You are on page 1of 16

INSTITUTE OF PHYSICS PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 49 (2004) 371–386 PII: S0031-9155(04)71699-1
Acoustic rhinometry measurements in stepped-tube
models of the nasal cavity
H¨ useyin C¸ elik
1
, Mehmet Cankurtaran
1
and Ozcan Cakmak
2
1
Department of Physics, Faculty of Engineering, Hacettepe University, Beytepe, 06532 Ankara,
Turkey
2
Department of Otorhinolaryngology, Faculty of Medicine, Baskent University, 06490 Ankara,
Turkey
E-mail: cankur@hacettepe.edu.tr
Received 10 November 2003
Published 16 January 2004
Online at stacks.iop.org/PMB/49/371 (DOI: 10.1088/0031-9155/49/3/002)
Abstract
Stepped-tube models with a constriction in the anterior section were used
to evaluate the effects that nasal valve passage area and nasal cavity shape
have on acoustic rhinometry (AR) measurements. The AR-determined
cross-sectional areas beyond a constriction of small passage area were
consistently underestimated, and the corresponding area–distance curves
showed pronounced oscillations. Also, the AR technique did not accurately
reproduce abrupt changes in passage area. The results suggest that, regardless
of the particular shape of the nasal cavity model, AR does not provide reliable
information about cross-sectional areas posterior to a severe constriction. The
experimental results are discussed in terms of theoretically calculated acoustic
input impedance for the models studied, the physical limitations of AR, and
assumptions made in ARalgorithms. The study demonstrated that energy losses
and sound wave attenuation due to air viscosity do not significantly affect AR
measurements. It was also shown that passage area beyond a severe constriction
is underestimated because the barrier created by the constriction reflects most
of the incident sound power. The results also indicate that the oscillations in
area–distance curves are due to low-frequency acoustic resonances in the nasal
cavity model.
1. Introduction
Acoustic rhinometry (AR) was introduced by Hilberg et al (1989) as an objective method
for examining the nasal cavity. This technique is based on the principle that a sound pulse
propagating in the nasal cavity is reflected by local changes in acoustic impedance. However,
certain factors inherent to the physics and algorithms used in AR limit the accuracy of this
method. One potential problemin using ARto study the geometry of the nasal cavity is that the
0031-9155/04/030371+16$30.00 © 2004 IOP Publishing Ltd Printed in the UK 371
372 H C¸ elik et al
area of a region beyond a severe constriction may not be measured accurately (Cakmak et al
2001, Cankurtaran et al 2003, Hilberg et al 1989, 1998, Hilberg and Pedersen 2000). This
means that, when considering AR accuracy in relation to the complex anatomy of the nasal
passage, special attention must be paid to the influence of the anterior narrowsegment, the nasal
valve. Various researchers have suggested that the nasal valve may cause loss of energy from
the incident sound wave, which would lead to underestimation of AR-measured area beyond
the narrowed site (Hamilton et al 1995, Hilberg et al 1989, 1998, Hilberg and Pedersen 2000,
Terheyden et al 2000). However, the specific physical cause of the area underestimation in
this portion of the nasal passage was clarified only recently by Cankurtaran et al (2003). These
authors examined the effects of nasal valve passage area on accuracy of AR measurements
using simple pipe models with a constriction. They demonstrated that passage area beyond
the constricted site is underestimated because the barrier created by the constriction reflects
most of the incident sound power. Specifically, they found that, when the passage area at the
site of constriction was smaller than normal adult size, the AR-measured cross-sectional areas
beyond the constriction were consistently underestimated and the corresponding area–distance
curve showed pronounced oscillations.
This previous work suggests that individual anatomical variations of the anterior narrow
segment might significantly limit the role of AR as a diagnostic tool for the entire nasal cavity.
If these limitations are not kept in mind during clinical examinations, AR results could lead
to misinterpretation of a patient’s condition. However, further studies on more complicated
models for the nasal cavity, supported by theoretical considerations, are necessary in order to
better understand how nasal valve passage area and the complex geometry of the nasal cavity
influence AR-measured area.
The aim of this study was to investigate how nasal valve passage area and nasal cavity
shape affect the accuracy of area–distance curves derived using commercially available AR
equipment. We also evaluated the ability of the AR method to accurately reproduce a sudden
(abrupt) change in passage area. To carry out the investigation, we used stepped-tube models
with a constriction in the anterior section. We calculated the acoustic input impedance of
the stepped-tube models used for the study, because it is possible to estimate both the cross-
sectional area as a function of distance and the acoustic input impedance fromthe input impulse
response of the system (Fredberg et al 1980, Hilberg et al 1989, Hoffstein and Fredberg 1991,
Jackson et al 1977, Jackson and Olson 1980). We compared the cross-sectional areas derived
by AR with the ‘effective’ cross-sectional areas based on the acoustic input impedance of the
stepped-tube models. We accounted for the effects of air viscosity on AR-derived area by
calculating the acoustic impedance for the propagation of planar sound waves with complex
wave vector.
2. Materials and methods
We used a transient signal acoustic rhinometer to perform the acoustic measurements. The
processed bandwidth for this rhinometer ranged from 100 Hz to 10 kHz. A 10 kHz low-pass
filter was used to reduce the errors associated with cross modes in the actual nasal cavity.
To assess the influence of the nasal cavity shape on AR-derived area–distance curve, and
to evaluate the ability of AR to accurately reproduce sudden changes in nasal passage area,
we used stepped-tube models made of brass, each consisting of three cylindrical sections
(figure 1(a)). The anterior section was 7.0 cm long with 1.3 cm inner diameter, the second
section was 5.0 cm long with 1.9 cm inner diameter and the third section was either 5.0 cm
or 2.5 cm long with 2.8 cm inner diameter. The sections were connected such that there was
a sudden change in passage area at each joint. To assess the effects that a constriction in the
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 373
(a)
(b)
7.0 cm 5.0 cm L= 5.0 (or 2.5) cm
1.3 cm
1.9 cm
2.8 cm
l
2.0 cm
1.3 cm
5.0 cm L= 5.0 (or 2.5) cm
d
1.9 cm 2.8 cm
Figure 1. The shape and sizes of the stepped-tube models (a) with no insert and (b) with an insert
(length l = 1.5 cm and inner diameter d) in the anterior section. The independent variables were
the inner diameter of the insert and the length (L) of the final (most distal) section of the nasal cavity
model. The anterior end of the model was connected to the nosepiece of the acoustic rhinometer.
anterior portion of the nasal cavity model had on the area estimates beyond the constriction,
a cylindrical insert (nasal valve simulator) of length l = 1.5 cm and inner diameter d was
fitted into the first (anterior) section of the stepped-tube model. In this way, we produced
five-stage stepped-tube models (figure 1(b)). Each insert was placed 2.0 cm from the
anterior end of the first section, and the inner diameters of the inserts tested ranged from
d = 0.4 to 1.0 cm in 0.1 cm increments. The passage areas of the inserts used matched those
of the actual human nasal valve, in order to imitate the normal anatomy and pathologies of
the nasal valve region (Cakmak et al 2003b, Hilberg and Pedersen 2000, Tomkinson and
Eccles 1998). These diameters also matched those tested in pipe models by previous
investigators (Buenting et al 1994, Cakmak et al 2001, Cankurtaran et al 2003). All the
dimensions of the five-stage stepped-tube model fitted with an insert of 0.7 cm inner diameter
were approximately the same as the model used by Hilberg and Pedersen (2000).
In order to prevent acoustic leakage, the stepped-tube models were designed such that
there was secure contact between the model and the nosepiece of the acoustic rhinometer.
All AR measurements were repeated at least five times to ensure that the results were
reproducible.
374 H C¸ elik et al
-4 -2 0 2 4 6 8 10 12 14 16
1
10
(a)
(b)
A
r
e
a
(
c
m
2
)
Distance (cm)
model
AR measurements
-4 -2 0 2 4 6 8 10 12 14
1
10
A
r
e
a
(
c
m
2
)
Distance (cm)
model
AR measurements
Figure 2. AR-derived area–distance curves for the three-stage stepped-tube models with final
section of length L = 5.0 cm (a) and L = 2.5 cm (b). The lines connecting the experimental data
points are to guide the eye. The solid line represents the actual cross-sectional areas of the model.
3. Experimental results
Stepped-tube models without inserts (figure 1(a)) were used to assess the influence of nasal
cavity shape on AR-measured area, and to evaluate the ability of AR to accurately reflect
sudden changes in nasal passage area. The passage area of the narrowest (first) section of
this model was large (1.32 cm
2
), so there were no sites of constriction to affect AR in this
case. Figures 2(a) and (b) illustrate experimental area–distance curves for the three-stage
stepped-tube models with final section of length L = 5.0 cm and 2.5 cm, respectively. The
actual cross-sectional areas of the models are shown on the diagrams for comparison. The data
in figure 2 show that the experimental area–distance curve does not accurately depict the sharp
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 375
increase in passage area at each junction between cylindrical sections: the AR method shows
a more gradual change than actually occurs. The AR-measured area–distance curve starts to
rise approximately 0.5 cm before the step (junction), increases smoothly and then levels off
approximately 0.7 cm after the step. The distance over which the AR-derived curve reflected
the change in area at each step (approximately 1.2 cm) was almost completely independent
of the magnitude of the acute increase in area at each junction. Previous experiments with
acoustic pulse reflectometry revealed similar trends for single-step models with rigid walls
(Fredberg et al 1980, Hilberg et al 1998, Jackson et al 1977, Sondhi and Gopinath 1971,
Sondhi and Resnick 1983). These results indicated more gradual changes in passage area
than actually existed, and the authors concluded that acoustic pulse reflectometry does not
accurately reproduce a sudden change in passage area of a tubular object.
For our stepped-tube model with final-section length 5.0 cm, apart from the above-
mentioned findings at each step and oscillations in the measured areas for each cylindrical
section, there were no significant discrepancies between AR-derived area and actual area
(figure 2(a)). In contrast, towards the end of the stepped-tube model with final-section length
2.5 cm, the AR-derived area deviated significantly from actual area (figure 2(b)). Specifically,
the cross-sectional areas of the second and third sections of the model were underestimated
and the corresponding AR-derived area–distance curve showed oscillations. These results
suggest that the length of the final section of the stepped-tube model imposes an important
boundary condition that affects the accuracy of AR measurements.
Ever since acoustic pulse response analysis was first described (Sondhi and Gopinath
1971), it has been of special interest to investigate how narrowing (constriction) of the
anterior portion of the acoustic pathway influences area estimates at locations beyond
such a constriction (Brooks et al 1984, Fredberg et al 1980, Hilberg et al 1989, 1998,
Hilberg and Pedersen 2000, Jackson et al 1977, Jackson and Olson 1980, Sondhi and Resnick
1983). As detailed above, we also recorded AR measurements in our stepped-tube models
after narrowing the passage area with different-sized inserts placed in the anterior section.
Figures 3(a) and (b) display typical examples of the variation we observed in AR-derived
area–distance curves relative to the inner diameter (d) of the insert for models with final-
section length 5.0 cm and 2.5 cm, respectively. For clarity, only the data sets obtained for three
different insert diameters are presented in these graphs. The actual cross-sectional areas of the
models are also shown for comparison. As figure 3 shows, the AR-measured cross-sectional
areas anterior to the insert were similar and almost completely independent of insert diameter.
The results suggest that AR provides a reasonably accurate measure of cross-sectional area
from the anterior opening of the model to approximately 1.0 cm before the insert (nasal
valve simulator). The measured passage area of the insert, which corresponds to the deepest
minimum in the corresponding area–distance curve, was consistently overestimated for inserts
with inner diameter smaller than 0.6 cm. The degree of overestimation was less when the
inner diameter of the insert was greater than 0.6 cm.
For the stepped-tube models with inserts of inner diameter d >0.6 cm, the cross-sectional
areas of the uniformsections (i.e. away fromthe site of sudden change in passage area, the step)
were estimated with considerable accuracy (figure 3). However, with inserts of inner diameter
d < 0.6 cm (passage area <0.283 cm
2
), the AR method consistently underestimated the cross-
sectional area of regions beyond the constriction. The data presented in figure 3 illustrate
that, regardless of the shape of the nasal cavity model, a severe anterior constriction leads
to underestimation of passage area in all sections of the model distal to this narrowing. The
degree of area underestimation increased as the passage area of the constriction decreased. In
addition, regardless of the passage area of the insert, the AR measurements did not accurately
reflect the passage areas in the vicinity of each step.
376 H C¸ elik et al
-4 -2 0 2 4 6 8 10 12 14 16
0.1
1
10
(a)
(b)
A
r
e
a
(
c
m
2
)
Distance (cm)
d = 1.0 cm
d = 0.7 cm
d = 0.4 cm
Model
-4 -2 0 2 4 6 8 10 12 14
0.1
1
10
A
r
e
a
(
c
m
2
)
Distance (cm)
d = 1.0 cm
d = 0.7 cm
d = 0.4 cm
Model
Figure 3. The effects of inner diameter (d) of the insert on the experimental area–distance curves
for stepped-tube models with final section of length L = 5.0 cm (a) and L = 2.5 cm (b). The
different symbols refer to data sets obtained for models with inserts of different diameter, as shown
inside the graph. The lines connecting the experimental data points are to guide the eye. The solid
lines represent the actual cross-sectional areas of the models.
One feature that is common to all the experimental data sets presented in figures 2
and 3 is oscillation of the AR-measured areas in all cylindrical sections of the stepped-
tube models. These oscillations were even apparent for models without inserts (figure 2),
and for models with inserts that provide the largest passage area (inner diameter d = 1.0 cm;
passage area 0.785 cm
2
) (figure 3). The oscillation amplitude tended to increase as the
inner diameter of the insert decreased from 1.0 to 0.4 cm. Previous reports have noted
similar oscillations in area–distance curves for straight-tube models with an insert (Buenting
et al 1994, Cakmak et al 2001, 2003a, Cankurtaran et al 2003, Hamilton et al 1995,
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 377
Hilberg et al 1998, Louis et al 2001, Mlynski et al 2003), and for a stepped-tube model
(Hilberg and Pedersen 2000).
4. Theory and numerical results
Cross-sectional area as a function of distance in a tubular airway, and the acoustic input
impedance of such a structure are closely related, and both can be calculated from
measurements of the input impulse response of the airway (Fredberg et al 1980, Hilberg et al
1989, Jackson et al 1977, Marshall et al 1991, Sharp 1996, Sondhi and Gopinath 1971). In
this section, we derive an expression for the acoustic input impedance of a nasal cavity model
of known shape (i.e. stepped-tube model), and compare cross-sectional area measured by AR
with cross-sectional area determined according to theoretically calculated acoustic impedance.
Using the well-known theoretical expression for the acoustic input impedance of a
cylindrical pipe of finite length (Hall 1987, Kinsler and Frey 1962), and assuming plane wave
propagation but including viscosity losses, Sharp (1996) derived an expression for the acoustic
input impedance of a three-stage stepped-tube model. We expanded this method described
by Sharp to calculate the acoustic input impedance of a stepped-tube model consisting of
n cylindrical sections with differing lengths and radii. We showed that the acoustic input
impedance (Z
i
) of the ith section, and (Z
n
) of the i = nth (final or most distal) section of the
model are given by
Z
i
=
ρω
K
i
πr
2
i
Z
i+1
πK
i
r
2
i
ρω
+ j tan K
i
L
i
1 + jZ
i+1
K
i
πr
2
i
tan K
i
L
i
ρω
for i = n −1, n −2, . . . , 1 (1)
and
Z
n
=
ρω
K
n
πr
2
n
0.25K
2
n
r
2
n
+ j (0.6K
n
r
n
+ tan K
n
L
n
)
(1 −0.6K
n
r
n
tan K
n
L
n
) + j0.25K
2
n
r
2
n
tan K
n
r
n
(2)
respectively. Here ρ is the air density, ω = 2πf is the angular frequency, L
i
and r
i
are the
length and inner radius of ith cylindrical section, j = (−1)
1/2
, and K
i
is the complex wave
vector (wave propagation constant). K
i
is defined as follows (Sharp 1996):
K
i
= k −jα
i
=
ω
c
−j
1
r
i
c

ηω

+ (γ −1)

κω
2ρC
p

for i = 1, 2, . . . , n (3)
where k and α
i
are the real and imaginary components of the complex wave vector, c is the
speed of sound in air, γ is the ratio of the principal specific heats of air, C
p
is the specific heat
of air at constant pressure, η is the coefficient of shear viscosity of air and κ is the thermal
conductivity of air.
The above equations allow us to quantitatively assess the effects that energy losses due
to air viscosity have on accuracy of AR measurements by comparing the impedance curves
calculated for the propagation of planar acoustic waves with real and complex wave vectors.
We calculated the acoustic input impedance of the stepped-tube model as a function of
sound frequency, for selected inner diameter d values of the insert in the anterior section, by
taking the wave vector complex. The values for ρ, c, γ, C
p
, η, and κ at room temperature
(20

C) that were used in the calculations are taken from the literature (Putland 1996,
Sharp 1996). Figure 4 shows plots of the magnitude of acoustic input impedance versus
frequency, which is commonly known as the impedance curve of a system (Jackson
et al 1977, Sharp 1996). For clarity, only the impedance curves for three different insert
diameters are presented. As the figure shows, the impedance curves pass through deep
378 H C¸ elik et al
0 2000 4000 6000 8000 10000
10
3
10
4
10
5
10
6
10
7
10
8
10
9
0.2 0.4 0.6 0.8 1.0 1.2
3.0x10
6
3.5x10
6
4.0x10
6
4.5x10
6
M
e
a
n
i
m
p
e
d
a
n
c
e
(
o
h
m
)
Insert diameter (cm)
I
n
p
u
t
i
m
p
e
d
a
n
c
e
(
o
h
m
)
Frequency (Hz)
d = 1.0 cm
d = 0.7 cm
d = 0.4 cm
Figure 4. The impedance curve for the stepped-tube model with final-section length L = 5.0 cm
and insert diameter d. The plots correspond to three different insert diameters, as shown inside the
graph. The inset shows the frequency-average of acoustic input impedance as a function of insert
diameter.
minima at about 980, 1900, 2940, 4140, 4600, 5440, 6920, 7620 and 8780 Hz, which
roughly correspond to the resonant frequencies of our stepped-tube models. Based on this,
sharp changes (oscillations) would also be expected on plots of sound-power reflection
coefficient versus frequency, because the reflection coefficient is related to the acoustic
input impedance of the model system (Cankurtaran et al 2003, Kinsler and Frey 1962,
Sidell and Fredberg 1978). As a consequence, AR-derived cross-sectional areas would also
be expected to oscillate, since AR measures the intensity of reflected sound waves relative
to that of incident waves. Recently, Cankurtaran et al (2003) proved that the oscillations in
AR-derived area–distance curves are mainly due to the low-frequency acoustic resonances in
the nasal cavity model.
To more clearly demonstrate the correlation between oscillations in the experimental
area–distance curve and those of the impedance curve, we calculated the frequency-average
of the acoustic input impedance of the stepped-tube model, defined by Cakmak et al (2001)
and Cankurtaran et al (2003):
Z
i,m
(x) =
1
9900

10 000
100
Z
i
(x, f ) df. (4)
Here x is the axial distance from the beginning to the end of each cylindrical section of the
model. Figure 5 compares the calculated Z
i,m
(x) curves and the impedance Z
i
= ρc/S
i
for the cylindrical sections of the stepped-tube model with an insert of inner diameter d =
1.0 cm (passage area 0.785 cm
2
). Although there is no change in the cross-sectional area (S
i
)
of each cylindrical section of the model, the calculated mean impedance Z
i,m
(x) oscillates
as a function of distance about Z
i
= ρc/S
i
.
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 379
0 2 4 6 8 10 12 14 16
10
6
10
7
M
e
a
n
i
m
p
e
d
a
n
c
e
(
O
h
m
)
Distance (cm)
Figure 5. Comparison of the frequency-average of acoustic input impedance (solid circles) and
the impedance Z
i
= ρc/S
i
(solid lines) calculated for the stepped-tube model with final-section
length L = 5.0 cm and insert diameter d = 1.0 cm. Here S
i
is the actual cross-sectional area of
each cylindrical section of the model.
The theoretically calculated mean impedance Z
i,m
(x) allows us to determine an ‘effective’
cross-sectional area S
eff
i
for each cylindrical section of the stepped-tube model using the
relationship:
S
eff
i
(x) =
ρc
Z
i,m
(x)
. (5)
The ‘effective’ cross-sectional areas also showoscillations similar to those in the area–distance
curve measured by AR (figure 6). Apart from the apparent phase shift, there is one-to-one
correspondence between the oscillations in the AR-determined area–distance curve and those
in the ‘effective’ cross-sectional area curve. These results demonstrate that the AR algorithm
translates oscillations of acoustic impedance in each cylindrical section of the stepped-tube
model into oscillations in the corresponding area–distance curve. Since the length of the
stepped-tube model is finite, some of the incident sound power is reflected back from its
open distal end. Furthermore, the sound waves transmitted through the anterior constriction
undergo multiple reflections at all locations where the acoustic impedance changes, including
the impedance change at each step, where there is a sudden change in passage area. Hence,
superposition of the sound waves travelling in opposite directions generates a complicated
pattern of standing waves in the stepped-tube model. The resonant frequencies of the stepped-
tube models used in this study exceed the lowest frequency (100 Hz) of the acoustic rhinometer
and extend well into its frequency bandwidth. Therefore, the oscillations of acoustic impedance
in each cylindrical section of the stepped-tube model (and hence those in the corresponding
area–distance curve measured by AR) are due to these low-frequency acoustic resonances in
the model.
The mean input impedance of the stepped-tube model increases substantially as the inner
diameter of the insert (anterior constriction) decreases from1.2 to 0.4 cm(see inset in figure 4).
Therefore, the area underestimation beyond a constriction of small passage area is explained
by the large values found for the mean input impedance of the model.
380 H C¸ elik et al
0 2 4 6 8 10 12 14 16
2
4
6
8
A
r
e
a
(
c
m
2
)
Distance (cm)
Area calculated from impedance
Stepped-tube model
AR-measured area
Figure 6. Comparison of the AR-determined area–distance curve (solid circles) and the ‘effective’
cross-sectional area (open circles) calculated from the frequency-average of acoustic input
impedance. Both the experimental and calculated data correspond to the stepped-tube model
with final-section length L = 5.0 cm and insert diameter d = 1.0 cm. The solid line represents the
actual cross-sectional areas of the model.
Finally, to assess the influence of viscous losses on the accuracy of AR measurements,
we calculated the acoustic input impedance for the stepped-tube model with an insert of the
smallest passage area (inner diameter d = 0.4 cm, passage area 0.126 cm
2
) by taking real and
complex wave vectors. Figure 7 compares the impedance curve calculated by assuming that
there is viscous loss (i.e. K
i
complex) with the impedance curve calculated on the assumption
of zero viscous loss (i.e. K
i
real). The inset shows the variation in mean input impedance
as a function of inner diameter of the insert. The differences between the two impedance
curves obtained for the complex and real wave vectors were negligible. These results indicate
that energy losses due to air viscosity did not significantly influence the accuracy of AR
measurements in any of the stepped-tube models used in this study.
One must also consider the attenuation of sound waves due to air viscosity (Jackson et al
1977, Sondhi and Resnick 1983). However, we found that the sound wave attenuation
coefficient, which is taken to be equal to the imaginary component of the complex wave
vector (see equation (3)), was very small at all frequencies in the range from 100 Hz to
10 kHz. Therefore, the attenuation of sound waves as they propagate along the stepped-tube
model is not significant. For example, if the amplitude of planar sound waves of frequency
f = 5 kHz is equal to A
0
at the input of the stepped-tube model with an insert of diameter
d = 0.4 cm, the amplitude decreases to A
0
exp(−αx) = 0.91A
0
at distance x = 10 cm inside
the model. Such a small reduction in the amplitude of sound waves would not explain the
substantial area underestimation observed in our models with inserts of the smallest passage
area (see figure 3). The stepped-tube models used in this study were short enough for energy
losses due to the attenuation of sound waves to be significant. It is evident from equation (3)
that the attenuation coefficient is inversely proportional to insert diameter. Therefore, area
underestimation that occurs with AR beyond a severe constriction cannot be solely attributed
to energy losses and/or sound wave attenuation due to air viscosity.
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 381
0 2000 4000 6000 8000 10000
10
3
10
4
10
5
10
6
10
7
10
8
10
9
0.2 0.4 0.6 0.8 1.0 1.2
3.0x10
6
3.5x10
6
4.0x10
6
4.5x10
6
M
e
a
n

i
m
p
e
d
a
n
c
e

(
o
h
m
)
Insert diameter (cm)
I
n
p
u
t
i
m
p
e
d
a
n
c
e
(
o
h
m
)
Frequency (Hz)
K complex
K real
Figure 7. Comparison of the impedance curves for the stepped-tube model with final-section
length L =5.0 cm and an insert of the smallest passage area (passage area 0.126 cm
2
, diameter d =
0.4 cm) calculated by taking the wave vector K
i
complex (solid line) and real (dashed line). The
inset shows the frequency-average of acoustic input impedance as a function of insert diameter, as
calculated for the complex and real wave vectors.
5. Discussion
To understand the causes of area underestimation with AR, oscillations in AR-derived
area–distance curves and misrepresentation of sudden area changes, it is necessary to review
the physical elements of the technique and the assumptions made in AR algorithms. The
physical principle of acoustic pulse reflectometry is that sound waves propagating in a tube
are reflected by local changes in acoustic impedance that result from changes in the cross-
sectional area of the tube (Fredberg et al 1980, Hilberg et al 1989, 1998, Hilberg and Pedersen
2000, Hoffstein and Fredberg 1991, Jackson et al 1977). The sound waves undergo partial
reflection and partial transmission at each change in acoustic impedance along the tube,
creating a reflection sequence. The reflection sequence at the input to the tube is termed the
input impulse response of the tube, and cross-sectional area as a function of axial distance
and acoustic input impedance can be calculated from this (Fredberg et al 1980, Jackson et al
1977, Marshall et al 1991, Sondhi and Gopinath 1971). Experimental data for input impulse
response are usually converted to an area–distance function using the algorithm developed by
Ware and Aki (1968).
The assumptions involved in acoustic pulse reflectometry technique are negligible sound
loss, rigid airway walls, symmetrical branching and planar wave propagation (Fredberg et al
1980, Hilberg et al 1989, 1998, Hilberg and Pedersen 2000, Hoffstein and Fredberg 1991,
Jackson et al 1977, Jackon and Olson 1980, Sondhi and Gopinath 1971, Sondhi and Resnick
1983). The assumption of planar wave propagation is fundamental to passage area
measurements made with AR. This assumption determines the spatial resolution and the
frequency bandwidth of the method, and imposes limitations on the transverse sizes of
382 H C¸ elik et al
the airway model (Djupesland and Lyholm 1998, Fredberg et al 1980, Hilberg et al 1989,
Hoffstein and Fredberg 1991, Jackson et al 1977). The departures from planar wave
propagation in our stepped-tube models were not significant, because the inner diameters
of all cylindrical sections are smaller than the shortest sound wavelength (approximately
3.43 cm) produced by the AR equipment. Furthermore, our stepped-tube models were made
of metal (brass), with rigid walls and no branching; thus, there were no effects related to wall
non-rigidity or asymmetric branching. As shown in the previous section of this paper, energy
losses and sound wave attenuation due to air viscosity were also insignificant. In other words,
the nasal cavity models used in this study met all the assumptions made in AR. This means that
there must be other reasons for the area underestimation, oscillation in area–distance curves
and inaccurate reflection of sudden passage area changes that are seen with ARin stepped-tube
models. These issues are discussed below.
5.1. Area underestimation beyond a constriction
It is well established that a narrowing in the anterior part of the nasal cavity causes errors in
AR-derived areas posterior to the site of constriction. The data from this study of stepped-
tube models also show that AR consistently underestimates the cross-sectional areas beyond
an insert (anterior constriction) of small passage area. We found that, as the passage area
of the insert was reduced from 0.283 cm
2
(insert diameter 0.6 cm), the degree of area
underestimation increased markedly. In other words, regardless of the particular shape of
the nasal cavity model, decreasing the passage area of the anterior constriction significantly
interferes with AR measurement of cross-sectional areas beyond the constriction. Such area
underestimations beyond the constricted site and oscillations in the corresponding portion
of the area–distance curve indicate that AR measurements of this region may easily lead
to misinterpretation of a patient’s nasal anatomy or condition. Pathologies that narrow
the anterior nasal passage, such as septal deviation, polyp, tumour, web or stricture may
cause significant errors in AR measurements of passage areas posterior to such severe
constrictions.
In previous reports, the area underestimation that occurs with AR has been attributed
to viscous forces, transmission losses and internal losses that take place as the sound
wave is transmitted through the constriction in an airway model (Brooks et al 1984,
Buenting et al 1994, Djupesland and Lyholm 1998, Hamilton et al 1995, Hilberg et al 1989,
1998, Hilberg and Pedersen 2000, Hoffstein and Fredberg 1991, Jackson et al 1977, Jackon
and Olson 1980, Mlynski et al 2003, Terheyden et al 2000). It has been argued that any form
of energy loss or sound wave attenuation would reduce the amplitude of the reflected wave,
which, in turn, would lead to area underestimation. It has also been suggested that narrowing
of the bandwidth of the incident sound pulse may reduce the size of the area beyond a severe
constriction as measured by AR (Hilberg et al 1998, Sondhi and Resnick 1983). With regard
to the actual nasal cavity, Hilberg et al (1989) noted that viscous losses in the narrow anterior
portion of the nasal cavity are responsible for underestimation of the area beyond. However,
as we demonstrated in the above section of this paper, the effects of energy losses and sound
wave attenuation due to air viscosity are negligible in stepped-tube models with inserts larger
than 0.2 cm diameter.
In a recent study on simple pipe models of the nasal cavity, Cankurtaran et al (2003) proved
that the reason distal passage area is underestimated with AR is that the barrier created by the
anterior constriction reflects most of the incident sound power. The authors showed that the
presence of a constriction in the model affected the homogeneity of the frequency spectrum of
the transmitted wave, because the acoustic input impedance of the model system is frequency
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 383
dependent. In models with constrictions (inserts) of small passage area, the high-frequency
components of the acoustic pulse generated by AR equipment do not reach the portion of
the model beyond the constriction, because these waves are reflected back from the barrier
created by the constriction. This finding is of vital importance for AR, since the transmitted
sound waves probe, and hence provide information about, the cross-sectional area posterior
to the constriction. In other words, an examiner should expect relatively higher degrees of
error when measuring the cross-sectional area of a nasal cavity model beyond a constriction of
small passage area. Since AR measures the intensity of reflected sound waves and compares
this with the intensity of incident waves, one would expect the measured cross-sectional areas
beyond the constriction to be lower than the actual cross-sectional area.
The results we obtained with stepped-tube models of the nasal cavity confirm our recent
experimental findings in and theoretical studies of simple pipe models (Cankurtaran et al
2003). A similar interpretation may apply to the actual nasal cavity, with its anterior narrow
segment and its acoustic pathway of finite length. Of course, the geometry of the nasal cavity
is much more complicated than the axially symmetric stepped-tube models we investigated.
However, Hilberg et al (1998) argued that the complex geometry of the actual nasal cavity had
no significant effects on AR measurements.
5.2. Oscillations in AR-derived area–distance curves
The AR-derived area–distance curve for each cylindrical section of the stepped-tube model
exhibits significant oscillations (see figures 2 and 3). In a nasal cavity model consisting
of many cylindrical sections, each time the incident sound wave encounters a new section,
part is reflected and part is transmitted further along the passage. Our theoretical results
demonstrate that superposition of sound waves travelling in opposite directions in each
cylindrical section leads to changes in acoustic impedance, even though the cross-sectional
area of the section remains constant. These changes (oscillations) in acoustic impedance
are due to low-frequency acoustic resonances in the stepped-tube model. Consequently, the
incident sound waves will be reflected not only due to axial variations in acoustic impedance
from changes in cross-sectional area along the length of the model, but also due to axial
variations in acoustic impedance caused by these acoustic resonances. When the incident
wave is reflected by changes in acoustic impedance that are not caused by changes in cross-
sectional area, the AR algorithm interprets these changes as variations in the cross-sectional
area of tube model. Therefore, if the sound-wave propagation occurs as lossless plane waves
and the walls of the acoustic pathway are rigid, the assertion (Hilberg et al 1989, Jackson et al
1977) that all changes in acoustic impedance are due to changes in cross-sectional area is not
valid.
5.3. Inaccurate reproduction of sudden changes in passage area
It has been argued that the limited frequency bandwidth of the acoustic pulse reflectometry
technique may increase the rise distance, and thereby produce a smoother incline in the
area–distance curve (Fredberg et al 1980, Hilberg et al 1998, Sondhi and Resnick 1983).
The frequency bandwidth (f ) of the incident sound pulse is important in determining
the spatial resolution of the technique, and hence has a major influence on the accuracy
of AR measurements. The spatial resolution (x) is defined as the smallest axial distance
that separates two cross-sectional areas that can still be resolved by the acoustic technique
(Fredberg et al 1980, Hilberg et al 1998, Hilberg and Pedersen 2000). In rigid-walled airways,
the spatial resolution is approximately equal to c/6f , or one-sixth of the shortest wavelength
384 H C¸ elik et al
of the incident sound pulse (Djupesland and Lyholm 1998, Fredberg et al 1980, Sidell and
Fredberg 1978). Using c = 34 300 cm s
−1
and f = 9900 Hz we obtained x = 0.6 cm,
which is much smaller than (approximately half of) the rise distance we observed in our
experiments (see figure 2). This limited spatial resolution partially explains why the AR-
derived area–distance curve does not precisely mirror the sudden changes in passage area that
occur at each junction between cylindrical sections of the stepped-tube model.
The inability of AR in reproducing these sharp changes in area can also be attributed in
part to the approximations made in the algorithm used to generate the area–distance curve
from the input impulse response of the system, because this algorithm affects the accuracy
of the method (Fredberg et al 1980, Hilberg et al 1998, Hilberg and Pedersen 2000). Most
previous investigations of airways (including the nasal cavity) made using acoustic pulse
reflectometry have used the algorithm developed by Ware and Aki (1968). This algorithm
assumes that the sound waves are plane waves, and it does not account for losses (airway wall
non-rigidity, viscous losses) or non-planar wave propagation effects (Fredberg et al 1980,
Hilberg et al 1989, Sidell and Fredberg 1978). All these assumptions were considered in most
of the earlier reports. However, there is a further requirement that is important: the Ware–Aki
algorithm is valid under the condition that the acoustic impedance of the one-dimensional
acoustic pathway is continuous (Ware and Aki 1968). If there is a finite sudden jump in the
acoustic impedance, the transformations and the potential functions used in the mathematical
formulation of this algorithm are not well defined (Ware and Aki 1968). In other words, it
appears that the Ware–Aki algorithm is not suitable for calculating the area–distance function
at locations where there are abrupt changes in the acoustic impedance of the model system.
Sondhi and Gopinath (1971) also developed a solution to the inverse problem of calculating
the area–distance function from the input impulse response of an airway. In their treatment,
as in the Ware–Aki formalism, sound wave propagation was assumed to be planar, and losses
in the airway were not taken into account. Their treatment is valid under the condition that
the cross-sectional area S(x) as a function of the distance x from the input to the airway is
finite and twice continuously differentiable. Therefore, the algorithm developed by Sondhi
and Gopinath is not applicable to airway models, which feature abrupt changes in passage
area.
Recently, Hilberg and Pedersen (2000) proposed a set of guidelines for optimal application
of ARand presented experimental area–distance curves for a step model with nasal valve area of
0.38 cm
2
(inner diameter 0.695 cm). They showed that AR underestimated the cross-sectional
areas of the step model beyond the valve region, and that the corresponding portion of the
area–distance curve showed marked oscillations. The results of our present study on
stepped-tube models confirm that AR underestimates the cross-sectional area beyond an
insert of diameter d < 0.6 cm (passage area <0.283 cm
2
), and that the degree of this area
underestimation decreases as d is increased above 0.6 cm. Hilberg and Pedersen (2000) argued
that ‘steep changes cause underestimation of the area’ and noted ‘We knowthat the area behind
a severe constriction is underestimated, but we do not yet know how to correct for this.’ They
also stated ‘The technique has some physical limitations, which cannot be changed, but some
technical aspects of the equipment can be adjusted to obtain higher accuracy.’ The results of
the model calculations we have presented in this study and in previous studies (Cakmak et al
2003a, Cankurtaran et al 2003) reveal the physical limitations of both the technique and
the algorithm used in AR. They also identify the physical causes and reasons for area
underestimation, oscillations in AR-derived area–distance curves and inaccurate reproduction
of sharp changes in passage area. These physical limitations of the AR technique should be
considered for further improvement in the design of AR equipment and related computer
software.
Acoustic rhinometry measurements in stepped-tube models of the nasal cavity 385
6. Conclusions
The experimental and theoretical results we obtained for stepped-tube nasal cavity models with
a constriction in the anterior section reveal the following: (i) oscillations in AR-derived area–
distance curves are due to low-frequency acoustic resonances in the model, (ii) energy losses
and sound wave attenuation due to air viscosity have negligible effects on area determination
by AR, (iii) area underestimation beyond a constriction of small passage area is mainly due to
reflection of most of the incident sound power by the barrier formed by the constriction and
(iv) AR technique does not accurately reproduce sudden changes in passage area. The latter
is due to the limited spatial resolution of AR and the approximations made in the inversion
algorithm. The results suggest that the accuracy of AR measurements of the nasal cavity
depends heavily on the passage area of the narrowest section of the nasal passage. Acoustic
rhinometry measurements of the anterior nasal passage are reasonably accurate if the nasal
valve passage area is within the normal range. However, reduction of nasal valve passage area
due to any cause or pathology weakens the diagnostic value of AR for measuring the entire
nasal cavity.
References
Brooks L J, Castile R G, Glass M G, Griscom N T, Wohl M E B and Fredberg J J 1984 Reproducibility and accuracy
of airway area by acoustic reflection J. Appl. Physiol. 57 777–87
Buenting J E, Dalston R M, Smith T L and Drake A F 1994 Artifacts associated with acoustic rhinometric assessment
of infants and young children J. Appl. Physiol. 77 2558–63
Cakmak O, Celik H, Cankurtaran M, Buyuklu F, Ozgirgin N and Ozluoglu L N 2003a Effects of paranasal sinus ostia
and volume on acoustic rhinometry measurements: a model study J. Appl. Physiol. 94 1527–35
Cakmak O, Celik H, Ergin T and Sennaroglu L 2001 Accuracy of acoustic rhinometry measurements Laryngoscope
111 587–94
Cakmak O, Coskun M, Celik H, Buyuklu F and Ozluoglu L N 2003b Value of acoustic rhinometry for measuring
nasal valve area Laryngoscope 113 295–302
Cankurtaran M, Celik H, Cakmak O and Ozluoglu L N 2003 Effects of the nasal valve on acoustic rhinometry
measurements: a model study J. Appl. Physiol. 94 2166–72
Djupesland P G and Lyholm B 1998 Technical abilities and limitations of acoustic rhinometry optimised for infants
Rhinology 36 104–13
Fredberg J J, Wohl M E B, Glass G M and Dorkin H L 1980 Airway area by acoustic reflections measured at the
mouth J. Appl. Physiol. 48 749–58
Hall D E 1987 Basic Acoustics (New York: Wiley)
Hamilton J W, Cook J A, Phillips D E and Jones A S 1995 Limitations of acoustic rhinometry determined by a simple
model Acta Otolaryngol. (Stockh) 115 811–4
Hilberg O, Jackson A C, Swift D L and Pedersen O F 1989 Acoustic rhinometry: evaluation of nasal cavity geometry
by acoustic reflection J. Appl. Physiol. 66 295–303
Hilberg O, Lyholm B, Michelsen A, Pedersen O F and Jacobsen O 1998 Acoustic reflections during rhinometry:
spatial resolution and sound loss J. Appl. Physiol. 84 1030–9
Hilberg O and Pedersen O F 2000 Acoustic rhinometry: recommendations for technical specifications and standard
operating procedures Rhinology Suppl. 16 3–17
Hoffstein V and Fredberg J J 1991 The acoustic reflection technique for non-invasive assessment of upper airway
area Eur. Respir. J. 4 602–11
Jackson A C, Butler J P, Millet E J, Hoppin F G Jr and Dawson S V 1977 Airway geometry by analysis of acoustic
pulse response measurements J. Appl. Physiol. 43 523–36
Jackson A C and Olson D E 1980 Comparison of direct and acoustical area measurements in physical models of
human central airways J. Appl. Physiol. 48 896–902
Kinsler L E and Frey A R 1962 Fundamentals of Acoustics 2nd ed (New York: Wiley)
Louis B, Fodil R, Jaber S, Pigeot J, Jarreau P-H, Lofaso F and Isabey D 2001 Dual assessment of airway area profile
and respiratory input impedance from a single transient wave J. Appl. Physiol. 90 630–7
Marshall I, Rogers Mand Drummond G1991 Acoustic reflectometry for airway measurements. Principles, limitations
and previous work Clin. Phys. Physiol. Meas. 12 131–41
386 H C¸ elik et al
Mlynski R, Gr¨ utzenmacher S, Lang C and Mlynski G 2003 Acoustic rhinometry and paranasal cavities: a systematic
study in box models Laryngoscope 113 290–4
Putland G R 1996 Modeling of horns and enclosures for loudspeakers PhD Thesis University of Queensland
Sharp DB1996 Acoustic pulse reflectometry for the measurements of musical wind instruments PhDThesis University
of Edinburgh
Sidell R S and Fredberg J J 1978 Noninvasive inference of airway network geometry from broadband lung reflection
data J. Biomech. Eng. 100 131–8
Sondhi M M and Gopinath B 1971 Determination of vocal-tract shape from impulse response at the lips J. Acoust.
Soc. Am. 49 1867–73
Sondhi MMand Resnick J R1983 The inverse problemfor the vocal tract: numerical methods, acoustical experiments,
and speech synthesis J. Acoust. Soc. Am. 73 985–1002
Terheyden H, Maune S, Mertens J and Hilberg O 2000 Acoustic rhinometry: validation by three-dimensionally
reconstructed computer tomographic scans J. Appl. Physiol. 89 1013–21
Tomkinson A and Eccles R 1998 Acoustic rhinometry: an explanation of some common artefacts associated with
nasal decongestion Clin. Otolaryngol. 23 20–6
Ware J A and Aki K 1968 Continuous and discrete inverse-scattering problems in a stratified elastic medium. I. Plane
waves at normal incidence J. Acoust. Soc. Am. 45 911–21