in vitro comparison of water displacement method and 3 tesla mri for mr-volumetry of the olfactory...
TRANSCRIPT
Original Investigation
In Vitro Comparison of WaterDisplacement Method and 3 Tesla
MRI for MR-Volumetryof the Olfactory Bulb:
Which Sequence Is Appropriate?
Hartmut Peter Burmeister, MD, Constanze M€oslein, Thomas Bitter, MD, Rosemarie Fr€ober, MD,Karl-Heinz Herrmann, PhD, Pascal Andreas Thomas Baltzer, MD, Hilmar Gudziol, MD,Matthias Dietzel, MD, Orlando Guntinas-Lichius, MD, Werner Alois Kaiser, MD, MSc
Ac
FrP.HoJeOLeFr(Rde
ªdo
Rationale and Objectives: Magnetic resonance imaging olfactory bulb (OB) volumetry (OBV) is already used as a complementary prog-nostic tool to assess olfactory disorders. However, a reference standard in imaging forOBVhas not been established. The aimof this in vitro
study was to compare volumetric results of different magnetic resonance sequences for OBV at 3 T to genuine OB volumes measured by
water displacement.
Materials and Methods: The volumes of 15 human cadaveric OBs were measured using the water displacement method in this institu-
tional review board–approved prospective study. The magnetic resonance imaging protocol at 3 T included constructive interference in
steady state (CISS), T2-weighted (T2w) three-dimensional (3D) sampling perfection with application-optimized contrasts using different
flip-angle evolutions (SPACE), T2w two-dimensional (2D) turbo spin-echo (TSE), and T1-weighted (T1w) 3D fast low-angle shot (FLASH)sequences. Two blinded observers independently performed two OB volumetric assessments per bulbus and sequence. Intraobserver
and interobserver reliabilities were assessed by intraclass correlation coefficients. Bland-Altman plots were analyzed to evaluate system-
atic biases and concordance correlation coefficients to assess reproducibility.
Results: For both observers, intraclass correlation coefficient analysis yielded almost perfect results for intraobserver reliability (CISS,
0.94–0.98; T2w 3D SPACE, 0.93–0.98; T2w 2D TSE, 0.98–0.98; T1w 3D FLASH, 0.95–0.99). Interobserver reliability showed almost perfect
agreement for all sequences (CISS, 0.98; T2w 3D SPACE, 0.89; T2w 2D TSE, 0.93; T1w 3D FLASH, 0.97). The CISS sequence yielded the
highest mean concordance correlation coefficient (0.95) and the highest combination of precision (0.97) and accuracy (0.98) values. Incomparison with the water displacement method, Bland-Altman analyses revealed the lowest systematic bias (�0.5%) for the CISS
sequence, followed by T1w 3D FLASH (�1.3%), T2w 3D SPACE (–7.5%), and T2w 2D TSE (�10.9%) sequences.
Conclusions: Compared to the water displacementmethod, the CISS sequence is suited best to validly and reliably measure OB volumesbecause of its highest values for accuracy and precision and lowest systematic bias.
Key Words: Olfactory bulb; volumetry; cadaver study; reproducibility; 3 Tesla MRI.
ªAUR, 2011
Olfactory bulb (OB) volumetry (OBV) has become
a growing topic (1) in the field of understanding
olfactory dysfunction, for which studies have indi-
ad Radiol 2011; -:1–8
om the Institute of Diagnostic and Interventional Radiology (H.P.B., C.M.,A.T.B., M.D., W.A.K.); Medical Physics Group (K.-H.H.), Universityspital - Friedrich Schiller University Jena, Philosophenweg 3, D-07740na, Germany; and the Department of Otorhinolaryngology (T.B., H.G.,.G.-L.), University Hospital - Friedrich Schiller University Jena,ssingstrasse 2, D-07743 Jena, Germany; and the Institute of Anatomy I,iedrich Schiller University Jena, Teichgraben 7, D-07743 Jena, Germany.F.). Received May 16, 2011; accepted June 27, 2011. Address correspon-nce to: H.P.B. e-mail: [email protected]
AUR, 2011i:10.1016/j.acra.2011.06.008
cated a prevalence of about 20% in the Western world (2,3).
OB volume decreases have been identified in all of the five
most frequent etiologies (4) of smelling disorders (ie, sinonasal
disease [5], postinfectious [6], posttraumatic [6], neurodegen-
erative diseases [7], and idiopathic olfactory disorders [8]).
However, volume increases have also been shown (eg, after
successful olfactory rehabilitation in sinonasal diseases [9]),
which exemplifies the reversibility of OB volume changes.
These dynamic changes are most probably due to the fact
that the human OB retained the ability of neuroneogenesis
(10) and therefore exhibits high structural plasticity, in which
its volume is correlated to afferent neural activity transmitted
by the olfactory receptor neurons (11). OBV has already been
1
BURMEISTER ET AL Academic Radiology, Vol -, No -, - 2011
used as a complementary prognostic tool for radiologic diag-
nosis to predict outcomes in olfactory disorders (7).
In 1997, Yousem et al (12) described magnetic resonance
(MR) imaging (MRI) as a feasible method for OBV on the
basis of T1-weighted (T1w) sequences at 1.5 T. Further devel-
opment of MRI techniques and the use of MRI at higher field
strength (ie, 3 T) could facilitate the quantification of volu-
metric results of this small paleocortical structure because
the increased signal-to-noise ratio can be invested in better
spatial resolution (13). Therefore, future studies dealing with
the subject of high-resolution MROBV should be based on
standardized imaging, among other factors, for compatibility
reasons. However, until now, a systematic radiologic compar-
ison of various standardized sequences for OB volumetric
purposes, which logically justifies a reference standard in
imaging for upcoming studies, has not been finalized.
Previous in vivo studies have evaluated image quality (14)
and systematic biases (15) of different MR sequences for
OBVat 3 T. However, these studies were not suited for com-
menting on the true accuracy of volumetric measurements,
because of the lack of a direct anatomic correlation (12). An
additional measurement of genuine cadaveric OB volumes
using the water displacement method (WDM) would be suit-
able (12) to verify the accuracy of MROBVand remains the
missing link in the collection of prerequisites to define a valid
and reliable reference standard. Until now, this type of verifi-
cation procedure had to be precluded because of the paucity of
cadaveric OBs available for research purposes.
Consequently, in this in vitro MRI study, we evaluated the
reliability, reproducibility, and systematic bias of different
MRI sequences used for MROBV at a field strength of 3 T
and correlated with realistic OB volumes measured by water
displacement.
MATERIALS AND METHODS
Subject Recruitment, Sampling, and ExperimentalDesign
In this prospective study, we examined a series of 15 human
cadaveric OBs from 11 human body donors from our institute
of anatomy. Authorized agreement for scientific medical
research and education was provided before death, and insti-
tutional review board approval was obtained.
OBs were included in this study if the following criteria
were fulfilled: donor age $ 18 years at the date of death,
macroscopic normal OB shape, an estimated regular OB
size, and completeness of the OB and the transitional zone
between the OB and the olfactory tract after resection. Exclu-
sion criteria were as follows: anamnestic data (if available)
indicating anosmia, a traumawith fracture of the anterior skull
base, or any known previous surgery of the anterior skull base.
Immediately after transfer to our institute of anatomy,
embalmment was initialized by injecting embalming fluid
into the human cadaver (aqueous ethanol-formaldehyde solu-
tion; application of about 10 L via the right femoral artery).
2
The complete brain (including the OBs) was removed 1 to
2 days after the donor’s death, and all brains were subsequently
stored within aqueous formalin solution (formaldehyde 5%).
Furthermore, the principal prosector (R.F.) carefully resected
all OBs, including the adjacent part of the olfactory tract.
Pipetting tubes were filled with aqueous formalin solution
(formaldehyde 5% stabilized with methanol) for immersion
fixation purposes, and the OBs were placed in the pipetting
tubes with the frontal pole of the OBs pointing downward
(Fig 1). Subsequently, the tubes were stored at room temper-
ature for 2 weeks to complete the chemical reactions of fixa-
tion and to equalize the grade of formalin impregnation of the
bulbs at the best possible rate. Then, the OBs were washed
over 48 hours with four isotonic saline solution (sodium chlo-
ride 0.9%) changes at room temperature. Repetitive T1w and
T2-weighted (T2w)MRI of the OBs was performed between
each of the four saline solution changes to define the end point
of the washout procedure on the basis of a finalization of signal
intensity changes. After the fourth pass, pipetting tubes were
filled with saline solution, the OBs were cautiously inserted
into the tubes, and the saline solution’s supernatant was care-
fully taken up with a pipette for OB volume measurement
analogous to the WDM (T.B.). Consequently, the formalin
solution was completely replaced by isotonic saline solution,
and the tubes were filled bubble-free with saline solution.
Furthermore, an embedding of the pipetting tubes in an
agar-agar phantom (600 mL water, 7% agar-agar) was per-
formed, with a special focus on avoiding adjacent air bubbles.
The tubes were arranged in circular and clockwise order (Fig
1) according to the body register of the institute of anatomy
(order based on date of death and the OB’s side).
Imaging Procedures
All OBs underwent 3-T MRI examinations (Magnetom Tim
Trio; Siemens Medical Systems, Erlangen, Germany) at our
institute of radiology using a dedicated 12-channel head coil
provided by the manufacturer for image acquisition. The
OBs’ orientations were equivalent to those of patients exam-
ined in the prone position with a coronal slice orientation
tilted by 90% to the horizontal lamella of the cribriform plate.
The imaging protocol included the following sequences:
a coronal constructive interference in steady state (CISS)
three-dimensional (3D) Fourier transformation sequence
(repetition time [TR], 11.68 ms; echo time [TE], 5.84 ms;
matrix size, 512 � 512; field of view, 256 mm; voxel size,
0.5� 0.5� 0.5 mm [0.125 mm3]; flip angle, 70�; bandwidth,130 Hz/pixel; averages, 1; acquisition time [TA], 7 minutes
48 seconds), a coronal T2w 3D sampling perfection with
application-optimized contrasts using different flip-angle
evolutions (SPACE) sequence (TR, 1000 ms; TE, 132 ms;
matrix size, 384 � 384; field of view, 200 mm; voxel size
0.5 � 0.5 � 0.5 mm [0.125 mm3]; flip angle, 120�; band-width, 289 Hz/pixel; averages, 2; TA, 4 minutes 18 seconds),
and a coronal T2w two-dimensional (2D) turbo spin-echo
(TSE) sequence (TR, 4800 ms; TE, 150 ms; matrix, 512 �
Figure 1. Experimental design for volumetricmeasurements to compare magnetic reso-
nance (MR) sequences to the water displace-
ment method (WDM) for olfactory bulb (OB)
volumetry at 3 T. (1) Resection of the OBincluding an adjacent part of the olfactory tract.
(2) Immersion fixation with aqueous formalin
solution (formaldehyde 5% stabilized with
methanol) and storage of the OBs for 2 weeks,followed by a quadruplicated washout proce-
dure with isotonic saline solution (sodium chlo-
ride 0.9%). Taking up saline solution’ssupernatant with a pipette analogous to the
WDM to determine OB volumes (V1). (3) Embed-
ding the pipetting tubes including the OBs
(frontal pole pointing downward) in an agar-agar phantom (4). (5) MR imaging (MRI) and
acquisition of the four sequences (Fig 2) for
comparison to the WDM. (6) Use of multiplanar
reconstructions to perform MR volumetry (7)and to determine OB MRI volume (V2).
Academic Radiology, Vol -, No -, - 2011 IN VITRO 3-T OLFACTORY BULB MR VOLUMETRY
512; field of view, 230 mm; voxel size, 0.4 � 0.4 � 2.0 mm
[0.32 mm3]; no distance factor; flip angle, 150�; bandwidth,130 Hz/pixel; averages, 2; TA, 8 minutes 50 seconds).
Furthermore, the imaging protocol was completed with an
additional coronal T1w 3D fast low-angle shot (FLASH)
sequence (TR, 20.0 ms; TE, 2.46 ms; matrix size, 384 �384; flip angle, 20�; voxel size, 0.6 � 0.6 � 0.6 mm
[0.216 mm3]; averages, 2; TA, 16 minutes; no fat saturation).
Imaging Data Analysis
Acquired images were transferred to two Advantage Worksta-
tion VolumeShare 2 workstations (Advantage Workstation
version 4.4; GE Medical Systems, Milwaukee, WI) for
analysis.
Two observers, one specialized in skull base imaging and
volumetry (H.P.B., 9 years’ experience) and the other experi-
enced in skull base imaging (C.M., second year of experience)
independently performed volumetry of all 15 OBs separately
using the four sequences applied (Fig 2).
Each observer performed MR volumetry of all OBs twice.
Repeatability conditions in this study included the same
measurement procedure (sequence), the same observers, the
same measuring instrument (used under the same conditions,
ie, 3-T MRI, Advantage Workstation VolumeShare 2 work-
stations), the same anatomic structure (the OB), and repeti-
tion over a short period of time (1 day). The adjacent part
of the olfactory tract was included in the volumetric
assessment because the transitional zone between the OB
and olfactory tract is sometimes more difficult to determine,
as previously stated (16).
The selection of the volume was performed by planimetric
manual contouring of the OB using a circular tool 1 mm in
diameter, which is described in detail in our previous in vivo
study (15). Multiplanar reconstructions in the axial, sagittal,
and coronal planes were used for performing volumetry.
Each slice orientationwas automatically indicated on the other
reconstruction planes simultaneously showing the currently
selected volume. This approach allows for an immediate visual
correlation and corrections. For 3D sequences (CISS, T2w 3D
SPACE, and T1w 3D FLASH), a selection of the volume was
possible using every multiplanar reconstruction plane ad
libitum. However, for volumetry of the T2w 2D TSE
sequence, preconditions were set analogous to the previously
preferred method for volumetry of 2D T1w and T2w
sequences, as described in detail by Rombaux et al (16). The
selection of the volume in the T2w 2D TSE sequence was
exclusively performed in coronal planes. Using the latter
method, OB volumes were calculated by planimetric manual
contouring (surface of coronal planes in square millimeters),
and all surfaces were added and multiplied by 2 because
of the 2-mm slice thickness to obtain a volume in cubic
millimeters (17).
Artifacts should be assessed only with special focus on the
contour sharpness of the OB’s surface. To prevent habituation
effects during MR volumetry, the order of performing MR
volumetry was randomized concerning OBs and sequences
in each of the two passes. During the second pass, the
observers were blinded to the results of the first pass.
Statistical Analysis
All data analyses were performed using SPSS version 15.0
(SPSS, Inc, Chicago, IL) andMedCalc version 11.0 (MedCalc
Software, Mariakerke, Belgium).
Interobserver and intraobserver reliability with corre-
sponding 95% confidence intervals (CIs) for all sequences
was determined by using intraclass correlation coefficients
(ICCs; type 3.k) (18). Results were benchmarked according
to the classification of Landis and Koch (19) for measurements
of observer agreement for categorical data.
Because interobserver reliability results using ICCs do not
address both precision and accuracy, additional concordance
correlation coefficients (CCCs) were calculated (20). The
3
Figure 2. Comparison of sequences appliedfor in vitro olfactory bulb volumetry at 3 T. (a)Constructive interference in steady state
sequence (repetition time [TR], 11.68 ms;
echo time [TE], 5.84 ms; voxel size,0.125mm3); (b) T2-weighted (T2w) three-
dimensional (3D) sampling perfection with
application-optimized contrasts usingdifferent flip-angle evolutions sequence (TR,
1000 ms; TE, 132 ms; voxel size, 0.125mm3);
(c) two-dimensional T2w turbo spin-echo
sequence (TR, 4800 ms; TE, 150 ms; voxelsize, 0.32mm3); (d) T1-weighted 3D fast low-
angle shot sequence (TR, 20.0 ms; TE,
2.46 ms; voxel size, 0.216 mm3). All images
are in the sagittal plane and depict the sameolfactory bulb.
BURMEISTER ET AL Academic Radiology, Vol -, No -, - 2011
CCC (rc) provides a means for examining the reproducibility
of continuous volumetric measurements made by multiple
observers using a single method (21).
To estimate systematic bias of the four different sequences
compared to theWDM, Bland-Altman plots (22) were gener-
ated by graphing the percentage difference of each volumetric
measurement result (average of two observers and two passes),
the mean percentage difference, and the confidence limits
(95% limits of agreement) on the vertical against the absolute
average volumetric results of the two methods on the
horizontal.
All statistical analyses in this study were based on a signifi-
cance level of a = 5%. To address possible a error accumula-
tion, P values in this paper are given as calculated. For
interpretation of results, classical Bonferroni correction was
applied (23).
RESULTS
Fifteen of 22 OBs (nine right sided, six left sided) met the
inclusion criteria and underwent both WDM and 3-T MRI
volumetry between February 5 and February 15, 2011. Prior
to resection, seven OBs were excluded because of estimated
hypotrophy and/or irregular shape. Regarding our exclusion
criteria, no OB had to be excluded, and consecutively all 15
resected OBs were eligible for volumetry and statistical evalu-
ation. The ages of the body donors ranged from 59 to 90 years
(mean, 76.5� 13.8 years), and the dates of death ranged from
2008 to 2011. Because in five body donors, some of the
demographic data were anonymized before this study, no
gender-specific differences were evaluated.
OB volumes determined by water displacement ranged
from 75 to 155 mm3 (mean, 106.33 � 21.05 mm3). Corre-
sponding MR volumetric measurements ranged from 77.5
to 157.5 mm3 (mean, 105.58 � 20.00 mm3) for the CISS
sequence, 64.5 to 135 mm3 (mean, 98.31 � 16.64 mm3) for
the T2w 3D SPACE sequence, 66 to 131.5 mm3 (mean,
95.11 � 16.89 mm3) for the T2w 2D TSE sequence, and
4
74.5 to 145 mm3 (mean, 104.83 � 19.34 mm3) for the
T1w 3D FLASH sequence. Artifacts did not significantly
restrict the evaluation of the OBs and olfactory tracts con-
cerning the aim of this study.
Intraobserver reliability for the first observer resulted in
ICCs of 0.94 (95% CI, 0.83–0.98) for the CISS sequence,
0.93 (95% CI, 0.80–0.98) for the T2w 3D SPACE sequence,
0.98 (95% CI, 0.94–0.99) for the T2w 2D TSE sequence, and
0.95 (95% CI, 0.86–0.98) for the T1w 3D FLASH sequence.
Intraobserver reliability for the second observer resulted in
ICCs of 0.98 (95% CI, 0.93–0.99) for the CISS sequence,
0.98 (95% CI, 0.95–0.99) for the T2w 3D SPACE sequence,
0.98 (95% CI, 0.96–0.99) for the T2w 2D TSE sequence, and
0.99 (95% CI, 0.95–0.99) for the T1w 3D FLASH sequence.
Interobserver reliability resulted in ICCs of 0.98 (95% CI,
0.90–0.99) for the CISS sequence, 0.89 (95% CI, 0.21–
0.97) for the T2w 3D SPACE sequence, 0.93 (95% CI,
0.78–0.98) for the T2w 2D TSE sequence, and 0.97 (95%
CI, 0.92–0.99) for the T1w 3D FLASH sequence.
The results of reproducibility evaluation using CCCs with
corresponding values for 95% CI, precision, and accuracy are
summarized in Table 1.
The mean biases and 95% limits of agreement for
volumetric measurement results (percentage values in cubic
millimeters) regarding the comparison of the WDM and the
four sequences applied are displayed in the corresponding
Bland-Altman plots in Figure 3.
DISCUSSION
According to our results, for both observers, ICC analyses of
intraobserver reliability yielded almost perfect agreement for
the CISS, T2w 2D TSE, and T1w 3D FLASH sequences.
Additionally, the widths of associated 95% CIs did not extend
beyond the limits of 0.81 to 1.00 for almost perfect agreement
according to the classification of Landis and Koch (19).
Although the T2w 3D SPACE sequence also resulted in
almost perfect agreement for both observers, the range of
TABLE 1. CCCs, Interobserver Reliability, and Reproducibility (n = 15)
Sequence CCC (rc) 95% CI Pearson’s r (Precision)
Bias Correction
Factor cb (Accuracy)
CISS
First evaluation 0.92 0.80–0.97 0.95 0.97
Second evaluation 0.90 0.73–0.96 0.91 0.99
Average 0.95 0.88–0.98 0.97 0.98
T2w 3D SPACE
First evaluation 0.70 0.41–0.86 0.84 0.83
Second evaluation 0.83 0.60–0.93 0.90 0.92
Average 0.80 0.57–0.91 0.91 0.88
T2w 2D TSE
First evaluation 0.85 0.63–0.94 0.86 0.99
Second evaluation 0.83 0.58–0.94 0.84 0.99
Average 0.85 0.63–0.95 0.86 0.99
T1w 3D FLASH
First evaluation 0.93 0.82–0.98 0.94 0.99
Second evaluation 0.89 0.70–0.96 0.89 0.99
Average 0.94 0.84–0.98 0.94 0.99
CCC, concordance correlation coefficient; CI, confidence interval; CISS, constructive interference in steady state; FLASH, fast low-angle shot;
SPACE, sampling perfection with application-optimized contrasts using different flip-angle evolutions; 3D, three-dimensional; T1w,
T1-weighted; TSE, turbo spin-echo; T2w, T2-weighted; 2D, two-dimensional.
Figure 3. Systematic biases of magnetic reso-nance sequences compared to the water
displacement method (WDM) for in vitro olfac-
tory bulb volumetry. Bland-Altman plots of
percentage differences of volumetric measure-ment results (y axis; percentage) against
average volumetric measurement results
(x axis; absolute values), with mean percentagedifference (ie, bias; continuous line) and 95%
limits of agreement (dashed lines) for (a)comparison of the constructive interference in
steady state (CISS) sequence and the WDM,(b) comparison of the T2-weighted (T2w) three-
dimensional (3D) sampling perfection with
application-optimized contrasts using different
flip-angle evolutions sequence and the WDM,(c) comparison of the T2w two-dimensional
turbo spin-echo sequence and the WDM, and
(d) comparison of the T1-weighted 3D fastlow-angle shot sequence and the WDM. The
dash-dotted regression lines proved not to be
significant in (a) (slope = �0.988, P = .332), (b)(slope = �1.111, P = .276), (c) (slope = �0.969,P = .341), and (d) (slope = �0.601, P = .553).
All values are given in cubic millimeters. SD,
standard deviation.
Academic Radiology, Vol -, No -, - 2011 IN VITRO 3-T OLFACTORY BULB MR VOLUMETRY
the associated 95% CI for the first observer (0.80–0.98) only
barely missed the limits for almost perfect agreement.
Regarding interobserver reliability, the CISS, T2w 2D
TSE, and T1w 3D FLASH sequences showed almost perfect
agreement, whereas the T2w 3D SPACE sequence resulted in
only substantial agreement. However, apart from the fact that
the width of the associated 95% CI (0.21–0.97) for T2w 3D
SPACE clearly extended beyond the limits for almost perfect
agreement, the associated 95% CI (0.78–0.98) for T2w 2D
TSE narrowly missed those limits.
Nevertheless, intraobserver reliability for both observers
and interobserver reliability using the CISS, T2w 3D SPACE,
and T2w 2D TSE sequences were noticeably higher
compared to the previously reported in vivo results from
our previous study at 3 T (15). Comparing intraobserver reli-
abilities between the two observers, the greater experience of
5
BURMEISTER ET AL Academic Radiology, Vol -, No -, - 2011
the first observer had no favorable impact; conversely, the less
experienced observer performed slightly better. These latter
two facts are most likely attributable to the peculiarity of
this in vitro study design (eg, no surrounding soft tissue or
osseous structures, no motion artifacts; see below).
On the topic of reproducibility, the CISS sequence yielded
the best results for CCC and the best combination of precision
and accuracy values (Table 1) compared to the T2w 3D
SPACE and T2w 2D TSE sequences. It must be stated that
except for the T2w 3D SPACE sequence, all sequences
showed lower precision than accuracy.
According to Bland-Altman plot analysis, in relation to the
WDM, the mean biases were�7.5% for the T2w 3D SPACE
sequence,�10.9% for T2w 2D TSE, and�1.3% for T1w 3D
FLASH. Additionally, associated limits of agreement of the
latter sequences had (in part) a considerably broader range
compared to the CISS sequence (Fig 3). The CISS sequence,
however, yielded the relatively most accurate mean bias of
�0.5% and the narrowest range for associated limits of agree-
ment (8.7% to �9.8%).
Bland-Altman plot analysis implies prima facie that the
methods (MR sequences) and agreement tended to be related
to the underlying size of the OB being measured. However,
none of the dash-dotted regression lines (Fig 3) is related to
a significant regression (for slope and P values, refer to the
legend of Fig 3). Therefore, it is not mandatory to create
size-related conversion tables to adapt the volume measured
by MROBV to the real individual OB’s volume. However,
we recommend this latter conclusion to be restricted to the
volume bandwidth evaluated in this study. Consequentially,
the necessity to use size-related conversion tables for OBs
with hypoplasia or disease-related atrophy should be evaluated
in a separate study.
In terms of intraobserver reliability, interobserver reli-
ability, reproducibility, and systematic bias analyses, we inter-
pret the results as indicating a clear tendency in favor of the
CISS sequence, in particular in consideration of the corre-
sponding results of our previous in vivo studies (14,15). For
the reasons we focused exclusively on T2w imaging in
these two preceding studies, please refer to Burmeister et al
(14,15). However, under the condition of an in vitro
evaluation, the idea was taken up anew to evaluate a high-
resolution T1w 3D FLASH sequence at 3 T for comparison
purposes. It must be stated that although T1w 3D FLASH
imaging did not reach the resolution of the CISS and T2w
3D SPACE sequences, in this study, T1w 3D FLASH
imaging performed second best, and the results were sur-
passed only by the CISS sequence. However, the T1w 3D
FLASH sequence is an exception, because an acquisition
time of 16 minutes was necessary to obtain approximately
comparable resolution. The latter issue and the consecutive
aggravated risk for developing motion artifacts make it diffi-
cult to include such a sequence in clinical routine. For these
reasons, T1w imaging should not yet be favored for OBV
purposes, but the possibility should be kept in mind for future
research.
6
Although this systematic evaluation of sequences regarding
their suitability for OBV purposes was performed under opti-
mized conditions as much as possible, this study was not
immune to limitations. Formaldehyde-fixed nervous tissues
are well suited for high-resolution, time-intensive MRI
acquisitions without motion artifacts (24). However, it is
well known that the use of aldehyde fixatives may significantly
alter tissue MRI properties. For this very reason, we tried to
equalize the grade of formalin impregnation of the bulbs at
the best possible rate, as described under ‘‘Materials and
Methods.’’ Nevertheless, after the end point of the washout
procedure and a finalization of signal intensity changes, we still
had to accept minimal contrast differences between the OBs
of body donors with dates of death after (Fig 1, specimens
1–9) and before (Fig 1, specimens 10–15) 2010. We cannot
rule out that minimal contrast differences between the OBs
and surrounding saline solution in these two subgroups led
to a systematic error. However, we were aware of the fact
that this systematic error could have been a constant factor,
possibly slightly skewing results in general but not regarding
the relation among the sequences. Fortunately, however, it
was possible to adjust these residual contrast differences by
fine-tuning the window width and level, and therefore we
do not expect the results of MROBV to be significantly influ-
enced. Furthermore, adhesive forces at the pipetting tubes’
lateral walls may have led to minimal impreciseness when
assessing volumes using the WDM. Therefore, the WDM
was performed twice, and the second pass confirmed all
first-pass WDM results. In this context, we should point out
that because of enhanced susceptibility at 3 T, the CISS
in vivo tends to underestimate the cerebrospinal fluid, whereas
the OB soft tissue consequently might be overestimated. We
hypothesize that the minimal imprecision using the WDM
(presumably a systematic overestimation of OB volume)
might have been the reason the mean volumetric results using
the WDM surpassed those using the CISS sequence by 0.5%,
although this relation should in fact have been the reverse.
However, because of the negligible difference of 0.5%, we
think that these overestimation effects of both the CISS
sequence and the WDM cannot be as pronounced as previ-
ously assumed.
Another limitation was that the occurrence of artifacts
(eg, Gibbs ringing artifacts due to sharp changes in intensity
at the pipetting tubes’ surfaces). These artifacts cannot be
avoided completely, and it was not possible to eliminate these
artifacts from the region of interest (25).
Unfortunately, no studies dealing with the subject of MR
volumetry of comparably small nervous structures of the
head exist using 1.5-T or 3-T MRI. The only studies at
1.5 T that seem most likely comparable to this study evaluated
the volume of the vestibulocochlear organ (26,27), whose
volume is at least twice as large (230 mm3) as that of the
OB. The only phantom study evaluating the reproducibility
and reliability of volumetric measurements of olfactory
eloquent structures (including the OB and olfactory tract)
was performed in 1997 by Yousem et al (12). The smallest
Figure 4. Comparison of the configuration of two left olfactory
bulbs, the posterior transitional zone, and the small adjacent part
of the olfactory tract in vitro (top view, scale bar on the left sideamounts to 1.2 cm, edge length of the grid 1 mm). (a) Olfactory
bulb diameter of the equatorial section approximately 4.5 mm (white
arrowhead), transitional zone (expected to be between millimeters 8
and 10; dashed white lines, 2-mm range) continuously merging intoan also conically tapering distal olfactory tract (red arrowheads)
with a diameter of the equatorial section of approximately 2.5 mm
(measured 11-mm distance from anterior end; bulb-to-tract ratio,1.8). (b) Olfactory bulb diameter of the equatorial section approxi-
mately 5 mm (white arrowhead), slightly tapering transitional zone
(expected to be between millimeters 8 and 10; dashed white lines,
2-mm range) merging into a constantly broad (green arrowheads)distal olfactory tract (diameter of the equatorial section approxi-
mately 3.5 mm, bulb-to-tract ratio, 1.4).
Academic Radiology, Vol -, No -, - 2011 IN VITRO 3-T OLFACTORY BULB MR VOLUMETRY
phantom measured had a volume of 1.5 cm3. Repeatedly per-
formed MR volumetric measurements yielded a mean varia-
tion (12%, or about 180 mm3), which, compared to the true
phantom’s volume, clearly surpasses the volume of a normal
genuine OB. However, those three studies neither performed
a systematic comparison of sequences for MR volumetry nor
compared to volumetric results of cadaveric organs.
The potential for inaccuracy or imprecision is large when
one is dealing with volumes as low as that of the OB. The
smaller the anatomic structure becomes, particularly if the
OB volume is reduced because of illness, the more important
it is to minimize the potential error using smaller voxel sizes
(27,28). Additionally, high accuracy and precision play
important roles when comparing sequences for OBV.
Furthermore, the smaller the structure measured with MRI,
the more the range for intraobserver and interobserver
reliability is expected to increase (12). Therefore, to facilitate
OBV for future research purposes, it might be useful to apply
standardized modern automated volumetric methods.
However, until now, it has not been possible to replace the
manual segmentation procedure of this primary olfactory
paleocortical structure by modern automated volumetric
methods such as voxel-based morphometry (29,30), which
is used for assessing secondary olfactory structures.
Conversely, a reason why manual segmentation should still
be used is the problem of exactly defining the OB’s
posterior end (7) using high-resolution 3-T imaging. The
posterior part of the OB does not abruptly end with a sudden
change in diameter between the OB and the olfactory tract, as
was pretended using 2-mm slice thicknesses in coronal planes.
However, OBs’ posterior ends form intraindividual and inter-
individual variably shaped transitional zones (Figs 1a–h in
Bhatnagar et al [31]) showing a smooth transition (Fig 4)
where the OB intermingles with and merges into the olfac-
tory tract (which only might be exactly seen on histologic
sections). Our opinion is that defining the posterior end of
an individual OB cannot be automated or standardized (eg,
in particular in atrophic or hypoplastic OBs) and remains
a task for the assessing radiologist. Therefore, to define the
OB’s posterior end to the best possible extent for MROBV
purposes, we propose that this task preferably be performed
by manual segmentation in multiplanar reconstructions (eg,
Figs 1a–c in Burmeister et al [15]).
Considering our previous results of image quality ratings
(14) and in vivo OBV evaluation (15), we recommend using
a high-resolution CISS sequence at 3 T as the reference
standard sequence in future imaging studies to gain precise
OB volumetric results. Thus, future studies should gain
the following advantages: a precisely rendered correlation
curve of OB volume with age-related olfactory function
(32), an easier categorization of pathologic findings, compa-
rability to other volumetric studies (eg, in case of neurologic
and/or psychological disorders) (33), and the possibility for
OBV to be used as a valid and reliable complementary prog-
nostic tool for the prediction of outcomes in olfactory
disorders (6).
CONCLUSIONS
The in vitro comparison of T1w and T2w sequences for MRI
OBVat 3 Twith WDM confirms previous in vivo results for
the CISS sequence to yield the best results for intraobserver
and interobserver reliability and reproducibility and the high-
est combination of accuracy and precision. On the basis of
direct anatomic correlation, the CISS sequence allows MRI
OBV to gain exact volumetric results. Therefore, OBV can
be implemented in routine clinical practice as a valid and reli-
able method for the evaluation of olfactory disorders.
ACKNOWLEDGMENTS
In particular, our special thanks go to all body donors who
bequeathed their bodies for scientific medical research and
education; without their generous donations, it would not
have been possible for us to do this research. Furthermore,
we would like to thank our photographer, Mrs Astrid Wetzel,
for her excellent cooperation.
REFERENCES
1. Duprez TP, Rombaux P. Imaging the olfactory tract (cranial nerve #1). Eur J
Radiol 2010; 74:288–298.
2. Hoffman HJ, Cruickshanks KJ, Davis B. Perspectives on population-
based epidemiological studies of olfactory and taste impairment. Ann N
Y Acad Sci 2009; 1170:514–530.
3. Vennemann M, Hummel T, Berger K. The association between smoking
and smell and taste impairment in the general population. J Neurol 2008;
255:1121–1126.
4. Hummel T. Therapy of olfactory loss [article in German]. Laryngorhinooto-
logie 2003; 82:552–554.
7
BURMEISTER ET AL Academic Radiology, Vol -, No -, - 2011
5. Rombaux P, Potier H, Bertrand B, et al. Olfactory bulb volume in patients
with sinonasal disease. Am J Rhinol 2008; 22:598–601.
6. Mueller A, Rodewald A, Reden J, et al. Reduced olfactory bulb volume in
post-traumatic and post-infectious olfactory dysfunction. Neuroreport
2005; 16:475–478.
7. Rombaux P, Duprez T, Hummel T. Olfactory bulb volume in the clinical
assessment of olfactory dysfunction. Rhinology 2009; 47:3–9.
8. Rombaux P, Potier H, Markessis E, et al. Olfactory bulb volume and depth
of olfactory sulcus in patients with idiopathic olfactory loss. Eur Arch
Otorhinolaryngol 2010; 267:1551–1556.
9. Gudziol V, Buschhuter D, Abolmaali N, et al. Increasing olfactory bulb
volume due to treatment of chronic rhinosinusitis—a longitudinal study.
Brain 2009; 132:3096–3101.
10. B�edard A, Parent A. Evidence of newly generated neurons in the human
olfactory bulb. Dev Brain Res 2004; 151:159–168.
11. Abolmaali N, Gudziol V, Hummel T. Pathology of the olfactory nerve.
Neuroimaging Clin North Am 2008; 18:233–242.
12. Yousem DM, Geckle RJ, Doty RL, et al. Reproducibility and reliability of
volumetric measurements of olfactory eloquent structures. Acad Radiol
1997; 4:264–269.
13. Schmitt F, Grosu D, Mohr C, et al. 3 Tesla MRI: successful results with
higher field strengths [article in German]. Radiologe 2004; 44:31–47.
14. Burmeister HP, Baltzer PAT, Moslein C, et al. Visual grading characteris-
tics (VGC) analysis of diagnostic image quality for high resolution 3 Tesla
MRI volumetry of the olfactory bulb. Acad Radiol 2011; 18:634–639.
15. Burmeister HP, Baltzer PAT, Moslein C, et al. Reproducibility and repeat-
ability of volumetric measurements for olfactory bulb volumetry: which
method is appropriate? An update using 3 tesla MRI. Acad Radiol 2011;
18:842–849.
16. Rombaux P, Grandin C, Duprez T. How to measure olfactory bulb volume
and olfactory sulcus depth? B-ENT 2009; 5(suppl):53–60.
17. Rombaux P, Mouraux A, Bertrand B, et al. Olfactory function and olfactory
bulb volume in patients with postinfectious olfactory loss. Laryngoscope
2006; 116:436–439.
18. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater
reliability. Psychol Bull 1979; 86:420–428.
19. Landis JR, Koch GG. The measurement of observer agreement for
categorical data. Biometrics 1977; 33:159–174.
8
20. Crawford SB, Kosinski AS, Lin H-M, et al. Computer programs for the
concordance correlation coefficient. Comput Methods Programs Biomed
2007; 88:62–74.
21. Lin LI. A concordance correlation coefficient to evaluate reproducibility.
Biometrics 1989; 45:255–268.
22. Bland JM, Altman DG. Statistical methods for assessing agreement
between two methods of clinical measurement. Lancet 1986; 1:307–310.
23. AickinM, Gensler H. Adjusting for multiple testing when reporting research
results: the Bonferroni vs Holm methods. Am J Public Health 1996; 86:
726–728.
24. Shepherd TM, Thelwall PE, Stanisz GJ, et al. Aldehyde fixative solutions
alter the water relaxation and diffusion properties of nervous tissue.
Magn Reson Med 2009; 62:26–34.
25. DietrichO, ReiserMF, Schoenberg SO. Artifacts in 3-TMRI: physical back-
ground and reduction strategies. Eur J Radiol 2008; 65:29–35.
26. Kendi TK, Arikan OK, Koc C. Volume of components of labyrinth: magnetic
resonance imaging study. Otol Neurotol 2005; 26:778–781.
27. Melhem ER, Shakir H, Bakthavachalam S, et al. Inner ear volumetric
measurements using high-resolution 3D T2-weighted fast spin-echo MR
imaging: initial experience in healthy subjects. AJNR Am J Neuroradiol
1998; 19:1819–1822.
28. Blatter DD, Bigler ED, Gale SD, et al. Quantitative volumetric analysis of
brain MR: normative database spanning 5 decades of life. AJNR Am J
Neuroradiol 1995; 16:241–251.
29. Bitter T, Bruderle J, Gudziol H, et al. Gray and white matter reduction in hy-
posmic subjects—a voxel-based morphometry study. Brain Res 2010;
1347:42–47.
30. Bitter T, Gudziol H, Burmeister HP, et al. Anosmia leads to a loss of gray
matter in cortical brain areas. Chem Senses 2010; 35:407–415.
31. Bhatnagar KP, Kennedy RC, Baron G, et al. Number of mitral cells and the
bulb volume in the aging human olfactory bulb: a quantitative morpholog-
ical study. Anat Rec 1987; 218:73–87.
32. Yousem DM, Geckle RJ, Bilker WB, et al. Olfactory bulb and tract and
temporal lobe volumes. Normative data across decades. Ann N Y Acad
Sci 1998; 855:546–555.
33. Negoias S, Croy I, Gerber J, et al. Reduced olfactory bulb volume and
olfactory sensitivity in patients with acute major depression. Neuroscience
2010; 169:415–421.