AJR 2003; 181:857-859
© American Roentgen Ray Society
Using Volume Segmentation of Cine MR Data to Evaluate Dynamic Motion of the Airway in Pediatric Patients
M. Bret Abbott1,2,
Bernard J. Dardzinski1 and
Lane F. Donnelly1
1 Department of Radiology, Cincinnati Children's Hospital Medical Center, 3333
Burnet Ave., Cincinnati, OH 45229-3039.
Received October 30, 2002;
accepted after revision March 5, 2003.
Address correspondence to L. F. Donnelly.
(Lane.Donnelly{at}chmcc.org).
2 Present address: Department of Internal Medicine, The Christ Hospital, 2139
Auburn Ave., Cincinnati, OH 45219.
Introduction
Airway imaging techniques such as cephalometrics, fluoroscopy, CT, and MRI
provide a powerful array of tools to evaluate the airway
[1-8].
Of these imaging techniques, cine MRI provides high-quality cross-sectional
images of the airway over time, documenting airway dynamics throughout the
respiratory cycle
[4-8].
To create cine MRIs, we obtained multiple fast gradient-echo images over time
in the identical anatomic plane. These images can then be displayed as a
real-time "movie." The technique has been shown to be a useful
tool in the evaluation of certain populations of patients with obstructive
sleep apnea
[4-8]
by showing abnormalities of the dynamic motion of the airway such as
glossoptosis and pharyngeal collapse.
Qualitatively, cine MRI can show such mechanisms of airway obstruction
[4-8].
However, quantifiable evaluation has been limited to comparing measurements of
the maximal and minimal diameters of the anatomic portion of the airway in
question at two individual points in time
[4,
5,
8]. Volume segmentation of cine
MRIs is shown as a novel technique to evaluate airway dynamics. Our objective
for creating volume segmentation techniques for the evaluation of cine MR data
of the airway was to create a means of quantifiably evaluating airway motion
over time. Such a tool may be useful in comparing patterns of airway motion
among patients and in the same patient before and after an intervention.
Volumetric segmentation of cine MR data may prove to be a useful diagnostic
and research tool.
Materials and Methods
At our institution, MRI studies of the airway, which include cine MR
sequences, have become a standard tool for the evaluation of certain
populations of children with obstructive sleep apnea. The technique for cine
MR studies and clinical applications of such studies have been described in
detail [4,
5,
8]. The clinical indications
for cine MR studies at our institution include persistent obstructive sleep
apnea despite prior surgical therapy such as tonsillectomy and adenoidectomy,
obstructive sleep apnea with predisposition to obstruction at multiple sites
(craniofacial anomalies and Down syndrome), and preoperative evaluation of
obstructive sleep apnea before complex airway surgery. Patients are typically
evaluated under conscious sedation administered via a radiology sedation
program [4,
5,
8]. A respiratory therapist is
present during the entire study. A 1.5-T MRI unit is used. The patients are
placed in the head-neck vascular coil in the supine position with the neck in
neutral position. Among the various MRI series performed in these patients,
cine MRIs are created by obtaining a fast gradient-echo pulse sequence
(fTR/TE, 8.2/3.6; flip angle, 80°; TR; slice thickness, 12 mm; field of
view, 24 cm; in-plane resolution, 256 x 256).
Cine MRIs are obtained in the sagittal midline plane and in the axial plane
at the level of the mid portion of the tongue. One hundred twenty-eight
consecutive images are obtained with a total imaging time of approximately 2
min. Therefore, an image is generated every 0.94 sec. Cine MRIs are obtained
at times when the child shows noisy breathing or oxygen desaturation
[8]. These images are displayed
in cine format, creating a real-time movie of airway motion for qualitative
evaluation.
For volume-segmentation analysis, the cine MRIs are transferred to a
computer workstation in the Digital Imaging and COmmunications in Medicine
file format. The series of images are loaded into a CCHIPS-based
image-analysis software program (Cincinnati Children's Hospital Image
Processing Software, an in-house interface definition language, Research
Systems, Boulder, CO)
(www.irc.chmcc.org).
CCHIPS converts the 128 separate images into to a single data matrix of a
signal intensity containing the entire time course of the cine MR sequence.
The region of interest for volume segmentation is manually selected to
encompass either the naso- or the hypopharynx on the sagittal images or the
entire cross section of the hypopharynx on axial images
(Fig. 1A). The matrix of
intensity data within the region of interest is then analyzed with the
segmentation routine within CCHIPS, using a k-means clustering algorithm
[9]. The region of interest
being evaluated is segmented using the signal intensity for the entire data
set (all slices over all time evaluated). The number of intensity levels
segmented can be defined by the user.
View larger version (64K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1A. —Representative cine MRIs and volume segmentations in
5-year-old girl with obstructive sleep apnea. Axial (A) and sagittal
(B) fast gradient-echo cine images were obtained at level of
hypopharynx. Images on left show time at which airway is near completely
collapsed. Images on right show time of maximal opening of airway. In each
image, white regions of interest show anatomic area for volume segmentation.
For sagittal images, regions of interest are marked for both naso- and
hypopharynx. Horizontal white line on sagittal images indicates approximate
location of axial plane.
|
|
After testing multiple choices for the number of signal intensity segments,
we determined that the use of three segments yielded the most accurate
depiction of the airway volumes by subjective assessment of the visibility of
the airway borders on the basis of clinical imaging expertise. The segmented
intensity data are displayed by assigning each intensity level a particular
shade of gray (Fig. 1B). The
darkest gray area represents the signal intensity of the patent airway,
whereas the intermediate and lighter shades represent the signal intensity of
adjacent tissues. Two segments are used to represent soft tissue to compensate
for motion artifacts and intrinsic air-soft tissue interface artifacts. By
using a third segment to account for the noise generated, we believe that the
volumes calculated more accurately reflect the patent portion of the airway.
The image segment representing the patent airway is selected, and airway
volume is plotted as a function of time. Segmentation of two-dimensional
images quantifies a volume because the imaging planes have an intrinsic
thickness (12 mm). Because the imaged plane is of nearly uniform thickness,
the segmentation volumes are directly proportional to the cross-sectional area
of the imaged airway. These images also clearly show the dramatic changes in
airway volume throughout the respiratory cycle. Change in volume over time can
be normalized by average airway volume for the patient and plotted across the
time course. Normalization allows the change in the volume of the area to be
viewed in relationship to the relative size of the patient's airway because
this volume may vary among patients of various sizes and because of the
subjective selection of regions of interest. The changes in the normalized
volume over time then can be shown in a graph format
(Fig. 2). Various patients can
also be compared with this technique (Fig.
2).
View larger version (69K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1B. —Representative cine MRIs and volume segmentations in
5-year-old girl with obstructive sleep apnea. Axial (A) and sagittal
(B) fast gradient-echo cine images were obtained at level of
hypopharynx. Images on left show time at which airway is near completely
collapsed. Images on right show time of maximal opening of airway. In each
image, white regions of interest show anatomic area for volume segmentation.
For sagittal images, regions of interest are marked for both naso- and
hypopharynx. Horizontal white line on sagittal images indicates approximate
location of axial plane.
|
|
View larger version (27K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2. —Graph shows comparison of change in normalized axial airway
volume over time between patients with and without obstructive sleep apnea.
Hypopharyngeal airway volumes were obtained as function of time by volume
segmentation of cine MRIs from 5-year-old girl with obstructive sleep apnea
(Figs. 1A,
1B,
1C, and
1D) and 5-year-old asymptomatic
boy being imaged for other indications. Volumes segmented from axial cine MRIs
in each patient were normalized and are depicted over time. Dark tracing
represents asymptomatic patient and light tracing represents patient with
obstructive sleep apnea. Asymptomatic patient shows minimal variation in
volume over time. Patient with obstructive sleep apnea shows marked variation
in volume over time.
|
|
View larger version (9K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1C. —Representative cine MRIs and volume segmentations in
5-year-old girl with obstructive sleep apnea. Corresponding
volume-segmentation images show quantification of airway cross sections in
collapsed and patent conformations for hypopharynx in axial plane (C)
and nasopharynx and hypopharynx in sagittal plane (D).
|
|
View larger version (12K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1D. —Representative cine MRIs and volume segmentations in
5-year-old girl with obstructive sleep apnea. Corresponding
volume-segmentation images show quantification of airway cross sections in
collapsed and patent conformations for hypopharynx in axial plane (C)
and nasopharynx and hypopharynx in sagittal plane (D).
|
|
Discussion
Multiple imaging techniques such as CT, MRI, radiography, and
cephalometrics have offered useful and unique information in the study of
patients with abnormalities of the central airways
[1-8].
Many of these techniques rely almost entirely on static anatomic information.
In the evaluation of patients with obstructive sleep apnea, dynamic processes
such as glossoptosis and pharyngeal collapse are often not optimally evaluated
with static imaging alone. Imaging techniques that can show motion over time,
such as cine MR techniques and sleep fluoroscopy, have shown a number of
advantages in studying certain populations of patients with obstructive sleep
apnea
[4-8].
It has been shown that cine MR studies can depict abnormal dynamic processes
qualitatively and that they are useful in planning patient care
[2-8].
Volume segmentation of cine MR data offers a way to evaluate changes in
airway volume as opposed to changes in linear measurements. In addition, it
allows evaluation of changes in the airway volume over time during the
respiratory cycle. Quantifiable data can be displayed graphically, subject to
statistical analysis; evaluated for patterns in change over time; and used to
objectively characterize patient populations and airway disorders. Graphic
display of airway volume as a function of time may reveal irregularities in
the respiratory cycle that are not readily observed with the standard cine MR
display technique or by measurement of airway diameters. Statistical
manipulations such as the SD, range, coefficient of variance, and normalized
range of airway volumes could be easily applied to volume-segmentation data
and may provide useful measures of airway dynamics. Quantifiable measurements
that reflect the entire time course of data, such as the mean for the SD in
change of volume over time, can be created. These analyses would be
prohibitively cumbersome to apply to anatomic measurements in a large series
of images for each patient.
Quantifiable and objective measures of airway dynamics provided by volume
segmentation may facilitate characterization of patterns of airway dysfunction
in specific patient populations with a high incidence of breathing dysfunction
such as in patients with Down syndrome, obesity, and Pierre Robin
malformation. They may also help define the patterns of airway motion that
should be considered physiologic and how to differentiate normal physiologic
motion from abnormal patterns of motion. In addition, they may provide
information concerning the coordination of airway motion among different
portions of the airway, such as the hypopharynx, oropharynx, and nasopharynx,
and define normal and abnormal relationships between airway motion in these
areas. In the examples shown in this article, marked differences in the
patterns of airway motion are seen in a child without airway symptoms and one
with obstructive sleep apnea (Fig.
2). Markedly greater changes in airway volume occur over time in
the patient with obstructive sleep apnea.
In conclusion, volumetric segmentation of information from cine MRIs may
provide a useful tool in quantifiably evaluating airway motion for both
clinical and research purposes. One of the largest limitations has been the
availability of adequate quantifiable information from such cine MR studies.
Volume segmentation is beginning to provide us with a tool to create more
detailed quantifiable data than were previously available. Further studies are
needed to determine the true potential uses of volume segmentation of cine MR
airway data.
References
- de Miguel-Diez J, Alvarez-Sala JL, Villa-Asensi JR. Magnetic
resonance imaging of the upper airway in children with Down syndrome. (reply
to letter) Am J Respir Crit Care Med2002; 165:1187[Free Full Text]
- Donnelly LF, Strife JL, Myer CM. Dynamic sleep fluoroscopy in
children with obstructive sleep apnea. Appl Radiol2001; 178:30
-34
- Gibson SE, Myer CM 3rd, Strife JL, O'Connor DM. Sleep fluoroscopy
for localization of upper airway obstruction in children. Ann Otol
Rhinol Laryngol 1996;105:678
-683[Medline]
- Donnelly LF, Casper KA, Chen B, Koch BL. Defining normal upper
airway motion in asymptomatic children during sleep by means of cine MR
techniques. Radiology2002; 223:176
-180[Abstract/Free Full Text]
- Donnelly LF, Casper KA, Chen B. Correlation on cine MR imaging of
size of adenoid and palatine tonsils with degree of upper airway motion in
asymptomatic sedated children. AJR2002; 179:503
-508[Abstract/Free Full Text]
- Suto YT, Matsuo T, Kato T, et al. Evaluation of the pharyngeal
airway in patients with sleep apnea: value of ultrafast MR imaging.
AJR 1993;160:311
-314[Abstract/Free Full Text]
- Shellock FG, Schatz CJ, Julien P, et al. Occlusion and narrowing of
the pharyngeal airway in obstructive sleep apnea: evaluation by ultrafast
spoiled GRASS MR imaging. AJR1992; 158:1019
-1024[Abstract/Free Full Text]
- Donnelly LF, Surdulescu V, Chini BA, Casper KA, Poe SA, Amin RS.
Upper airway motion depicted at cine MR imaging performed during sleep:
comparison between young patients with and those without obstructive sleep
apnea. Radiology2003; 227:239
-245[Abstract/Free Full Text]
- Bezdek JC, Hall LO, Clarke LP. Review of MR image segmentation
techniques using pattern recognition. Med Phys1993; 20:1033
-1048[Medline]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
|
|
 
L. F. Donnelly
Obstructive Sleep Apnea in Pediatric Patients: Evaluation with Cine MR Sleep Studies
Radiology,
September 1, 2005;
236(3):
768 - 778.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. B. Abbott, L. F. Donnelly, B. J. Dardzinski, S. A. Poe, B. A. Chini, and R. S. Amin
Obstructive Sleep Apnea: MR Imaging Volume Segmentation Analysis
Radiology,
September 1, 2004;
232(3):
889 - 895.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Introduction
Materials and Methods
