Eur Arch Otorhinolaryngol (2004) 261 : 452–455
DOI 10.1007/s00405-003-0675-y
452
RHINOLOGY
Ivo Weinhold · Gunter Mlynski
Numerical simulation of airflow in the human nose
Received: 23 January 2003 / Accepted: 12 August 2003 / Published online: 3 December 2003
© Springer-Verlag 2003
Abstract Unobstructed air passageways as well as sufficient contact of the air stream with the mucous membrane
are essential for the correct function of the nose. For that,
local flow phenomena, which often cannot be captured by
standard diagnostic methods, are important. We developed and validated a method for the numerical simulation
of the nasal airflow. Two anatomically correct, transparent
resin models of human nasal cavities, manufactured by a
special casting technology, and the nasal cavities of two
patients were reconstructed as Computer Aided Design
models based on computed tomography (CT) scans. One
of the nasal models and one clinical case represented a
normal nasal anatomy, while the others were examples of
pathological alterations. The velocity and pressure fields
in these reconstructed cavities were calculated for the entire
range of physiological nasal inspiration using commercially available computational fluid dynamics software.
To validate the results rhinoresistometric data were measured and characteristic streamlines were videotaped for the
resin models. The numerical results were in good agreement with the experimental data for the investigated cases.
An example of a complex clinical case demonstrates the
potential benefit of the developed simulation method for
rhinosurgical planning. The results support the assumption that even under the specific conditions of the clinical
practice the application of numerical simulation of nasal
airflow phenomena may become realistic in the near future. However, important technical issues such as a completely automated reconstruction of the nasal cavity still
need to be resolved before such simulations are efficient
and cost effective enough to become a standard tool for
the rhinologist.
Keywords Nasal airflow · Flow simulation ·
Rhinosurgical planning · Nasal cavity · Nasal physiology
I. Weinhold (✉) · G. Mlynski
Department of Otorhinolaryngology and Head and Neck Surgery,
Ernst Moritz Arndt University of Greifswald,
Walther-Rathenau-Strasse 43-45, 17487 Greifswald, Germany
Tel.: +49-3834-866228, Fax: +49-3834-866201
Introduction
In order to ensure the proper function of the human nose,
an unobstructed air passageway as well as sufficient contact of the air stream with the mucous membrane is essential. Under normal conditions the healthy nasal anatomy
forms special aerodynamic elements that guide and condition the air stream in such a way that all respiratory functions can be fulfilled sufficiently [6, 8]. Often anatomical,
pathological or traumatic modifications of nasal functions
are an indication for functional rhinosurgery. The planning
of the therapeutic strategy relies considerably on high
quality diagnostic results. In some difficult cases the reasons for an insufficient respiratory function of the nose
are complex global and local flow and turbulence phenomena. Those phenomena are very difficult to realize using the normal rhinological diagnostic tools available today, such as rhinomanometry, rhinoresistometry or acoustic
rhinometry. A study statistically supports this by the finding that 9 months after a functional septoplasty only 51%
of patients were subjectively free from obstruction [4].
Apart from this, the complex nasal anatomy characterized
by numerous very thin airway channels does not allow direct experimental measurements of flow patterns inside
the nose. The recent developments in medical imaging,
three-dimensional geometrical modeling, numerical mathematics as well as computer science open new possibilities for physically realistic numerical simulations of nasal
airflow based on anatomically precise computer models
of the nasal cavity. Today, Computational Fluid Dynamics
(CFD) methods are widely used in science and industry to
design and optimize products and to simulate natural
processes, including numerous biomedical applications.
To increase the understanding of the detailed flow phenomena inside the human nasal cavity without any intervention and clinical risk for the patient, we applied CFD
methods to simulate nasal airflow. The three-dimensional
visualization of the simulation results allow a detailed picture of the local and global distribution of physical flow
parameters like air velocities and pressure to be obtained
as a basis for specific, validated conclusions for the plan-
453
ning of the rhinosurgical therapy. Recently published results of numerical simulations of nasal airflow [2, 5, 9]
showed limitations with respect to neglecting important
physical conditions or insufficient agreement with experimental data. For instance, the simulations were restricted
to the application of either pure laminar or pure turbulent
flow models, although it is known that in the range of
physiological breathing laminar, turbulent and transitional
flow regimes may exist at the same time [3, 6, 8]. We developed a method for the numerical simulation of nasal airflow and the visualization of the three-dimensional flow
and pressure fields in the human nasal cavity considering
anatomically precise geometrical models and the physically complex laminar, turbulent and transitional characteristics of nasal airflow. We report the results of the validation of the numerical flow simulation and the first clinical application of the method.
Materials and methods
Assurance of the quality of the results should always be an indispensable element of each CFD simulation. The European organization ERCOFTAC developed best practice guidelines to ensure
high quality computational fluid dynamics results [1]. All simulations in this work were carried out applying these suggested rules.
This included additional tasks such as a general validation of the
simulation software used, mesh dependency studies and sensitivity
studies for the location and the type of boundary conditions.
In the context of the clinical relevance the simulation method
should meet the following requirements: (1) the use of medical images from conventional medical imaging devices, (2) the consideration of complex anatomical details and the three-dimensional
laminar, turbulent and transitional characteristics of the nasal airflow, (3) the possibility to efficiently simulate a large number of
physical variations, (4) robust and reliable computations, (5) assurance of quality for the simulation results and (6) vivid visualization of the simulation results.
The method was validated in two steps. First, two transparent
resin models of anatomically correct nasal cavities (one side only,
without sinuses) were investigated by both fluid dynamics experiments and numerical simulations of airflow. These models were
manufactured using an established in vivo casting technology [7].
One model (27-year-old male) represented a normal nasal anatomy,
while the second model (29-year-old male) was characterized by a
septal deviation. As the second step, the experiences obtained with
the nasal models were used to investigate the nasal airflow of two
patients. One patient (37-year-old male) had a normal nasal anatomy,
and the other (26-year-old male), after having undergone various
rhinosurgical operations, had a final total inferior turbinectomy,
but still complained about insufficient breathing and sicca symptoms.
Rhinoresistometric data of flow rate versus pressure drop over
the physiological range of breathing were measured for each of the
four cases. These curves served as quantitative reference data for
the validation of the computed results. Additionally, flow experiments using the resin models were carried out applying an established experimental methodology [7]. As a result, characteristic
streamlines of the flow inside the models were recorded on video
tape for selected flow rates. These video images were the reference
for qualitative validations.
Geometrical modeling
The geometries of the four nasal cavities were reconstructed as 3D
CAD models (Computer Aided Design) using a commercially
available CAD software (SolidWorks 2001, SolidWorks Corpora-
tion, Concord, USA). The basis for the reconstruction was for all
cases a stack of 1 mm spaced, high resolution coronary computed
tomography (CT) scans. In order to enforce a nearly constant geometry of the patient’s nasal cavities by suppressing vasomotor changes
during the imaging and measurement period, a decongestant spray
was applied prior to imaging.
The contours of the nasal cavities (one side only) were then
manually digitalized for each slice and lofted to form a solid model
using standard functionality of the software. During digitalization
of the patient’s cavities, the sinuses and the vibrissae were neglected.
Physical modeling
The following conditions were assumed for physical modeling:
(1) pure airflow (dry, steady-state, incompressible, isothermal),
(2) inspiration flow direction only, (3) no gravitational effects,
constant geometry with hydraulic smooth walls (no vasomotor
changes) and (4) no change of the nasal mucous membranes.
The inflow conditions were located for all models on a planar
face at the ostium externum. The inlet flow was defined as a constant volume flow rate assuming a very low turbulence level at this
location. Depending on the model, volume flow rates between
20 ml/s and 1,400 ml/s were simulated.
The outflow condition of a constant pressure of 1 bar was applied at a location corresponding to the outlet opening of the nasal
models, and at a location corresponding to the Choane for the patient’s nasal cavities, respectively.
Computations
For each of the four models a numerical rhinoresistometric curve
was computed, represented by a total of 8 to 17 points per curve.
All computations were done using the commercial CFD software
package FloWorks 2001 (NIKA GmbH, Frankfurt am Main).
FloWorks 2001 employs a Finite Volume scheme to solve the
Navier-Stokes equations. This software qualifies for the simulation
of nasal airflow, because it offers fully automatic mesh generation
of complex geometries with local adaptive refinement, automatic
treatment of boundary layers and an automatic handling of laminar, turbulent and transitional flow regimes within one model,
based on an enhanced two-equation turbulence model of the k-ε
type.
Results
Figure 1 exemplarily shows the computed curve of the
nasal model representing a normal anatomy, visualized together with the corresponding measured data and their estimated tolerance. Similar results were obtained for the
nasal model with septal deviation. The diagram in Fig. 1
shows that the results of the computer simulation are
within the estimated tolerance of the measured data. Additionally, the simulation results for the two reconstructed
nasal models were qualitatively compared with the videotaped characteristic streamlines at flow rates of 50 ml/s,
100 ml/s, 150 ml/s, 200 ml/s, 300 ml/s and 400 ml/s. Figure 2a and b shows an example of the comparison of
recorded streamlines at a volume flow rate of 200 ml/s with
the corresponding computed flow trajectories for the nasal
model with septal deviation. A good correspondence of
the simulations with the experiments was found.
The qualitative evaluation of the simulated flow pattern in the nasal cavity of the patient with normal anatomy
454
Fig. 1 Results of the resin
model representing normal nasal
anatomy
Fig. 3 Computed flow trajectories of the patient model with pathological nasal anatomy at 250 ml/s
Fig. 2 a Video image of the resin model representing a septum deviation at 200 ml/s. b Computed flow trajectories of the resin model
representing a septum deviation at 200 ml/s
confirmed the understanding of a normally ventilated
nose. The air is in touch with almost the entire surface of
the turbinates. This patient’s nose functions correctly, and
he does not feel any impairment of his nasal breathing.
The simulation results for the pathological clinical case
after a total inferior turbinectomy disclosed possible reasons for the continuous complaints of the patient about
obstructed nasal breathing. The rhinoresistometric measurements indicated a medium nasal obstruction (resistance at 250 ml/s =0.31 Pa s/ml) and increased turbulent
effects (predominantly turbulent flow already at a inspiratory flow rate of about 300 ml/s, drag coefficient
λ=0.037). The visualized airflow pattern disclosed an increased “nozzle effect” of the nasal vestibule. This effect
resulted in a jet type flow along the bottom surface of the
medium turbinate and quite large recirculation zones in
the lower and upper regions of the nasal cavity (Fig. 3).
The total pressure drop in the internal ostium appeared to
be approximately two times higher than in a normal nasal
cavity.
Discussion
Currently, by far the most time consuming and errorprone step of the developed method is the manual geo-
455
metrical reconstruction of the nasal cavity from coronary
CT scans. There is a great variety of possible causes for
uncertainties, such as the spatial resolution and quality of
the medical images, difficulties in precisely determining
the boundary between the flow domain and the mucous
membrane, or the transient nature of the interior nasal
anatomy (nasal cycle). In order to keep in vivo rhinoresistometric measurement data comparable with the simulation results, it is necessary to do the measurements immediately after getting the CT scans. Also, it is suggested always to employ the maximum possible resolution and image size for the CT scans. Perhaps, an additional geometrical calibration process can help to determine and verify
the reconstruction parameters best fitting the reality.
Comparisons of the computed numerical rhinomanometric curves for the four nasal cavities with the corresponding measurements showed good agreement of the
data over the entire range of physiological breathing. This
indicates that the modeling assumptions were basically
admissible, and the physical models, especially the turbulence modeling schemes of FloWorks 2001, appear to be
able to simulate sufficiently the main flow phenomena in
the nasal cavity. In addition, when qualitatively comparing simulation results with video images, the simulations
reproduced the primary global and local flow characteristics quite well. However, a significantly larger number of
cases need to be simulated and quantitatively compared
with measurements to obtain the necessary amount of data
for a statistical evaluation of the accuracy level. To make
this feasible, the availability of a fully automated geometrical reconstruction method of the nasal cavity would be
absolutely essential. Again, this should be the primary
task for the further development of the method.
For a better understanding of local flow phenomena as
a basis of functional rhinosurgical planning we demonstrated the application of nasal airflow simulations in a
complex clinical case. Based on existing experiences, we
would have recommended a surgical narrowing of the
lower nasal cavum for the patient with a condition after
turbinectomy. After numerical flow simulation, the conclusion for further rhinosurgical planning was that additionally to the surgical narrowing of the lower nasal cavum,
an extension of the internal ostium was recommended to
decrease the jet effect and to eliminate the large recirculation zones. This should ensure a lower pressure drop, a
better distribution of the air flow and a normal turbulence
situation in the entire nasal cavity as a prerequisite for a
normal respiratory function of the nose.
Consequently, the method developed for the simulation
of nasal airflow can be considered as being validated in general and available for further applications. It provides the basis to potentially carry out a large number of simulations for
further statistical validations, and for systematic investigations to better understand the basic principles of the interaction between shape and function of the human nasal cavity.
It may be integrated in a future computer-based rhinosurgical planning system supporting the specialist’s therapeutic
decisions. But before those simulations actually will become
a diagnostic tool for a day-to-day clinical application, the
cost-benefit ratio needs to be improved significantly. However, for selected cases such as expert opinions in lawsuits,
the benefit may justify the current costs of the application of
the simulation method even today.
References
1. Casey M, Wintergerste T (2000) Best practice guidelines. SIG
on quality and trust in industrial CFD. European Research Community On Flow, Turbulence And Combustion, Hampshire
2. Chometon F, Ebbo D, Gillieron P, Koifman P, Lecomte F, Sorrel-Dejerine N (2000) A numerical simulation of the aerodynamics of the nasal cavity. Annales d’ Otolaryngologie et de
Chirurgie Cervico Faciale 117:98–104
3. Cole P (2000) Biophysics of nasal airflow: a review. Am J Rhinol 14:245–250
4. Jessen M, Ivarsson A, Malm L (1989) Nasal airway resistance
and symptoms after functional septoplasty: comparison of findings at 9 months and 9 years. Clin Otolaryngol Allied Sci 14:
231–234
5. Keyhani K, Scherer PW, Mozell MM (1995) Numerical simulation of airflow in the human nasal cavity. J Biomech Eng 117:
429–441
6. Mlynski G, Gruetzenmacher S, Lang C, Mlynski B (2000) Aerodynamik der Nase. Physiologie und Pathophysiologie. HNO Praxis
Heute 20:61–81
7. Mlynski G, Gruetzenmacher S, Plontke S, Grützmacher W,
Mlynski B, Lang C (2000) A method for studying nasal air flow
by means of fluid dynamics experiments. Z Medizin Physik 10:
207–214
8. Mlynski G, Gruetzenmacher S, Plontke S, Mlynski B, Lang C
(2001) Correlation of nasal morphology and respiratory function. Rhinology 39:197–201
9. Subramaniam RP, Richardson RB, Morgan KT, Kimbell JS,
Guilmette RA (1998) Computational fluid dynamics simulations
of inspiratory airflow in the human nose and nasopharynx. Inhalat Toxicol 10:473–502