A New Method of Reconstructing the Human

Laryngeal Architecture Using Micro-MRI

*,‡Ting Chen, †Ann M. Chodara, †Alicia J. Sprecher, §Fang Fang, §Wen Song, †Chao Tao, and *,†Jack J. Jiang,
*Shanghai zFuzhou xNanjing, China, yMadison, Wisconsin

Summary: Objectives/Hypothesis. A realistic three-dimentional (3D) model of the larynx could be of value for
the understanding of normal laryngeal motion and for studying pathological changes as well as in modeling potential
therapy outcomes. The objective of this research was to present a new method of creating a computer model of the
human larynx using data obtained through micro-magnetic resonance imaging (MRI) scanning.
Study Design. Prospective study.
Methods. A 7-T micro-MRI scanner was used to scan an excised larynx obtained postmortem from a 68-year-old
woman. MRI data was manually segmented and compiled into 3D images using Mimics12.1 reconstruction software.
Measurements of the reconstructed structures were also calculated using Mimics12.1.
Results. The 3D laryngeal model contained the thyroid, cricoid, and arytenoid cartilages. Nearly all of the intrinsic
muscles could be segmented. Although the cricoarytenoid joint could be visualized, the features of the cricothyroid joint
were not clear. Muscle and cartilage volumes and surface areas were calculated from the 3D model.
Conclusions. The combination of MRI and 3D reconstruction generates promising results in the hopes of creating
a highly realistic and detailed model of the human larynx.
Key Words: Micro-MRI–3D reconstruction–Human larynx–Model.

INTRODUCTION data, yielding voxels with side lengths of 0.35 mm or

A better understanding of the anatomical and physical charac-
teristics of the larynx is required to create realistic models with
true potential for research and clinical implementation. These
features include the shape and relative positions of the laryngeal
cartilages, intrinsic muscles, and ligaments. There are numer-
ous studies that report original data on laryngeal anatomy
obtained through gross anatomical examination of multiple
specimens.1–4 However, it is difficult to combine the data
from such examinations into a complete three-dimensional
(3D) model. Histological slicing has been used in other organ
systems to create 3D reconstructions after two-dimensional
(2D) segmentation5; however, this method is problematic in la-
ryngeal modeling because different laryngeal tissues have dif-
ferent deformation ratios during the slicing process,
obscuring the true anatomical shape. Computed tomography
(CT) has also been used for laryngeal imaging and reconstruc-
tion6–8 and in surgical planning,9 but CT is not able to clearly
reflect soft tissue. Models created using magnetic resonance im-
aging (MRI) are more successful because of its superior ability
to image soft tissue. Selbie et al.10 used a 2-T MRI to collect
Accepted for publication March 29, 2011.
Financial Disclosure Information: This research was supported by the Sina-America
Voice Research Center Project from the Ear, Eye, Nose and Throat Hospital of Fudan
University, Shanghai, China.
From the *Department of Otolaryngology, Eye, Ear, Nose and Throat Hospital, Fudan
University, Shanghai, China; yDepartment of Surgery, Division of Otolaryngology—Head
and Neck Surgery, University of Wisconsin School of Medicine and Public Health,
Madison, Wisconsin; zDepartment of Otolaryngology—Head and Neck Surgery, Fujian
Provincial Clinical College of Fujian Medical University, Fuzhou, China; and the
xKey Laboratory of Molecular and Functional Imaging, School of Medicine, Southeast
University, Nanjing, China.
Address correspondence and reprint requests to Jack J. Jiang, Department of Surgery,
Division of Otolaryngology, University of Wisconsin School of Medicine and Public
Health, 5745 Medical Sciences Center, 1300 University Avenue, Madison 53706,
Wisconsin. E-mail: jjjiang@wisc.edu
Journal of Voice, Vol. 26, No. 5, pp. 555-562
0892-1997/$36.00
Ó 2012 The Voice Foundation
doi:10.1016/j.jvoice.2011.03.012
0.43 mm. Using the 2D images for segmentation, they created
a 3D model using volume rendering; however, the field strength
and resolution were low, which led to inaccurate boundaries
between both soft tissues and cartilages and within soft tissues
themselves. Muscle segmentation was difficult and fiber num-
bers and directions could not be collected. Moreover, the clarity
of the cartilages, specifically the cricoid cartilage, decreased
near the edges of the scan.10
The use of high-resolution MRI can generate a more accurate
understanding of the volume, contours, and location of each
component of the laryngeal framework. Ultrahigh magnetic
field strengths (above 7 T) can provide submillimeter imaging
detail,11 approaching a spatial resolution of 0.04 mm and pixel
resolution of 15 mm at fields above 21 T.12 The signal-to-noise
ratio increases almost linearly with the magnetic field13 and can
be used to improve resolution or decrease scan time.14,15
Ultrahigh-field MRI has great potential in an organ such as
the larynx with its small size and involved structures. In this
study, we used a 7-T micro-MRI scanner to obtain high-
resolution scans of a freshly excised human larynx. Using
3D image processing and editing software, we generated a geo-
metric 3D model for use in laryngeal biomechanical analysis.
MATERIALS AND METHODS
Specimen
The Institutional Review Board of the Eye, Ear, Nose and
Throat Hospital affiliated with the Fudan University in
Shanghai, China approved and oversaw the protocol of this
investigation. A normal healthy larynx from a 68-year-old
woman was carefully dissected, postmortem. Immediately after
death, the cadaver was frozen and the tissue was thawed before
excising the larynx. Heavy nylon sutures were used to suspend
the larynx within an acrylic cylinder for scanning (Figure 1).
556
FIGURE 1. Posterior (A), anterior (B), and lateral (C) views of the excised human larynx after dissection. All cartilages and intrinsic muscles were
maintained intact.
Micro-MRI scanning
Scanning was completed with a 7.0-T micro-MRI scanner
(Bruker, Rheinstetten, Germany) with a 16-cm horizontal bore.
A volume coil of 61 mm diameter was used. A TurboRARE 3D
T2-weighted scanning sequence was used for axial scanning. In
previous studies, we noted severe ghosting near the head and
tail of the larynx because of the decreased magnetic signal mov-
ing away from the isocenter. To remedy this, we scanned the
specimen twice, once for the top half and once for the bottom
half. Each scan used an echo time of 43.0 ms, a repetition
time of 1000 ms, a flip angle of 180.0 , and a field of view of
4.50 3 4.50 3 3.20 cm with a slice thickness of 0.25 mm, with-
out intervals. The total imaging time for each scan was 2 hours
16 minutes 32 seconds with a number of excitations of 4 and
a matrix of 256 3 256 3 128. The final resolution was
175 3 175 3 250 mm per voxel. We combined the two sets of
images after scanning using Mimics12.1 software (The Materi-
alise Group, Leuven, Belgium).
3D reconstruction
The micro-MRI images were processed with Mimics12.1 to cre-
ate a 3D model of the larynx that included the thyroid, cricoid,
and arytenoid cartilages, internal laryngeal muscles, and the
FIGURE 2. MRI images from the sagittal (A), coronal (B), and transverse (C) planes, showing the tracing of cartilage boundaries for compilation
and 3D reconstruction. Outlined in yellow is the thyroid cartilage, in blue the arytenoids, and in lavender the cricoid cartilage.
Journal of Voice, Vol. 26, No. 5, 2012
cricoarytenoid (CA) joint. Using manual segmentation, a 1:1 re-
construction was generated. The framework of the larynx was
constructed using surface rendering and Gaussian smoothing.
RESULTS
In this study, we found that micro-MRI of the laryngeal frame-
work provides an excellent anatomical depiction of its architec-
ture. The regional muscles, laryngeal cartilages, and CA joint
were described. The interstitial fluids and spaces were useful in
determining the boundaries of the laryngeal components. We
focused on the structural components known to be important in
phonation and potentially of use in future modeling experiments.
Rigid body reconstruction
Cartilages. From the sagittal, coronal, and transverse views,
we can clearly observe the features and spatial arrangements
of the six laryngeal cartilages: the thyroid, cricoid, and
epiglottis, and the paired arytenoids, corniculates, and cunei-
forms. As the specimen is from an older woman (68 years
of age), cartilaginous ossification resulted in areas of the
cartilage that appear black in the MRI images. The first step
in generating a 3D model consisted of manually tracing the
contours of the cartilages as shown in Figure 2. From these
Ting Chen, et al Micro-MRI and Modeling of the Larynx 557

FIGURE 3. 3D reconstruction of the thyroid (A), cricoid (B), and arytenoid (C) cartilages and their composite framework (D).

traces, we produced 3D reconstructions of the thyroid vex surface of the cricoid (B). Minimal joint space is visible in
cartilage, the arytenoids, and the cricoid cartilage and the MRI image; however, the capsular ligament (C) can be
combined these to create composite images showing the rela- seen to run from the posterior end of the joint through the inner
tive positions of these cartilages (Figure 3). Cartilage volumes surface connecting with the conus elasticus. By displaying the
and surface areas were calculated with Mimics12.1 (Table 1). cricoid and one arytenoid cartilage, we gain an impression of
To simplify our model and because of their limited involve- the joint, which is improved by generating a 3D model of the
ment in phonation, the epiglottis and the corniculate and
cuneiform cartilages were not reconstructed.
Joints. We examined the two of the major joints in the larynx:
the CA and cricothyroid joints. The two articulated cartilages of
the CA joint are shown in Figure 4. Evident are the concave
articular surface of the arytenoid (A) and the corresponding con-

TABLE 1.
Volume and Surface Area of Laryngeal Cartilages and
Conus Elasticus as Calculated by Mimics12.1
Surface Area
Tissues Volume (mm3) (mm2)
Thyroid cartilage 4196.97 3598.23
Cricoid cartilage 3513.53 2134.95
Arytenoid 206.62 (Left)/ 274.93 (Left)/ FIGURE 4. The CA joint shown from a sagittal view. Arrows indi-
cartilage 205.29 (Right) 272.60 (Right)
cate the articular surfaces of the arytenoid (A) and cricoid cartilages
Conus elasticus 2258.19 3254.15
(B). The capsular ligament (C) is shown in a lighter color.
558 Journal of Voice, Vol. 26, No. 5, 2012

FIGURE 5. The relationship of the CA joint and the conus elasticus. The conus elasticus travels around the medial and posterior borders of the joint.

capsular ligament (Figure 5). Unlike the CA joint, the articular
surfaces, ligaments, and capsule of the cricothyroid joint are
not readily visible on the MRI images and could not be
reconstructed.
Ligaments. The ligaments of the larynx set the boundaries of
the joints’ movement. With the high resolution provided by
micro-MRI, the primary outline of the elastic membrane is
well visualized and easily differentiated from the underlying
muscle. The conus elasticus can be traced as it connects the thy-
roid and cricoid cartilages. The lateral portion of the ligament
arises from the inner surface of the cricoid arch and extends up-
ward, where it is contiguous with the inferior margin of the
vocal ligament (Figure 6). The ligaments of the CA joint can
be seen to join with the elastic membrane of the larynx, which
extends downward through the cricoid cartilage. Loose liga-
ments such as the hyoepiglottic ligament were not well defined
because of the higher concentration of fat near these ligaments.
Control component reconstruction
Intrinsic muscles. Our MRI results showed excellent muscle
boundaries and gave an estimate of fiber direction when that di-
rection was parallel to the sagittal, coronal, or transverse axis.

FIGURE 6. In (A) arrows indicate the conus elasticus from the sagittal (left), coronal (middle), and a laterally shifted sagittal (right) view. (B) and
(C) show the conus elasticus (teal) in relation to the cartilage framework, the LP (white) and the thyroarytenoid muscle (light blue).
Ting Chen, et al Micro-MRI and Modeling of the Larynx
FIGURE 7. Segmentation of the intrinsic laryngeal muscles and LP on the right side. Evident are distinct boundaries between the posterior
CA (brown), lateral CA (salmon), cricothyroid (olive) and thyroarytenoid (light blue) muscles as well as the cartilages.
Figure 7 shows a series of images with laryngeal muscles high-
lighted. Of note is the ease with which these muscles can be dif-
ferentiated and the muscle fiber structure evident in many of the
images. After muscles were traced on the 2D images,
Mimics12.1 was used to compile the 3D reconstructions shown
in Figure 8. Muscle groups are clearly defined, and manipula-
tion of these groups and the underlying cartilages is simple.
The model also provided volume measurements for each of
the reconstructed muscles, shown in Table 2.
Oscillation element reconstruction
Vocal fold and lamina propria. The micro-MRI images
showed a clear boundary between the lamina propria (LP)
and the vocalis muscle, which could be easily segmented
(Figure 9). The layers of the LP could not be identified; it
appears homogenous in the MRI images. The LP volume was
calculated and reported in Table 2.
DISCUSSION
Computer models are useful in exploring and estimating the per-
formance of systems too complex for analytical solutions, with
one example being the dynamics of the larynx. Previously, com-
puter models of laryngeal dynamics have been lumped element
559
models,16,17 which are created, mathematically speaking, by
a simplified reduction of the state space of the system to
a finite number with no scientific visualization. Finite element
analysis, which uses simple geometry to approximately
simulate laryngeal posturing and biomechanics, has been used
to develop laryngeal models.18,19 A limitation of these studies
is that data were collected from invasive experiments
involving individual muscles and layers and thus do not
reflect the true laryngeal architecture. Of course, the more we
know about the architecture of the larynx, the more reliable
and valuable a computer model of the larynx will be for
study. Modern imaging technology in combination with
3D reconstruction provides the means to produce a highly
realistic dynamic model of the human larynx.20 Models created
using MRI have assisted in medical education,21 demonstrated
interactions between laryngeal elements,22 and tested model
predictions by scaling an individual larynx to a reference frame-
work.10 Accurate models of the larynx show potential not just
for research but for eventual clinical application in evaluating
laryngeal pathologies and treatments.
In this study, we evaluated the feasibility of creating
a computer model of the human larynx. The use of micro-MRI
provides true architectural information about the rigid body—
560 Journal of Voice, Vol. 26, No. 5, 2012

FIGURE 8. 3D composite images of the intrinsic laryngeal muscles showing their spatial location within the laryngeal framework (same colors as
indicated in Figure 7).

cartilages, joints, and ligaments; the control elements—the images.10 High-field MRI overcomes this limitation and appears
intrinsic muscles; and the oscillation elements—the vocal folds. to be an ideal tool for collecting data to complete a comprehen-
This technology allows visualization of structures not easily sive 3D reconstruction. The 7-T scanner allows a resolution of
evaluated using CT or ultrasound. As compared with traditional 175 3 175 3 250 mm voxels that results in clear enough images
gross anatomical observation or histological slicing, MRI pro- to complete excellent 2D segmentation and 3D reconstruction,
vides more accurate spatial information that can be translated which provide clear geometric data and boundary limitations
into a 3D model. In previous studies, the small details of the lar- for computer modeling.
ynx could not be completely described by low-resolution MRI As for the joint anatomy, the CA joint has been described as
a saddle-shaped synovial joint whose prominent capsule
ligament and articulated surfaces allow for three primary move-
ments: lateral-medial sliding, anteroposterior rocking, and
TABLE 2.
Volumes of Intrinsic Laryngeal Muscles and LP as
rotation. In our data, the conus elasticus wraps around the
Calculated by Mimics12.1 posterior and medial borders of the joint, allowing the lateral-
medial sliding and anteroposterior rocking, but limiting rotation
Left Volume Right Volume of the arytenoids, which is in accordance with previous
Tissues (mm3) (mm3) research.23 It is valuable to know the boundaries of the CA joint
Thyroarytenoid 862.75 851.56 for an accurate laryngeal computer model.
muscle Although the CT joint could not be well visualized in this
Posterior CA muscle 605.10 599.02 study, previous research has indicated that this joint can be di-
Cricothyroid muscle 586.11 566.61 vided into type A, with a well-defined facet, tight capsule, and
Lateral CA muscle 192.16 188.45
ligaments; type B, with no definable facet, only joined by
LP 95.65 92.34
loose soft connective tissue; and type C, with soft tissue facets
Ting Chen, et al Micro-MRI and Modeling of the Larynx
FIGURE 9. In (A, sagittal view) and (B, coronal view), arrows indicate the vocal fold LP. The LP was reconstructed in (C) and (D), which show the
relation of this layer (white) to the lateral cricoarytenoid (salmon) and thyroarytenoid muscles (light blue).
and a tiny thin capsular connection.24 This sample appears to
be of type B, which would explain the limited visualization of
the joint. From the images, the CT joint is evidently subject to
more freedom of movement as compared with the CA joint.
A higher magnetic field reduces the scanning time needed to
record high-resolution data; however, for the extremely high-
resolution images recorded in this study, the use of in vivo sam-
ples is currently prevented by long scanning times, the small
core size and the impact of breathing, swallowing, and other
random movements. Use of an excised larynx provides accurate
architectural data while conveniently avoiding these issues. The
larynx used in this study came from a 68-year-old woman.
Despite the ossification associated with the individual’s ad-
vanced age, cartilage contours remained well defined. Success-
ful 3D reconstruction provides the possibility of computerized
study of variations in laryngeal dynamics because of age, sex,
and pathology and even among individuals.
Additionally, a limitation observed in this study was in the vi-
sualization of the LP. Although the entire LP was easily seg-
mented from the vocalis muscle, divisions within the LP were
not evident. The division of the LP into three layers is based
on the distribution and ratio of elastic and collagen fibers.25
At 7 T, the LP appears to be homogenous and cannot be divided
into layers. In a recent study,26 an 11.7-T micro-MRI was used
to observe vocal fold tissues in excised ferret and canine laryn-
ges. Even with this strong magnetic field and a voxel side length
of 39 mm in the principal plane, the LP layers were not visible.
Although there may be subtle differences in the LP structure be-
tween species, it remains to be seen whether the layers can be
differentiated even at a higher resolution or magnetic field
strength or using a different scanning sequence.
561
Micro-MRI scanning data supply high enough resolution of
laryngeal components as to permit excellent 2D segmentation
for outstanding anatomic architecture reconstruction. More im-
portantly, when combined with other MRI technology such as
diffusion tensor imaging and tractography, it can noninvasively
provide biomechanical data such as muscle motor unit anisot-
ropy, tracking of muscle and nerve fiber paths in space, motor
unit numbers, and so on.27,28 These technologies have the
potential to greatly change the traditional methodology in
biological computer modeling and supply a whole new style
of laryngeal dynamic computer simulation. Anatomic
architecture reconstruction as achieved in this study is the
first step in this novel method.
CONCLUSION
The present study proposes a new method of creating
a computer model of the larynx. High-resolution data from
micro-MRI noninvasively provide a visual representation of
the real anatomical structure and spatial relationship of laryn-
geal components. It supplies true geometric information and
boundary limitations for the successful completion of a compa-
rable dynamic model for use in physical and pathophysiological
research. In the future, such a model may also be useful in the
evaluation of possible therapies for laryngeal disorders.
REFERENCES
1. Tayama N, Chan RW, Kaga K, Titze IR. Geometric characterization of the
laryngeal cartilage framework for the purpose of biomechanical modeling.
Ann Otol Rhinol Laryngol. 2001;110:1154–1161.
562
2. Sprinzl GM, Eckel HE, Sittel C, Pototschnig C, Koebke J. Morphometric
measurements of the cartilaginous larynx: An anatomic correlate of
laryngeal surgery. Head Neck. 1999;21:743–750.
3. Zrunek M, Happak W, Hermann M, Streinzer W. Comparative anatomy of
human and sheep laryngeal skeleton. Acta Otolaryngol. 1988;105:
155–162.
4. Randestad A, Lindholm CE, Fabian P. Dimensions of the cricoid cartilage
and the trachea. Laryngoscope. 2000;110:1957–1961.
5. Yelnik J, Bardinet E, Dormont D, et al. A three-dimensional, histological
and deformable atlas of the human basal ganglia. I. Atlas construction
based on immunohistochemical and MRI data. Neuroimage. 2007;34:
618–638.
6. Wang D, Zhang W, Xiong M, Xu J. Laryngeal and hypopharyngeal
carcinoma: comparison of helical CT multiplanar reformation, three-
dimensional reconstruction and virtual laryngoscopy. Chin Med J (Engl).
2001;114:54–58.
7. Zeiberg AS, Silverman PM, Sessions RB, Troost TR, Davros WJ,
Zeman RK. Helical (spiral) CT of the upper airway with three-
dimensional imaging: technique and clinical assessment. AJR Am J
Roentgenol. 1996;166:293–299.
8. Silverman PM, Zeiberg AS, Sessions RB, Troost TR, Zeman RK. Three-
dimensional imaging of the hypopharynx and larynx by means of helical
(spiral) computed tomography. Comparison of radiological and otolaryngo-
logical evaluation. Ann Otol Rhinol Laryngol. 1995;104:425–431.
9. Raappana A, Koivukangas J, Piril€a T. 3D modeling-based surgical
planning in transsphenoidal pituitary surgery—preliminary results. Acta
Otolaryngol. 2008;128:1011–1018.
10. Selbie WS, Gewalt SL, Ludlow CL. Developing an anatomical model of the
human laryngeal cartilages from magnetic resonance imaging. J Acoust Soc
Am. 2002;112(3 Pt 1):1077–1090.
11. Regatte RR, Schweitzer ME. Ultra-high-field MRI of the musculoskeletal
system at 7.0T. J Magn Reson Imaging. 2007;25:262–269.
12. Sharma R. Microimaging of hairless rat skin by magnetic resonance at
900 MHz. Magn Reson Imaging. 2009;27:240–255.
13. Collins CM, Smith MB. Calculations of B1 distribution, SNR, and SAR for
a surface coil adjacent to an anatomically-accurate human body model.
Magn Reson Med. 2001;45:692–699.
14. Uematsu H, Takahashi M, Dougherty L, Hatabu H. High field body MR
imaging: preliminary experiences. Clin Imaging. 2004;28:159–162.
Journal of Voice, Vol. 26, No. 5, 2012
15. DiFrancesco MW, Rasmussen JM, Yuan W, et al. Comparison of SNR and
CNR for in vivo mouse brain imaging at 3 and 7T using well matched
scanner configurations. Med Phys. 2008;35:3972–3978.
16. Wong D, Ito MR, Cox NB, Titze IR. Observation of perturbations in
a lumped-element model of the vocal folds with application to some
pathological cases. J Acoust Soc Am. 1991;89:383–394.
17. Titze IR, Story BH. Rules for controlling low-dimensional vocal fold
models with muscle activation. J Acoust Soc Am. 2002;112(3 Pt 1):
1064–1067.
18. Tao C, Jiang JJ. Mechanical stress during phonation in a self-oscillating
finite-element vocal fold model. J Biomech. 2007;40:2191–2198.
19. Jiang JJ, Diaz CE, Hanson DG. Finite element modeling of vocal fold vi-
bration in normal phonation and hyperfunctional dysphonia: implications
for the pathogenesis of vocal nodules. Ann Otol Rhinol Laryngol. 1998;
107:603–610.
20. Tao C, Jiang JJ. Anterior-posterior biphonation in a finite element model of
vocal fold vibration. J Acoust Soc Am. 2006;120:1570–1577.
21. Hu A, Wilson T, Ladak H, Haase P, Fung K. Three-dimensional education
computer model of the larynx: voicing a new direction. Arch Otolaryngol
Head Neck Surg. 2009;135:677–681.
22. Rubin JS, Summers P, Harris T. Visualization of the human larynx: a three-
dimensional computer modeling tool. Auris Nasus Larynx. 1998;25:
303–308.
23. Fink BR, Demarest RJ. Laryngeal Biomechanics. Cambridge, UK: Harvard
University Press; 1978. 45–63.
24. Hammer GP, Windisch G, Prodinger PM, Anderhuber F, Friedrich G. The
cricothyroid joint—functional aspects with regard to different types of its
structure. J Voice. 2010;24:140–145.
25. Madruga de Melo EC, Lemos M, Arag~ao Ximenes Filho J, Sennes LU,
Nascimento Saldiva PH, Tsuji DH. Distribution of collagen in the lamina
propria of the human vocal fold. Laryngoscope. 2003;113:2187–2191.
26. Herrera VL, Viereck JC, Lopez-Guerra G, et al. 11.7 Tesla magnetic reso-
nance microimaging of laryngeal tissue architecture. Laryngoscope. 2009;
119:2187–2194.
27. Heemskerk AM, Sinha TK, Wilson KJ, Ding Z, Damon BM. Quantitative
assessment of DTI-based muscle fiber tracking and optimal tracking
parameters. Magn Reson Med. 2009;61:467–472.
28. Mori S, van Zijl PC. Fiber tracking: principles and strategies—a technical
review. NMR Biomed. 2002;15:468–480.