to turbulent after it passes the valverea in cadaver models (Swift and Proctor, 1977; Girardin et al.,983). Nasal airflow at a flow rate of 180 ml/s showed disturbedaminar flow in an experimental model derived from MRI images,hough laminar-like flow is probably present in much of the nasalavity during normal breathing (Hahn et al., 1993). The airflow wasaminar for a simulated air flow rate of 125 ml/s in two DPIV studiesKelly et al., 2000; Kim and Chung, 2004a). The widths of the nasalirways probably account for discrepancies between studies.

In relation to eddy turbulence in airflow during inspiration, twomall eddies have been noted in front of the inferior turbinate andn the vestibule, while an eddy in the olfactory area suggested byther authors (Swift and Proctor, 1977) was not confirmed in DPIVtudies (Kelly et al., 2000; Chung et al., 2006).

.2. Nasal airflow during the respiratory cycle

The airflow pattern during normal quiet respiration waseported in coronal reconstructed images (Chung et al., 2006). Theaximum velocity occurred in the valve area and the main flow

tream occurred in the middle airway and middle meatus duringnspiration (Fig. 4). The inspiratory flow passed through the upper

116 S.-K. Chung, S.K. Kim / Respiratory Physiology & Neurobiology 163 (2008) 111–120

Fig. 4. Coronal reconstruction of mean velocity in the left nasal cavity. The maximum velocity occurred in the nasal valve area (C) and the main path of flow was in themiddle airway during respiration. The velocity during inspiration (B, C, and D) is faster than that during expiration (F–I). (A) The beginning of inspiration; (B) in the middleo ratorym piratoc Chun

pwrrm(

psp

t2iAtrt

4

s

snpofbmn

4

ehsTep

f inspiratory acceleration; (C) at the maximum velocity; (D) in the middle of inspiaximum velocity during expiration; (G–I) during expiration (each letter in the res

ycle (A–I) was 30 s in the model, equivalent to the real period of 3 s. Modified from

art of the nasopharynx. During expiration, the maximum velocityas lower than that during inspiration. The main flow during expi-

ation occurred in the middle airway, although expiratory air waseported to be approximately evenly distributed across the inferior,iddle, and olfactory areas between the septum and the turbinate

Proctor, 1966; Swift and Proctor, 1977).The flow patterns were similar in each inspiration and expiration

eriod, and only differed in the magnitude of flow. These findingsuggest similar flow patterns in the acceleration and decelerationhases of both inspiration and expiration (Fig. 4).

As reported by Tonndorf, eddies were noted when the inspira-ory and expiratory phases change (Tonndorf, 1939; Chung et al.,006). When expiration begins, multiple eddies can be identified

n the anterior portion of the middle turbinate and nasopharynx.small vortex was evident in the upper part of the soft palate in

he middle portion of expiration. In the latter part of expiration,elatively large eddies occurred in the posterior part of the inferiorurbinate and the upper part of the nasopharynx (Fig. 5).

. Changes to nasal airflow in pathologic conditions

With cadaver models, changes in the nasal air currents weretudied for pathologic conditions such as the presence of polyps,

DmaWf

deceleration; (E) during the change between inspiration and expiration; (F) at thery flow curve [center box] represents the location of each image). The period of oneg et al. (2006).

welling, atrophy or resection of turbinates, and for enlarged ade-oids (Tonndorf, 1939; Proetz, 1951; Swift and Proctor, 1977). Theresence of polyps in the superior meatus, and generalized swellingf the turbinates seem to have the most effect on airflow pattern. Asor the main pathway of airflow, the effects of various kinds of tur-inectomy on nasal airflow are of great interest to rhinologists whoust consider such aspects of surgical treatment prior to treating

asal obstruction.

.1. Deviated nasal septum (DNS)

Although the prevalence of DNS is high as 22.38% in Korea (Mint al., 1995), the number of asymptomatic persons with DNS isigher than that of symptomatic patients. The presence of a septalpur alone has scarcely any effect on airflow patterns (Proetz, 1951).here are no published reports on the effect of septal deviation asvaluated using DPIV methods. We thus used DPIV to investigateatterns of nasal airflow in both nasal cavities of one asymptomatic

NS patient. Generally the patterns of airflow, stream lines andajor pathways in the middle airway, were similar in both concave

nd convex sides with that of airflow in the normal nasal cavity.hile the patency of both nasal cavities are influenced by many

actors, including neurovascular control, and differs according to

S.-K. Chung, S.K. Kim / Respiratory Physiology & Neurobiology 163 (2008) 111–120 117

F beginfl he chaa tory flC

twmT

Fb

ig. 5. Streamline displays of nasal flow. Distribution of flow streams (A) during theow rate, and (D) in the middle of inspiratory deceleration. (E) Distribution during tlarge eddy at the posterior part of the inferior turbinate (each letter in the respirahung et al. (2006).

he nasal cycle (Eccles, 2000), usually uniform widths of nasal air-ays were noted on each side in the CT or MR images in a mannerore or less independent of the degree of septal deviation (Fig. 6).

his phenomenon can be explained by the presence of compen-

soaa

ig. 6. Reconstructed coronal CT images of the nasal cavity. In spite of the nasal septal doth sides.

ning of inspiration, (B) in the middle of inspiratory acceleration, (C) at the maximalnge between inspiration and expiration. (F–I) Distributions during expiration withow curve [center box] represents the location of each streamline). Modified from

atory hypertrophy of turbinates and developmental adaptationf other nasal structures (Grützenmacher et al., 2006). This mayccount for the nearly equal airflow patterns of nasal cavities withsymptomatic septal deviation.

eviation to the right side, the widths of the nasal airways are almost the same on

118 S.-K. Chung, S.K. Kim / Respiratory Physiology & Neurobiology 163 (2008) 111–120

F drawir omy. Iw s recot

4

4

mior(iaTtammdeaps

Ttm

4

adWittTmtm

ig. 7. Changes to nasal airflow after 3 kinds of middle turbinectomies. Schematicesection. c; Control model, 1; partial, 2; anterior partial, 3; total middle turbinectere noted in the space widened after middle turbinectomy. In (C), RMS velocity i

urbinectomized areas. Modified from Kim and Chung (2004a).

.2. On the turbinectomy

.2.1. Middle turbinectomyThe middle turbinate is located in the anterior portion of the

iddle nasal airway and its shape is thought to have an importantnfluence on nasal airflow. Although there is no exact definitionf the normal shape or size of the middle turbinate, it may beeshaped partially during endoscopic sinus surgery. Kim and Chung2004a) used DPIV methods to measure changes to nasal airflown three different models of turbinectomy: partial, anterior partialnd total middle turbinectomy (Fig. 7) (Kim and Chung, 2004a).he airflow through the upper airway including the middle mea-us was increased in the three turbinectomized models (Fig. 7B),s was the RMS velocity (Fig. 7C). These results are in rough agree-ent with the result that nasal airflow ventilated mainly in theiddle airway after partial middle turbinectomy in a model of the

econgested nasal cavity of a healthy test person (Grützenmachert al., 2003). However, they are different from results showing thatbsence of the middle turbinate did not influence the overall flowattern during inspiration and expiration with numerical and DPIVtudies using schematic nasal cavity models (Hörschler et al., 2006).

eti

w

ng of middle turbinectomies (A). Colored areas depict the location and amount ofn (B), reconstructed average velocity is displayed in coronal planes. Main airflowsnstructed in coronal planes. Values of RMS velocity were increased in the middle

he increase in RMS velocity after turbinectomy is consistent withhe occurrence of eddies following atrophy of, or for a resected,

iddle turbinate in the nasal cavity models (Tonndorf, 1939).

.2.2. Inferior turbinectomyBecause enlargement of the inferior turbinate is considered as

major cause of nasal obstruction, many turbinate procedures areone worldwide to enlarge the valve area and the inferior airway.e used the DPIV method to investigate the effect of three kinds of

nferior turbinectomy – partial, anterior partial and total inferiorurbinectomy – on nasal airflow. The three models of turbinec-omy were created and compared with a normal model (Fig. 8).he main airflow through the middle airway, including the middleeatus, was not overly different between the two partial turbinec-

omy models. The main pathway of airflow in the totally resectedodel was altered in the middle and inferior airways and the lat-

ral portion of the enlarged airway was filled with very low flowhat was indicative of eddies. The RMS velocity was also increasedn the two partially operated models (Fig. 8C).

In the total inferior turbinectomy model, the enlarged lower air-ay was not used as effective air tract. The inferior turbinate and

S.-K. Chung, S.K. Kim / Respiratory Physiology & Neurobiology 163 (2008) 111–120 119

Fig. 8. Changes to nasal airflow after 3 kinds of inferior turbinectomies. Schematic drawing of inferior turbinectomies (A). Colored areas depict the location and amount ofresection. c; control model, 1; anterior inferior partial, 2; anterior partial, 3; total inferior turbinectomy. In (B), average velocity is displayed in coronal planes. Main airflowsw nectot sterioV d in t

sirbctw

eocwrOaasrtTs

5

nCtcna

tspeTm

ere noted in the middle airway in the control model and the anterior partial turbihe middle and inferior airway in the mid portion and in the inferior airway in the poalues of RMS velocity were increased in partial turbinectomy models and decrease

purs seemed to serve as guide vane to ensure a homogenous veloc-ty distribution in the nasal cavity channels. In the case of totalesection of the inferior turbinate, the flow field was characterizedy massive vortices that impacted on flow distribution in the nasalavity (Hörschler et al., 2006). Similar results were reported afterotal resection of the middle and inferior turbinate investigatedith CFD methods (Lindemann et al., 2005b).

In an airflow study using CFD methods, conservative, unilat-ral reduction of the inferior turbinate induced a broad reductionf pressure along the nasal airway (Wexler et al., 2005). Airflowhanges were relatively regional: airflow in the valve area regionas only minimally affected, while airflow volume in the infe-

ior airway was increased and the flow velocity was decreased.n the other hand, in a direct flow visualization study, virtuallyll of the flow was found to run through the lower nose sectionfter an approximate 1-mm reduction of the medial and caudal

ide of the inferior turbinate (Grützenmacher et al., 2003). Ouresults of increased airflow through the middle airway after a par-ial inferior turbinectomy are different from the results of others.his difference may be due to different models and methods ofurgical manipulation being used.

tw

tp

my model. For the total turbinectomy model, the main stream was found betweenr portion of the nasal cavity. In (C), RMS velocity is reconstructed in coronal planes.he total turbinectomy model.

. Future role of digital particle image velocimetry

The availability of well-made computer-generated models of theasal cavity is the key to obtain reliable outcomes in the DPIV orFD experiments. Thus, for the basis of a model, it is recommendedo use the data of axial image taken from not decongested nasalavities reconstructed to less than 1.25 mm in slices thickness. Untilow, 2D vector of sagittal plane, which is major component of nasalirflow, has been investigated with DPIV techniques.

Although the secondary flow from major axial flow is small inhe nasal cavity, secondary flow is anticipated along the anteriorurface of the middle turbinate during quiet inspiration and in theosterior space of the inferior turbinate during expiration. Also, lat-rally located eddies are anticipated after structural modifications.his phenomenon cannot be resolved accurately with the 2D DPIVethod. To obtain a more detailed 3D airflow pattern, evaluation of

he 3 components of nasal airflow velocity should be investigatedith the 3D DPIV technique.

Evaluation of normal patterns of nasal airflow in certain popula-ions is very difficult because constructing many DPIV models anderforming detailed measurements requires a great deal of time.

1 siology

Tsm

lL

A

F(

R

A

A

B

C

C

EE

F

G

G

G

G

G

G

H

H

H

H

H

H

H

H

I

K

K

K

K

K

K

K

K

L

L

L

L

M

O

P

P

P

P

P

P

S

S

S

S

T

T

Uv

W

W

W

W

Z

20 S.-K. Chung, S.K. Kim / Respiratory Phy

his problem can be solved with CFD methods, for which DPIVerves as an essential method to verify results. As such, the twoethods complement each other.The DPIV technique can also be extended to include the pharynx,

arynx, trachea and bronchus (Fresconi and Prasad, 2007; Kim andee, 2007; van Ertbruggen et al., 2008).

cknowledgements

This work was supported by the Korea Science and Engineeringoundation (KOSEF) grant funded by the Korea government (MEST)No. R01-2008-000-20196-0).

eferences

drian, R.J., 1991. Particle-imaging techniques for experimental fluid mechanics.Ann. Rev. Fluid. Mech. 23, 261–304.

rroyo, M.P., Greated, C.A., 1991. Stereoscopic particle image velocimetry. Meas. Sci.Technol. 2, 1181–1186.

ailie, N., Hanna, B., Watterson, J., Gallagher, G., 2006. An overview of numericalmodelling of nasal airflow. Rhinology 44, 53–57.

hung, S.K., Son, Y.R., Shin, S.J., Kim, S.K., 2006. Nasal airflow during respiratory cycle.Am. J. Rhinol. 20, 379–384.

hurchill, S.E., Shackelford, L.L., Georgi, J.N., Black, M.T., 2004. Morphological varia-tion and airflow dynamics in the human nose. Am. J. Hum. Biol. 16, 625–638.

ccles, R., 2000. Nasal airflow in health and disease. Acta Otolaryngol. 120, 580–595.lsinga, G.E., Scarano, F., Wieneke, B., van Oudheusden, B.W., 2006. Tomographic

particle image velocimetry. Exp. Fluids 41, 933–947.resconi, F.E., Prasad, A.K., 2007. Secondary velocity fields in the conducting airways

of the human lung. J. Biomech. Eng. 129, 722–732.arcia, G.J., Bailie, N., Martins, D.A., Kimbell, J.S., 2007. Atrophic rhinitis: a CFD study

of air conditioning in the nasal cavity. J. Appl. Physiol. 103, 1082–1092.irardin, M., Bilgen, E., Arbour, P., 1983. Experimental study of velocity fields in

a human nasal fossa by laser anemometry. Ann. Otol. Rhinol. Laryngol. 92,231–236.

rant, O., Bailie, N., Watterson, J., Cole, J., Gallagher, G., Hanna, B., 2004. Numericalmodel of a nasal septal perforation. Medinfo 11, 1352–1356.

rützenmacher, S., Lang, C., Mlynski, G., 2003. The combination of acoustic rhinom-etry, rhinoresistometry and flow simulation in noses before and after turbinatesurgery: a model study. ORL J. Otorhinolaryngol. Relat. Spec. 65, 341–347.

rützenmacher, S., Robinson, D.M., Gräfe, K., Lang, C., Mlynski, G., 2006. First findingsconcerning airflow in noses with septal deviation and compensatory turbinatehypertrophy—a model study. ORL J. Otorhinolaryngol. Relat. Spec. 68, 199–205.

rützenmacher, S., Robinson, D.M., Lang, C., Lebe, E., Knape, U., Mlynski, G., 2005.Investigations of the influence of external nose deformities on nasal airflow. ORLJ. Otorhinolaryngol. Relat. Spec. 67, 154–159.

ahn, I., Scherer, P.W., Mozell, M.M., 1993. Velocity profiles measured for airflowthrough a large-scale model of the human nasal cavity. J. Appl. Physiol. 75,2273–2287.

insch, K.D., 1995. Three-dimentional particle velocimetry. Meas. Sci. Technol. 6,742–753.

insch, K.D., 2002. Holographic particle image velocimetry. Meas. Sci. Technol. 13,R61–72.

opkins, L.M., Kelly, J.T., Wexler, A.S., Prasad, A.K., 2000. Particle image velocimetrymeasurements in complex geometries. Exp. Fluids 29, 91–95.

ornung, D.E., Leopold, D.A., Youngentob, S.L., Sheehe, P.R., Gagne, G.M., Thomas, F.D.,Mozell, M.M., 1987. Airflow patterns in a human nasal model. Arch. Otolaryngol.Head Neck Surg. 113, 169–172.

örschler, I., Brücker, C., Schröder, W., Meinke, M., 2006. Investigation of the impactof the geometry on the nose flow. Eur. J. Mech. B Fluids 25, 471–490.

örschler, I., Meinke, M., Schröder, W., 2003. Numerical simulation of the flow fieldin a model of the nasal cavity. Comput. Fluids 32, 39–45.

uang, H., 1994. Limitations of and Improvements to PIV and its Application to aBackward-facing Step Flow. Verlag Koester.

shikawa, S., Nakayama, T., Watanabe, M., Matsuzawa, T., 2006. Visualization of flowresistance in physiological nasal respiration: analysis of velocity and vorticitiesusing numerical simulation. Arch. Otolaryngol. Head Neck Surg. 132, 1203–1209.

ähler, C.J., Kompenhans, J., 2000. Fundamentals of multiple plane particle imagevelocimetry. Exp. Fluids Suppl., S70–S77.

elly, J.T., Prasad, A.K., Wexler, A.S., 2000. Detailed flow patterns in the nasal cavity.J. Appl. Physiol. 89, 323–337.

eyhani, K., Scherer, P.W., Mozell, M.M., 1995. Numerical simulation of airflow in thehuman nasal cavity. J. Biomech. Eng. 117, 429–441.

im, S.K., Chung, S.K., 2004a. An investigation on airflow in disordered nasal cavityand its corrected models by tomographic PIV. Meas. Sci. Technol. 15, 1090–1096.

im, S.K., Haw, J.R., 2004b. An investigation on airflow in pathological nasal airwayby PIV. J. Vis. 7, 341–348.

Z

Z

& Neurobiology 163 (2008) 111–120

im, S.K., Lee, R.W., 2007. PIV application to the respiratory airflow. In: 6th Inter-national Symposium on Flow Visualization and Image Processing, Hawaii, USA,pp. 138–141.

im, S.K., Son, Y.R., 2002. Particle image velocimetry measurements in nasal airflow.Trans. KSME B 26, 566–569.

urtz, D.B., Zhao, K., Hornung, D.E., Scherer, P., 2004. Experimental and numericaldetermination of odorant solubility in nasal and olfactory mucosa. Chem. Senses29, 763–773.

indemann, J., Brambs, H.J., Keck, T., Wiesmiller, K.M., Rettinger, G., Pless, D., 2005a.Numerical simulation of intranasal airflow after radical sinus surgery. Am. J.Otolaryngol. 26, 175–180.

indemann, J., Keck, T., Wiesmiller, K., Sander, B., Brambs, H.J., Rettinger, G., Pless, D.,2004. A numerical simulation of intranasal air temperature during inspiration.Laryngoscope 114, 1037–1041.

indemann, J., Keck, T., Wiesmiller, K., Sander, B., Brambs, H.J., Rettinger, G., Pless,D., 2006. Nasal air temperature and airflow during respiration in numericalsimulation based on multislice computed tomography scan. Am. J. Rhinol. 20,219–223.

indemann, J., Keck, T., Wiesmiller, K.M., Rettinger, G., Brambs, H.J., Pless, D., 2005b.Numerical simulation of intranasal air flow and temperature after resection ofthe turbinates. Rhinology 43, 24–28.

in, Y.G., Jung, H.W., Kim, C.S., 1995. Prevalence study of nasal septal deformities inKorea: results of a nation-wide survey. Rhinology 33, 61–65.

zlugedik, S., Nakiboglu, G., Sert, C., Elhan, A., Tonuk, E., Akyar, S., Tekdemir, I., 2008.Numerical study of the aerodynamic effects of septoplasty and partial lateralturbinectomy. Laryngoscope 118, 330–334.

ark, K.I., Brücker, C., Limberg, W., 1997. Experimental study of velocity fields ina model of human nasal cavity by DPIV. Laser anemometry advances andapplications. In: Ruck, B., Leder, A., Dopheide, D. (Eds.), Proceedings of the7th International Conference. University of Karlsruhe, Karlsruhe, Germany, pp.617–626.

ereira, F., Gharib, M., Dabiri, D., Modarress, D., 2000. Defocusing digital particleimage velocimetry: a 3-components 3-dimensional DPIV measurement tech-nique, application to bubbly flows. Exp. Fluids Suppl., S78–S84.

ereira, F., Stüer, H., Graff, E.C., Gharib, M., 2006. Two-frame 3D particle tracking.Meas. Sci. Technol. 17, 1680–1692.

less, D., Keck, T., Wiesmiller, K.M., Lamche, R., Aschoff, A.J., Lindemann, J., 2004.Numerical simulation of airflow patterns and air temperature distribution dur-ing inspiration in a nose model with septal perforation. Am. J. Rhinol. 18,357–362.

roctor, D.F., 1966. Airborne disease and the upper respiratory tract. Bacteriol. Rev.30, 498–513.

roetz, A.W., 1951. Air currents in the upper respiratory tract and their clinicalimportance. Ann. Otol. Rhinol. Laryngol. 60, 439–467.

chreck, S., Sullivan, K.J., Ho, C.M., Chang, H.K., 1993. Correlations between flowresistance and geometry in a model of the human nose. J. Appl. Physiol. 75,1767–1775.

hi, H., Kleinstreuer, C., Zhang, Z., 2006. Laminar airflow and nanoparticle or vapordeposition in a human nasal cavity model. J. Biomech. Eng. 128, 697–706.

immen, D., Scherrer, J.L., Moe, K., Heinz, B., 1999. A dynamic and direct visualizationmodel for the study of nasal airflow. Arch. Otolaryngol. Head Neck Surg. 125,1015–1021.

wift, D.L., Proctor, D.F., 1977. Acess of air to the respiratory tract. In: Brain, J.D.,Proctor, D.F., Reid, L.M. (Eds.), Respiratory Defense Mechanisms, Part I. MarcelDekker, Inc, Newyork, pp. 63–93.

arabichi, M., Fanous, N., 1993. Finite element analysis of airflow in the nasal valve.Arch. Otolaryngol. Head Neck Surg. 119, 638–642.

onndorf, J., 1939. Der weg der atemluft in der menschlichen nase. Arch. Ohren.Nasen Kehlkopfheilkd 146, 41–63.

ddströmer, M., 1940. Nasal respiration. Acta Otolaryngol. Suppl. 42, 1–146.an Ertbruggen, C., Corieri, P., Theunissen, R., Riethmuller, M.L., Darquenne, C., 2008.

Validation of CFD predictions of flow in a 3D alveolated bend with experimentaldata. J. Biomech. 41, 399–405.

einhold, I., Mlynski, G., 2004. Numerical simulation of airflow in the human nose.Eur. Arch. Otorhinolaryngol. 261, 452–455.

exler, D., Segal, R., Kimbell, J., 2005. Aerodynamic effects of inferior turbinatereduction: computational fluid dynamics simulation. Arch. Otolaryngol. HeadNeck Surg. 131, 1102–1107.

illert, C.E., Gharib, M., 1991. Digital particle image velocimetry. Exp. Fluids 10,181–193.

illert, C.E., Gharib, M., 1992. Three-dimensional particle imaging with a singlecamera. Exp. Fluids 12, 353–358.

hao, K., Dalton, P., Yang, G.C., Scherer, P.W., 2006a. Numerical modeling of turbulentand laminar airflow and odorant transport during sniffing in the human and ratnose. Chem. Senses 31, 107–118.

hao, K., Pribitkin, E.A., Cowart, B.J., Rosen, D., Scherer, P.W., Dalton, P., 2006b.Numerical modeling of nasal obstruction and endoscopic surgical intervention:outcome to airflow and olfaction. Am. J. Rhinol. 20, 308–316.

hao, K., Scherer, P.W., Hajiloo, S.A., Dalton, P., 2004. Effect of anatomy on humannasal air flow and odorant transport patterns: implications for olfaction. Chem.Senses 29, 365–379.