Journal of Medical and Biological Engineering, 30(1): 57-63
57
A Virtual Reality Ear Ossicle Surgery Simulator Using
Three-dimensional Computer Tomography
Ming-Shium Hsieh1,2
Fei-Peng Lee3
Ming-Dar Tsai4,*
1
School of Medicine, Taipei Medical University, Taipei 110, Taiwan, ROC
Department of Orthopedics Research Center, Taipei Medical University Hospital, Taipei 110, Taiwan, ROC
3
Department of Otolaryngology, Taipei Medical University Hospital, Taipei 110, Taiwan, ROC
4
Institute of Information and Computer Engineering, Chung Yuan Christian University, Chungli 320, Taiwan, ROC
2
Received 27 Apr 2009; Accepted 26 Oct 2009
Abstract
This paper describes a virtual reality simulator for middle ear ossicle surgery. An automatic segmentation method has
been developed to segment tiny ear ossicles at transverse computer tomography (CT) slices. Surgeons can therefore easily
observe the three-dimensional (3D) geometry of the segmented ossicles by the volume reconstruction and the spatial
relation with the temporal bone to diagnose middle ear disease. Surgeons can use a virtual reality round cutting bur to cut
the temporal bone for opening the 3D tympanic cavity, and a virtual reality round polishing bur to cut for separating ear
ossicles from other bones or to polish the adhered tissue sclerosis on the ear ossicles. These burring simulations can achieve
real-time visual and haptic responses based on our reported volume manipulation methods. A technique is developed to
judge if a separation among the ossicles or from the temporal bone occurs during the burring simulations so that
repositioning simulations can be followed to align the ear ossicles. A simulation example of a real ear ossicle surgery
demonstrates these simulation functions work well even for tiny ossicles and thus shows the effectiveness of the surgical
simulator to rehearse the surgical procedures, confirm surgical plans and train interns and students.
Keywords: Middle ear surgery, Virtual reality surgery simulation, Ear ossicle, Volume visualization
1. Introduction
Surgeries for recovering functions of middle ear ossicles
(malleus, incus and stapes) through bone ossicular replacements,
alignment or cleaning are advanced surgical procedures for
treating conductive hearing loss. However, the tiny sizes of these
ossicles bring high surgical failure rates [1-5]. The X-ray based
cephalogram is a standard procedure to evaluate geometry of the
middle ear (tympanic) cavity and the inside ear ossicles, diagnose
diseases and manage surgery. However, projection errors occur
in using X-rays [6], which cannot provide correct spatial relations
among the ear ossicles and with the temporal bone to make
correct surgical plans. Meanwhile, computer tomography (CT)
slices can provide interior information and avoid projection
errors, and thus are considered as an excellent way to evaluate
the tiny ear ossicles [7-10]. Three-dimensional (3D)
reconstruction from a volume constituted by CT slices can be
used to visualize pathologies of the ear ossicles [11] or simulate
cutting operations on the 3D temporal bone [12,13].
However, the ear ossicles are too small to be distinguished
* Corresponding author: Ming-Dar Tsai
Tel: +886-3-2654718; Fax: +886-3-2654799
E-mail: [email protected]
from the neighboring temporal bone for visualization and then
surgical simulation. This paper therefore proposes an automatic
segmentation method to recognize the ear ossicles in the
tympanic cavity on transverse CT slices. Using segmented
ossicular areas on the two-dimensional (2D) CT slices, 3D ear
ossicles can be highlighted from the temporal bone during the
volume reconstruction. A surgeon can then easily zoom in on
the tympanic cavity to simulate opening the tympanic cavity,
polishing out tissue sclerosis on the ear ossicles and
repositioning to align these ossicles. This paper also develops a
volume-based simulation method that uses real-time visual and
haptic burring simulation functions to open the tympanic cavity
and polish the ear ossicles, checks whether the ossicles are
separate or separated from the temporal bone and uses
repositioning functions to align ossicles. In this preliminary
study, we evaluated the usefulness of our method and the
prototype system with a middle ear disease case.
2. Subjects and methods
2.1 Middle ear ossicle segmentation on transverse CT slices
The process of finding the middle ear ossicles on a
transverse slice is described as follows.
J. Med. Biol. Eng., Vol. 30. No. 1 2010
58
(1) Determine the two region of interest (ROI) areas to
include the left and right ears, respectively, on every
transverse CT slice. The left (or right) ROI is set near the
left (or right) ear to include the left (or right) tympanic
cavity. Because the field of view (FOV) of a transverse CT
slice for evaluating ossicular diseases usually includes
both the ears, the ROI is calculated using the horizontal
and anteroposterior widths of the skull on the slice. As
illustrated in Fig. 1, the horizontal and anteroposterior
positions (Pr, Pl and Pap in Fig. 1) and widths (hr and hap)
of the ROI are determined using the bone widths (sr and
sap) along the horizontal and anteroposterior directions on
the slice and the middle points of the two widths. To avoid
missing the tympanic cavity, the ROI width is set at
several times the general tympanic cavity width. The nose
or brain cavities can then be excluded from the
segmentation computation; however, the tympanic cavity
together with neighboring cavities or canals such as the
auricular canal may be included in the ROI.
s ap
sh
bone ranges is then not processed (such as the cavity
constituted by the pairs of gi and hi shown in Fig. 2(A)).
(4) Determine the center (C in Fig. 2(C)) of the cavity by
averaging the position of all the pixels inside this cavity and
use a vector starting from C to find the furthest inside bone
range (b2-c2 in Fig. 2(C)) from the continuous scanlines
representing inside bones in the cavity. The furthest range of
inside bones in a cavity is then calculated using the distance
of the cavity center (C) to every endpoint (e.g., b2 or c2) of
each inside bone range. The range with the endpoint of the
largest distance to the center is the furthest range. The
pixels on the furthest range are set as boundary pixels (e.g.,
pixels between b2 and c2) of the inside bones that is used to
distinguish the inside bones from the cavity. For example,
ear ossicles if connecting with the cavity wall by a bone bar
can be distinguished by this procedure.
(5) Determine separate inside bones in a cavity. Similar to the
third procedure, the middle point (ni in Fig. 2(D)) of each
range representing inside bones (such as ear ossicles) in a
cavity is used as a seed for flooding a separate inside bone
in the cavity. After this procedure, all ranges of an inside
bone in a cavity can be recognized as a continuous bone to
be highlighted even if only one range of this bone on the
scanlines is in the values pre-defined for inside bones.
(A)
pr
pl
p ap
th
th
t ap
(A)
(B)
(B)
Figure 1. Determination of two ROIs including the two tympanic
cavities on a transverse slice. Bold line: a transverse slice.
(A) Anteroposterior view. (B) Horizontal view.
(2) Determine the intersections of cavity boundaries using
continuous horizontal scanlines inside the ROI, as shown in
Fig. 2. A bone threshold is used to determine the cavity
boundaries. Two neighboring intersections with in-between
null (with gray-level values under the bone threshold) pixels
form a cavity range, such as the range between ai and bi,
and between ci and di, respectively. Neighboring pairs of
cavity ranges (such as the ranges ai-bi and ci-di) can be
considered as being in the same cavity (such as ai-di) with
an inside bone range (such as bi-ci). The ranges that are
possible to be a tympanic cavity and inside ossicles are
pre-defined. Too large or too small ranges of cavities and
inside bones are excluded.
(3) Search all pixels inside each cavity by using the middle
point (mi in Fig. 2(B)) of the cavity range (ai and di) on a
horizontal line as a seed to flood the cavity [14]. A seed and
flood algorithm can traverse all pixels inside a closed cavity
and consumes only a little time because of the small area of
the cavity. Multiple middle points may be obtained from
continuous horizontal lines that traverse the same cavity. A
cavity constituted by the horizontal lines including no inside
(C)
(D)
Figure 2. Image-matching of middle ear ossicles in a ROI of a
transverse slice. (A) Intersections of horizontal scanlines
with boundaries of the tympanic cavity and ear ossicles. (B)
Determination of the tympanic cavity by the seed and flood
algorithm. (C) Determination of the boundary pixels between
the tympanic cavity wall and ossicles (D) Determination of
separate ossicles by the seed and flood algorithm.
2.2 Cutting and topologic simulations for temporal bone and
ear ossicles
This ossicle surgery simulator combines the simulation
functions of our several previous works. First, CT transverse
slices with segmented ear ossicles are used to constitute a
volume in which every voxel is extended to six pairs of
distance-levels and face-flags. A face-flag indicates whether a
Ear Ossicle Surgery Simulator
voxel face is boundary (shared by two voxels of different
tissues) and thus the associated distance-level represents a
surface vertex between the two voxels. Therefore, six
distance-levels represent 6 possible tissue vertices along the
three volume-axis-parallel lines passing through the voxel
center. Our previous work manipulated the voxel
distance-levels to represent burred tissue changes on the
volume-axis-parallel lines and thus to simulate burring
operations with visual and haptic responses [15].
Then, whether a burred ossicle is separated from other
ossicles or the temporal bone by the burring simulations is
calculated as follows. Figure 3 indicates the changes of the
voxel faces during the burring simulations. The burs used in
the ear ossicle surgery are near or smaller than the voxel
width. The system checks whether a bone voxel center is
overlapped by the cutting tool. If it is, this voxel is nullified
(the voxel tissue type is set as air) and the face-flags indicate
the faces between this voxel with the neighboring bone voxels
are changed to indicate these face are boundary, as illustrated
in Fig. 3(B). The burred ossicle is separated from other
ossicles or the temporal bone if the bur traverses from
immersing into bone voxels and then leaving into a null voxel
(as V in Fig. 3(C)). Then, a seed is assigned to each side of
the last voxel (as W). The separate ossicles can be then
manipulated (deleted or repositioned) independently by the
two seeds (S1 and S2).
This simulator can manipulate (delete or reposition) a
separate bone through a fast 3D seed and flood algorithm [14].
A deletion function traverses to nullify all voxels of the
separate bone (set the tissue type of traversed voxels as null).
Meanwhile, a reposition function traverses to store all voxels
of the separate bone in a stack, deletes the bone and pops the
stored data into the voxels of the new position in the reverse
order as that in which they were stacked [16]. The system
uses the marching cubes method to reconstruct triangulated
bone surfaces [17]. However, the reconstruction for bone
surfaces in a whole volume consumes time, making it difficult
in real time. Therefore, the system uses a dynamical cube data
structure to record reconstructed bone triangles inside every
voxel. By manipulating the cube data structure, 3D
reconstruction can be implemented not for the whole volume
but only for the operated (burred) voxels to achieve real-time
visual responses [18]. This 3D reconstruction method can
reconstruct simultaneously surfaces of different tissues or
structures (the temporal bone or ossicles) to assign these
tissues or structures with different colors. Therefore, the
recognized ossicles can be highlighted by being assigned a
different color from other bones.
3. Implementation
Currently, our prototype system is implemented on a
general PC with an IntelCore 2 Duo E8500 (3.16 GHZ) CPU, 3
Gbytes of main memory and a NVIDIA GeForce9600 GT
graphics card. The software uses the OpenGL libraries to
render surfaces of bones through the volume visualization
method, and uses the HDDIV libraries (by SensAble Inc.) to
59
obtain 6D haptic device inputs and output 3D translational
forces through the PHANToM Desktop (by SensAble Inc.) that
is used to simulate a bur attached pen-like handpiece as
illustrated in Fig. 4.
(A)
(B)
(C)
Figure 3. Determination of bone separation during burring simulations.
Green and blue areas: separate bones. Bold lines: boundary
faces of bone voxels. Shown in 2D for clarity. (A) Beginning
of cutting a bone. (B) During the burring simulation in which
a bone voxel U is nullified. (C) End of cutting a bone in
which the burr entered a null voxel V, thus a new separate
bone is generated. Two seeds (black points) are assigned to
the two sides of the last nullified bone voxel W.
(A)
(B)
Figure 4. A bur attached real pen-like hand-piece simulated by a
pen-type haptic device. (A) A pen-type haptic device:
PHANToM Desktop. (B) A real pen-like handpiece.
3.1 Case of a middle ear disease study
A patient was a 39-year-old man suffering from a
conductive hearing loss in the right ear. Frontal and lateral
X-rays were performed and revealed a fixation between the
tympanic cavity wall with the middle ear ossicles and tissue
sclerosis on these ossicles. CT (spiral CT, GE LightSpeed) was
performed in July, 2008 with 53 transverse slices at 1.25 mm
intervals. These slices followed the DICOM protocol and were
converted as raw data to constitute a volume (256  256  53
intervals between slices and 0.625 mm pixel resolutions in the
slices).
Figure 5 shows the segmentation results for the ROI
including the right ear on several successive transverse CT slices
(in the gravity direction order). And shown in Figs. 5(A)-(D),
some ranges of two separate cavities (A and B) were in the cavity
range and thus were recognized as two independent cavities.
Some ranges of one inside bone in the cavity A were in the inside
bone range and thus were recognized as an inside bone.
Meanwhile, no inside bones in B were recognized. In Figs. 5(E)
and (G), one inside bone in the cavity A and two in B are
recognized. In Fig. 5(F), one inside bone in the cavity A and one
in B are recognized. In Fig. 5(H), the cavity A is not closed;
therefore, is considered an open canal such as the auricular canal
and the inside bone is thus not processed.
J. Med. Biol. Eng., Vol. 30. No. 1 2010
60
A
A
B
B
(A)
(B)
B
B
B
(C)
A
A
A
A
A
A
B
B
(D)
B
(E)
(F)
(G)
(H)
Figure 5. Ossicle segmentation result for the patient with a middle ear disease (see text). Gray area: bones. Blue area: segmented bones inside
a closed cavity.
(A)
(B)
(C)
(D)
Figure 6. Opening of the tympanic wall to reveal ear ossicles with a bone bar from the tympanic cavity to the incus. Gray area: bones. Blue
areas: highlighted ear ossicles. (A) Lateral view of the temporal bone in which tiny but highlighted ossicles partially appeared. (B)
Zooming in to observe the ossicles and neighboring bones. (C) Beginning of opening the tympanic cavity wall to reveal the ossicles.
(D) End of opening the tympanic cavity wall.
The segmentation result indicates one inside bone was
recognized in the tympanic cavity. The 2D inside bones in the
cavity A shown in Figs. 5(A)-(G) construct a continuous 3D
bone including the malleus and incus that can be used for
visualization and surgery simulation. However, the inside bone
that was not recognized, as shown in Fig. 5(H), because of
being in an open canal is actually a part of the malleus and was
also reconstructed and simulated together with the remaining
part of the malleus, as shown in Figs. 6 and 7. Meanwhile, the
stapes was too small to be in the predefined inside bone range
and thus could not be recognized. The inside bones recognized
in another cavity could not be observed from the auricular
canal, therefore would not confuse the system user to observe
the ossicles, even though these bones were also highlighted.
Figure 6 shows the 3D images reconstructed from the
volume before surgery simulations and after the opening of the
temporal bone. Figures 6(A) and (B) show 3D images before
surgery simulations. The tiny ear ossicles in Fig. 6(A) were
highlighted to lead the surgeon to zoom in to an optimal
perspective, as shown in Fig. 6(B), for visualizing and
simulating the ear ossicle surgery. Figures 6(C) and (D) show a
0.6 mm round cutting bur was used to cut the temporal bone for
opening the tympanic cavity. After the opening, the 3D ossicles
and the pathology of a bone bar fixating the incus could be
clearly observed.
Figure 7 shows a 0.4 mm round polishing bur was used to
cut the bone bar for separating the ossicles from the tympanic
cavity wall and polishing the ossicles to delete sclerosis.
Figures 7(A) and (B) show the bone bar has been cut to
separate the ossicles from the tympanic cavity wall. This
separation check lets the surgeon know if the malleus and incus
can be separated from the temporal bone during the burring
operations. Figure 7(C) shows the separate ossicles are
repositioned (by a virtual hand) outside of the cavity for easily
implementing the polishing operations. Figure 7(D) shows the
anterior surface of the incus being polished. Figure 7(E) shows
Ear Ossicle Surgery Simulator
61
(A)
(B)
(C)
(D)
(E)
(F)
(G)
(H)
Figure 7. Simulations of ear ossicle surgery. Gray area: bones. Blue areas: highlighted ear ossicles. (A) Cutting the incus bar to release the
incus fixation. (B) End of cutting the bone bar in which the incus fixation has been released. (C) Repositioning the ossicles out of
the tympanic cavity. (D) Polishing the anterior surface of the incus. (E) Polishing the superior surface of the incus. (F) Polishing the
posterior surface of the malleus. (G) Repositioning the ossicles back into the tympanic cavity. (H) The ossicles have been aligned in
a suitable position.
the superior surface of the incus being polished. Figure 7(F)
shows the posterior surface of the malleus being polished.
Figure 7(G) shows the surfaces of the ossicles have been
polished and was being repositioned. Figure 7(H) shows the
ossicles was repositioned to suitable positions. The surfaces of
the ossicles have become smooth after the polishing operations.
The spatial relation between the ossicles and the tympanic
cavity has become suitable. The above 3D images show our
system can give virtual-reality burring simulations for tympanic
cavity opening, bone bar cutting, and ossicle polishing, and
topological simulations for repositioning ossicles to align the
tiny ossicles in suitable positions.
(A)
(B)
(C)
(D)
3.2 Case of comparative study
Another volume by the same CT machine (256  256  53
with 1.25 mm intervals between slices and 0.7 mm pixel
resolutions in the slices) with no right ear diseases was used to
demonstrate the segmentation, bone surface reconstruction and
opening simulation result for comparative study. Figure 8
shows the segmentation results for the ossicles in the right ear
on several transverse CT slices that are demonstrated in the
gravity direction order. In Figs. 8(B)-(D), one inside bone in
the tympanic cavity is recognized. In Fig. 8(A), the cavity is
not closed; therefore, the inside bone is not highlighted. Figure
9 shows the reconstructed images in which the tiny ear ossicles
were highlighted so that the surgeon zoomed in (from Fig. 9(A))
to an optimal perspective (as shown in Fig. 9(B)) and opened
the tympanic cavity. Although the ossicles were segmented in
fewer slices than the slices in the first case, the sizes and shapes
of the malleus and incus in the two cases are similar. That
indicates the 3D reconstruction (marching cubes method) can
generate continuous ossicle surfaces by the CT slices with the
ossicles segmented (highlighted as blue) or not (gray).
Meanwhile, a bone bar from the temporal bone to fixate the
incus as the first case was not observed in this case.
Figure 8. Ossicle segmentation result for a normal middle ear (see
text). Gray area: bones. Blue area: segmented bones inside a
closed cavity.
(A)
(B)
Figure 9. Opening of the tympanic wall to reveal ear ossicles of a
normal middle ear (see text). Gray area: bones. Blue areas:
highlighted ear ossicles.
62
J. Med. Biol. Eng., Vol. 30. No. 1 2010
4. Discussion
In this study, we developed a method that automatically
segments tiny ear ossicles on transverse CT slices and then
highlights the 3D ossicles from the temporal bone. The surgeon
can visualize the shape of the ossicles and the spatial relations
between the ossicles with the temporal bone to assist diagnosis.
Through the ossicle separation determination function together
with the simulation functions reported in our previous works,
clinicians can implement simulations of cavity opening, and
ossicle separation, polishing and repositioning. Using the
simulator, clinicians can manage and rehearse a surgery plan
for any specific ear ossicle disease.
The segmentation method was developed to segment the
bones inside the cavities in the ROI including an ear. The
tympanic cavity with inside ossicles can be segmented;
however, some inner cavities with inside bones may be
segmented simultaneously. On a 3D virtual skull, these cavities
and inside bones are occluded and are not revealed as the
highlighted ossicles. The outmost auricular canal includes no
inside bones; therefore, the highlighted ossicles may be
partially occluded by the auricular canal wall but can still be
observed by a suitable perspective.
Our system can provide realistic 3D images for volume
data using the marching cubes method and real-time visual
responses through a dynamic cube data structure [18]. This
cube data structure can assign different colors for different
types of bone voxels; therefore, the ossicle voxels detected by
our segmentation method can be assigned a different color from
the temporal bone. However, ossicle pixels at every transverse
slice are detected only inside the closed tympanic cavity. The
ossicles, especially the inferior part of the malleus, may be
resolved together with the open auricular canal and excluded by
the segmentation. This inferior malleus is demonstrated as a
small gray (same as the temporal bone) area, as shown in Figs.
6 and 7. This area actually does not affect the judgment
regarding the ossicles and the following simulations, such as
burring operations or repositions. However, this indicates our
segmentation method is critical for a slice resolving
simultaneously the auricular canal and the tympanic cavity, and
a more accurate segmentation method should be discussed.
The CT slices in the two case studies showed the artifacts,
such as patient motion, metal and beam-hardening artifacts did
not appear. The stair-step artifacts that easily occur in 3D image
reconstructed from helical CT machines [19] were also not
clear on the 3D ossicles in the two sample cases. However, the
resolutions including the pixel widths (0.6–0.7 mm) and slice
intervals (1.25 mm) were insufficient for tiny ossicles (7–10
mm) and thin tympanic cavity walls to bring the following
artifacts. First, the anatomic direction of the small stapes is near
horizontal; therefore, the stapes easily lies between two slices
and thus not or only little resolved on transverse CT slices.
Second, partial volume artifacts easily arise at thin parts of the
tympanic cavity wall. Because a voxel containing the thin wall
may contain other tissues, the CT number or gray value of this
voxel is decreased by the other tissues to be segmented as not a
bone. As in the comparative case study, the voxel gray-levels at
the tympanic cavity on most the slices were decreased to be not
segmented as bones. As a result, the tympanic cavity became
open and inside bones in the cavity were not recognized as
ossicles, as shown in Figs. 8(A) and 9. These artifact problems
can be solved by using smaller pitches to obtain smaller slice
intervals and converting the DICOM data into raw data with
finer resolutions (e.g., 512  512). However, real-time (1000
Hz haptic and 20 Hz visual responses) burring simulations
become difficult to achieve under current PC platforms because
such resolutions bring too many voxels. Therefore, future work
includes discussion of using more advanced PC hardware to
achieve real-time simulations under sufficiently fine slice
intervals and resolutions.
A typical patient with ear ossicle pathology was used as
the example. This pathology demonstrated a bone bar from the
tympanic cavity wall fixating the ossicles and sclerosis on the
ossicles. The burring functions were used to open the tympanic
cavity, then cut the bone bar to separate and polish the ossicles.
These burring simulations resembled the real procedures [20].
In addition, the ossicles were seldom repositioned outside the
tympanic cavity but usually burred and aligned inside the
cavity [20]. Although the ossicles were repositioned outside in
our sample simulation for easy observation and operation,
surgeons can also use the system functions to operate the
ossicles directly inside the tympanic cavity. Meanwhile,
patients with different ear ossicle or middle ear pathology such
as ossicle absence, deformity, disruption and erosion should be
our future work to test whether different ear ossicle pathology
can be observed through our segmentation and volume
visualization methods. Moreover, simulations for the
procedures of different ear ossicle surgery such as ossicular
chain reconstruction or ossicular replacement should also be
considered.
5. Conclusion
The 3D geometry visualization of tiny middle ear ossicles
is important in deciding the appropriate diagnostic modality
and treatment procedures for a middle ear disease. In this study,
we have proposed a segmentation method to highlight the tiny
ear ossicles from the temporal bone for easy observation.
Diagnosis can be confirmed through the observation of the 3D
ossicle shapes and the spatial relations between the ossicles
with the temporal bone. Our separation method judges if a new
bone is generated during the burring simulation. Combining
this technique with our 3D volume reconstruction technique,
real-time haptic cutting simulations, and topological
simulations of a bone deletion and repositioning, our simulator
facilitates simulating surgical procedures on the temporal bone
and the ear ossicles to enable simulations for the ear ossicle
surgery. Our example of a patient with a bone bar from the
tympanic cavity wall to fixate the ear ossicles and sclerosis on
the ear ossicles demonstrated that the simulator allows
clinicians to visualize the pathology of ear ossicles and
simulate surgical procedures for treating the ossicle disease. In
conclusion, the proposed simulator can be a diagnostic, surgery
plan management and verification tool for ear ossicle diseases.
Ear Ossicle Surgery Simulator
Acknowledgment
This study was partially sponsored by the National
Science Council (NSC), Taiwan, ROC. Grant number:
NSC-97-2221-E33-048.
[11]
[12]
References
S. B. Ogale, S. B. Mahajan, S. Dutt and J. H. Sheode, “Fate of
middle ear implants,” Auris Nasus Larynx, 24: 151-157, 1997.
[2] P. S. Mundarda and S. J. Jaiswal, “A method for ossicular
reconstruction with tragal cartilage autografts,” Laryngoscope,
99: 955-962, 1989.
[3] A. G. Kerr and G. D. L. Smyth, “The fate of transplanted
ossicles,” J. Laryngol. Otol., 85: 337-347, 1971.
[4] S. B. Ogale, S. Dutt, S. Thakur and M. Pawar “The styloid
process in ossicular chain reconstruction. II: Long-term
analysis,” J. Laryngol. Otol., 108: 111-112, 1994.
[5] E. Steinbach and A. Pusalkar, “Long-term histological fate of
cartilage in ossicular reconstruction,” J. Laryngol. Otol., 114:
1231-1239, 1981.
[6] S. Baumrind, F. H. Moffitt and S. Curry, “The geometry of
three-dimensional measurement from paired coplanar x-ray
images,” Am. J. Orthod., 84: 313-322, 1983.
[7] F. Veillon, S. Riehm, M. N. Roedlich, P. Meriot, E. Blonde and J.
Tongio, “Imaging of middle ear pathology,” Semin.
Roentgenology, 35: 2-11, 2000.
[8] K. Urano, K. Nakayama, H. Miyashita, M. Isono, Y. Hijii and K.
Murata, “Evaluation of reconstructed 3-D images of the middle
ear using multi-slice scan CT,” Int. Congr. Ser., 1240: 1487-1490,
2003.
[9] Y. Kurosaki, Y. O. Tanaka and Y. Itai, “Malleus bar as a rare
cause of congenital malleus fixation: CT demonstration,” Am. J.
Neuroradiol., 19: 1229-1230, 1998.
[10] R. Maroldi, D. Farina, L. Palvarini, A. Marconi, E. Gadola, K.
[1]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
63
Menni and G. Battaglia, “Computed tomography and magnetic
resonance imaging of pathologic conditions of the middle ear,”
Eur. J. Radiol., 40: 78-93, 2001.
T. Nakasato, M. Sasaki, S. Ehara, Y. Tamakawa, K. Muranaka, T.
Yamamoto, H. Chiba, T. Ishida and K. Murai, “Virtual CT
endoscopy of ossicles in the middle ear,” Clin. Imaging, 25:
171-177, 2001.
D. Morris, C. Sewell, F. Barbagli, K. Salisbury, N. H. Blevins
and S. Girod, “Visuohaptic simulation of bone surgery for
training and evaluation,” IEEE Comput. Graph. Appl., 26: 48-57,
2006.
M. Agus, A. Giachetti, E. Gobbetti, G. Zanetti, N.W. John and R.
J. Stone, “Mastoidectomy simulation with combined visual and
haptic feedback,” in: J. D. Westwood, H. M. Hoffmann, G. T.
Mogel and D. Stredney (Eds.), Medicine Meets Virtual Reality,
Amsterdam, Netherlands: IOS Press, 17-23, 2002.
S. B. Jou and M. D. Tsai, “A fast 3D seed-filling algorithm,”
Visual Comput., 19: 135-149, 2003.
M. D. Tsai and M. S. Hsieh, “Accurate visual and haptic burring
surgery simulation based on a volumetric model,” Journal of
X-ray Science and Technology, 18, 2010. (in press)
M. D. Tsai and M. S. Hsieh, “Volume manipulations for
simulating bone and joint surgery,” IEEE T. Inf. Technol. Biomed.,
9: 139-149, 2005
W. E. Lorensen and H. E. Cline, “Marching cubes: a high
resolution 3D surface construction algorithm,” ACM SIGGRAPH
Comput. Graph., 21: 163-169, 1987.
J. F. Lee, M. S. Hsieh, C. W. Kuo, M. D. Tsai and M. Ma,
“Real-time three-dimensional reconstruction for volume-based
surgery simulations,” Biomed. Eng. Appl. Basis Commun., 20:
205-218, 2008.
D. Fleischmann, G. D. Rubin, D. S. Paik, S. Y. Yen, P. R. Hilfiker,
C. F. Beaulieu and S. Napel, “Stair-step artifacts with single
versus multiple detector-row helical CT,” Radiology, 216:
185-196, 2000.
H. Hildmann and H. Sudhoff (Eds.), Middle Ear Surgery,
Heidelberg (Germany): Springer, 2006.
64
J. Med. Biol. Eng., Vol. 30. No. 1 2010