Extraocular muscle deformation assessed by
motion-encoded MRI during eye movement
in healthy subjects
Institute for Biomedical Engineering,
University and ETH Zurich, Zurich, Switzerland, &
Department of Ophthalmology,
University Hospital of Zurich, Zurich, SwitzerlandMarco Piccirelli
Institute for Biomedical Engineering,
University and ETH Zurich, Zurich, SwitzerlandRoger Luechinger
Institute for Biomedical Engineering,
University and ETH Zurich, Zurich, SwitzerlandAndrea K. Rutz
Institute for Biomedical Engineering,
University and ETH Zurich, Zurich, SwitzerlandPeter Boesiger
Department of Ophthalmology,
University Hospital of Zurich, Zurich, SwitzerlandOliver Bergamin
Conventional magnetic resonance imaging (MRI) is useful for assessing morphological changes but not for assessing
deformations inside homogeneous structures (M. D. Abràmoff, A. P. Van Gils, G. H. Jansen, & M. P. Mourits, 2000). Since
no intrinsic contrast can be imaged for distinguishing heterogeneous patterns of muscle contraction, morphological changes
along the length of the extraocular muscles (EOMs) are not macroscopically detectable. However, an imaging method that
is able to directly encode motion could give evidence about the dynamics of the inhomogeneous deformation of the EOMs.
Thus, we developed a method for acquiring motion-encoded MRI images of the EOMs during eye movements. Seven
healthy subjects gazed at a horizontal sinusoidally oscillating target. A small surface coil was placed in front of the right
orbit. The contracting and relaxing horizontal rectus muscles and the noncontracting optic nerve were reliably tracked. The
differential contractility of the EOMs could be distinguished from the third time frame on (=140 ms from the beginning of the
right to left eye movement lasting 1 s). The muscle belly of the contracting medial rectus muscle could be distinguished from
the posterior and the anterior segment from the sixth time frame on (=350 ms). In conclusion, motion-encoded MRI resolved
the heterogeneous contraction of moving EOM segments in healthy subjects without using physical markers.
Keywords: motion-encoded MRI, tagging, physiology of extraocular muscle movement
Citation: Piccirelli, M., Luechinger, R., Rutz, A. K., Boesiger, P., & Bergamin, O. (2007). Extraocular muscle deformation
Introduction
The pattern of movement within the extraocular
muscles (EOMs) and the surrounding orbital connective
tissue is not yet understood (Miller, Rossi, Wiesmair,
Alexander, & Gallo, 2006). Knowledge about the physi-
ology of the contractility pattern of the EOMs during eye
movement is essential to understanding the pathology
progression in diseases that affect the EOMs. So far,
visualization of the EOM path by magnetic resonance
imaging (MRI) has been restricted to multi-positional
imaging.
However, the MRI signal has been known to be
sensitive to motion since the fifties (Suryan, 1951). First
applied to measure blood flow in healthy subjects (Morse
& Singer, 1970), the tagging of tissue magnetization was
introduced in 1988 to depict the heart deformation within
an image plane (Zerhouni, Parish, Rogers, Yang, &
Shapiro, 1988). Further improved by Axel and Dougherty
(1989) and Fischer, Mckinnon, Maier, and Boesiger
(1993), this method was later adapted to other tissues,
e.g., the brain (Soellinger, Ryf, Boesiger,&Kozerke, 2007).
Motion-encoded MRI is able to capture the motion of
tissues relative to a fixed spatial referential system. The
motion of tissue points can even be visually tracked
1
ISSN 1534-7362 * ARVO
Journal of Vision (2007) 7(14):5, 1–10 http://journalofvision.org/7/14/5/
http://journalofvision.org/7/14/5/, doi:10.1167/7.14.5.
assessed by motion-encoded MRI during eye movement in healthy subjects. Journal of Vision, 7(14):5, 1–10,
doi: 10 .1167 /7 .14 .5 Received April 20, 2007; published November 21, 2007

directly from the MR images without postprocessing, as
will be shown later in the first two figures. This method
will permit detailed studies of normal EOMs in vivo and
depict the deformation of soft orbital tissues during
smooth movements of the eye. The objective of this study
was to take advantage of motion-encoded MRI to measure
the heterogeneous deformation along the lateral and
medial rectus muscles and the optic nerve in healthy
subjects.
This study contains three parts. First, we present images
that can be obtained with motion-encoded MRI and
analyze their dissimilarities with conventional MRI
(Figure 1). For objective measurement of the relative
motion of tissue points, an automatic tracking method is
used (Figure 2). The second part of this study tests the
reliability of the tracking method. Using a supramillimetric
resolution in this study, we tracked several neighboring
tissue points that are expected to have a similar motion and
investigated the variation of the results (Figures 3 and 4).
The third part presents the physiological data that were
obtained (Figures 5 and 6).
Methods
Subjects and setup
The study was conducted according to the tenets of the
Declaration of Helsinki and approved by the local ethics
committee. Nine subjects were recruited for the imaging
part of the study, and informed consent was obtained from
the subjects after explanation of the nature and possible
consequences of the study. The visual acuity of all
subjects was sufficient to track the visual stimulus
described below. To compare the deformation of the
EOMs with the optic nerve (ON), we required the
complete muscles and ON to be visible within one image
plane of 4 mm thickness. Therefore, two subjects with a
strongly curved ON were excluded because their ON
could not be completely visualized in the same transversal
image plane as the horizontal EOMs. Seven healthy
volunteers (two women and five men; mean age: 36 years;
range: 27 to 62 years of age) were included for further
analysis.
MR images of the right orbit were acquired using a
small receive-only surface coil (47 mm diameter) on a
1.5 T system (Achieva 1.5 T; Philips Medical Systems,
Best, The Netherlands). Small surface coils have a greater
signal to noise ratio (SNR) that declines faster with depth
than larger coils. Preliminary experiments showed, how-
ever, that the microscopy coil used still had a better SNR
for the orbital apex than the next bigger surface coil in our
possession. The microscopy coil was placed like a
monocle, so that it was possible to see the target through
it. The head of the subject was immobilized with foam
pads. A typical functional MRI setup (computer projector,
screen, and the software “Presentation”VNeurobehavioral
Systems Inc., Albany CA, USA) was used for the
presentation of the visual stimulus. A mirror allowed the
subjects to gaze out of the bore to the projection screen.
The horizontal gaze range was 40 deg. The room light was
turned off to maximize the contrast of the stimulus.
Stimulus paradigms
A horizontal sinusoidally oscillating white square
(target size = 0.4 deg, luminance = 5.1 T 1 cd/mm2 on a
background of 0.05 T 0.02 cd/mm2) with an amplitude of
T20 deg and a period of 2 s (corresponding to a maximal
target velocity of 64 deg/s) was presented on a black
background to induce smooth pursuit eye movements. It is
known that humans can pursue targets moving at 60 deg/s
rather well (Meyer, Lasker, & Robinson, 1985). Prelimi-
nary experiments using scleral search coils confirmed that
a maximal target speed of 64 deg/s induced only few
catch-up saccades, see also Figure 3 in (Yee, Goldberg,
Jones, Baloh, & Honrubia, 1983). These preliminary
experiments were done on a subgroup of the same subjects
outside the scanner bore. The results at this velocity
confirmed that indeed only a negligible number of catch-
up saccades were present. During preparation, all subjects
confirmed the target was visible at all times. The sinus-
oidal oscillation of the target was repeated for approxi-
mately 4.5 min. Images were acquired on all subjects
from right to left horizontal gaze. A slight up-gaze
position, dependent on the subjects’ anatomy, was
necessary for imaging the ON in the same plane as the
horizontal rectus muscles.
MRI sequence
Transversal MR images were acquired with a fast
gradient echo sequence. The right-to-left horizontal eye
movement of 40 deg was split into fifteen time frames of
70 ms (1050 ms acquisition period). The remaining
950 ms of the 2 s periodic eye movement served for
signal recovery. The relatively short time frame duration
of 70 ms was chosen so that the deformation of the orbital
tissues between two time frames was small enough to
allow tracking. The following scan parameters were
selected: field of view: 140
140 mm2 (big enough to avoid
fold over artifacts), scan resolution: 1.2
1.2
4.0 mm3
(limited by scan duration and signal to noise ratio),
number of signal averages: 8 (to increase the signal to
noise ratio), reconstruction matrix 256
256. The use of
an echo planar imaging (EPI) factor of 5 shortened the
acquisition time to 4.5 min.
The MRI sequence CSPAMM (Complementary SPAtial
Modulation of Magnetization) (Fischer et al., 1993)
encodes the motion of each tissue point. Before the images
Piccirelli et al. 2Journal of Vision (2007) 7(14):5, 1–10

are acquired, the magnetization of the tissue is periodically
modulated in two perpendicular directions, so called tissue
tagging. The resulting images look like anatomical images
modulated by a grid pattern (see Figure 1B at time 0 s).
The period of the sinusoidal modulation is termed the
tagline distance. Since the modulation is used as marker,
equivalent tissue markers are present at every tagline
distance. As the magnetization moves with the underlying
tissue, the displacement of each tissue point can be
followed during the entire recording. The change of the
distance between each marker was measured.
The postprocessing software assigns the nearest equiv-
alent tissue point of the next time frame to each tissue
point. As the postprocessing software cannot differentiate
equivalent markers, the tagline distance should measure at
least twice the maximal tissue displacement between two
time frames. Otherwise, a movement in the wrong
direction could be assigned to a tissue point. To select
the appropriate tagline distance, the expected movement
of the orbital tissue was calculated as follows: assuming
the ocular globe including the adjacent connective tissues
builds a 25- to 35-mm diameter sphere that rotates 64 deg/s
during 70 ms (=4.5 deg) following the moving stimulus,
the surface of such a sphere shifts 1.4 mm at most
(4.5 deg*diameter*pi/360 deg). A tagline distance of at
least a double sphere shift (3 mm) was needed. A larger
tagline distance would increase the sensitivity of the
measurement to noise and would consequently need
extended scan time for compensation.
Postprocessing
An adapted software program based on TagTrack 1.5.6
(GyroTools Ltd.; Zurich; Switzerland) was used to track
the marked tissue points automatically. The postprocessing
Figure 2. CSPAMM images of all 15 time frames including the
manually embedded polylines (in color) in one subject (Subject 1)
gazing from 20 deg right to 20 deg left. Five polylines served to track
the two horizontal muscles and the optic nerve of the right eye.
Figure 1. (A) Static MR image of the right orbit without tagging. The optic nerve (ON), the lateral rectus muscle (LRM), and the medial
rectus muscle (MRM) are depicted in the image plane. (B) Four CSPAMM MR images of the same slice during eye movement. The
magnetization is modulated to create a grid bind with the tissue (see time 0 s). The deformation of this grid (during time) depicts the
differential movements within homogeneous tissues, such as in the extraocular muscles. The 1st, 5th, 10th, and 15th time frames are
shown. The temporal resolution was 70 ms. These images are from the same subject (Subject 1).
Piccirelli et al. 3Journal of Vision (2007) 7(14):5, 1–10

method HARP (HARmonic Phase) (Osman, Kerwin,
McVeigh, & Prince, 1999) with peak-combination (Ryf,
Tsao, Schwitter, Stuessi, & Boesiger, 2004) is integrated
into this software. HARP enables tracking of all tissue
points, not just the tagging lines (dark lines of the grid),
because it tracks the phase and not the MRI magnitude
information. To understand motion-encoded MRI with
HARP evaluation, it is important to distinguish the
distance between equivalent markers (the tagline distance)
from the acquired pixel size, which is always smaller. A
circular band-pass filter was applied to extract the
harmonic peaks in Fourier space and to diminish image
noise. A filter diameter that doubles the image pixel size is
the theoretical optimum of the HARP method. The size of
the filter (which corresponds to a diameter of 2.7 pixels
of the image) was selected as a trade-off between tracking
stability and movement resolution. The filter was centered
on the harmonic peak (of the sinusoidally modulated
image) and enabled us to resolve a maximal contraction of
53% of the original tissue length at the first time frame.
The maximal resolvable contraction with the optimal
theoretic filter is given by the scan resolution (1.2 mm)
divided by the tagging line distance (3 mm) (Osman et al.,
1999), i.e., 40% (i.e., 2.5 times shorter). Hence, the
degradation due to filtering is reasonable. There is no
limitation for the maximal trackable elongation of a tissue.
The horizontal extraocular muscle thickness of about
3 mm (Kaufmann & Decker, 1995) was covered with at
least one pixel that lay completely inside the muscle after
filtering. Therefore, landmark chains traced and tracked
on these pixels along the muscles were expected to
describe similar motion. A good tracking technique should
render similar motion for landmarks in the same pixel. To
test the quality of the tracking algorithm, five landmark
chains (=polylines) for each horizontal rectus muscle and
the ON were manually drawn on the 10th time frame
(approximately gaze straight ahead) such that the whole
tissue broadness could be used for tracking (see Figure 2).
For each polyline, TagTrack interpolated about 70 equally
spaced points, and all of them were tracked through the 15
time frames. The standard deviation of the motion pattern
of the polylines was calculated for the validation of the
tracking algorithm. A small standard deviation corresponds
to a good tracking quality of the algorithm. To ascertain
that the polylines lay on the expected tissue, we took
advantage of anatomical images and realigned the polylines
if they were not on the muscles of interest. Although the
signal to noise ratio dropped at the orbital apex, the tracking
of the polylines was still reliable. The whole postprocessing
procedure took in average 20 min for each subject.
Numerical and statistical evaluation of the
polylines
The length of each polyline was defined as the sum of
the distances between neighboring landmarks along the
polyline. As reference length, the average of the length of
the first and second time frames (20 deg right gaze) was
selected. This improved the signal to noise ratio and was
justified by the fact that the eye moved less than half a
pixel during the first two time frames. The deformation
(relative length change) of each polyline was calculated
by dividing its length at the actual time frame by its
reference length. Furthermore, each polyline was divided
into three segments of equal length.
The changes of length of the horizontal rectus muscles
as well as of the three EOM segments of the seven
subjects were statistically analyzed by the Kruskal–Wallis
test (ANOVA without Gaussian distribution assumption).
Results
Panels A and B of Figure 3 show representative tracking
results of the right lateral rectus muscle (LRM), medial
rectus muscle (MRM), and optic nerve (ON) of two
different subjects. The LRMs of the right eye elongated
+22% in Subject 1 and +19% in Subject 2 during 20 deg
right to 20 deg left gaze. All five polylines of the LRM
showed a consistent elongation as their standard devia-
tions did not exceed 1% during the 15 time frames.
Meanwhile, the MRMs of both subjects contracted from
baseline to 83% with a length variation between the
polylines of less than 2% in Subject 1 and 0.5% in Subject 2.
The four times larger standard deviation of the MRM
polylines of Subject 1 compared to Subject 2 is explained
by the tracking imperfection of the polylines of the MRM of
Subject 1: Polyline crossing can be observed on Figure 2 at
the first time frame. As expected, the deformation patterns
of both muscles were nearly sinusoidal corresponding to
the sinusoidally oscillating right eye. The average length
of the five polylines of the ON varied between 102% and
98% in Subject 1 and between 100% and 96% in
Subject 2. For Subject 2, the standard deviation of the
five polylines of the ON was slightly greater than in the
two rectus muscles but was still smaller than 2%. Subject 1
did not show a greater standard deviation of the ON
polylines compared to the EOM polylines. The mean of the
standard deviation of all subjects at each single time frame
was smaller than 1% for the three tissues.
The actual lengths of the polylines of the three tissues
at the first and last time frames are listed in Panels 3C
and 3D. To unambiguously compare the lengths of the
different polylines, we consider in this paragraph the
average of the lengths at the first and last time frames.
For Subject 1, the LRM was slightly longer than the
MRM, and the ON was the shortest, as expected.
However, the length analysis of the polylines of Subject 2
revealed that the polylines did not represent the whole
length of the labeled tissue, as the LRM polylines were
shorter than the MRM polylines. The anterior part of
Piccirelli et al. 4Journal of Vision (2007) 7(14):5, 1–10

the LRM of Subject 2 was relatively thin so that it was
not possible to get reliable tracking of this muscle
region.
Both horizontal EOMs and the ON of Subject 2 were
further investigated in more details. The top Panel of
Figure 4 shows the LRM of this subject. This muscle was
divided into three segments of equal lengths (at straight-
ahead gaze). The middle segment of the muscle (muscle
belly) elongated twice as much as the anterior and the
posterior segments. This result is expected, since the
anterior and the posterior segments of the muscles contain
more tendon that cannot stretch as easily as muscle tissue.
The larger standard deviation of the five polylines at the
15th time frame in the middle and posterior part of the
LRM segments was due to partial volume effects that
appeared at the interface of tissues with different move-
ments. These effects occurred also in the anterior part of
the EOM in other subjects.
Relative to the posterior and anterior part of the muscle,
the middle segment of the LRM relaxed earlier and to a
larger extent in later time frames. In contrast to the
heterogeneously contracting LRM segments, the three ON
segments all had a similar degree of longitudinal
deformation (Figure 4Vcentral Panel). In this subject,
the standard deviations among the five polylines of the
ON segments were greater than in the LRM; however, this
was not the case for all subjects.
The bottom Panel of Figure 4 shows the heterogeneous
contraction among the three medial rectus segments. The
middle segment contracted ahead of the anterior and
posterior segments. This pattern of earlier contraction
extended even to the later time frames. The standard
deviation of the five polylines in each segment was
comparable to the segments of the LRM (Figure 4Vtop
Panel). In fact, the average of all subjects’ standard
deviation at each single time frame was smaller than 3%
for each segment of the three tissues.
The Figures 5 and 6 summarize the results in all seven
subjects. Figure 5 shows a clear separation among the
LRM elongation, the ON deformation, and the MRM
contraction. The standard deviation among the subjects
increased over time and was slightly more pronounced in
the LRM and the ON than in the MRM. This may be due
to the nonlinearity of the deformation scale. Since the
standard deviation of the subjects was relatively small,
there was a statistically significant separation of the two
horizontal EOMs after the third time frame (correspond-
ing to a gaze deviation of 1.9 deg from the 20 deg right
gaze; see * at 140 ms), and of the three tissues from each
other after the fourth time frame (corresponding to a gaze
deviation of 4.2 deg from the 20 deg right gaze; see ** at
210 ms).
Figure 6 depicts the heterogeneous longitudinal seg-
mental deformation of the LRM (top Panel), the ON
Figure 3. (A, B) Change of length relative to the tissue length at the first time frame at 20 deg right gaze. The five polylines of the relaxing
lateral rectus muscle (LRM), the optic nerve (ON), and the contracting medial rectus muscle (MRM) of the right eye are averaged. The
tissue deformations of two subjects are depicted during 1 s (from 20 deg right to 20 deg left gaze). The movements of the muscles and the
optic nerve were distinguishable from each other (error bars: Tone standard deviation). (C, D) Length of the polylines of each tissue
(of Subjects 1 and 2) averaged for the five polylines at the first and last time frame (errors: Tone standard deviation).
Piccirelli et al. 5Journal of Vision (2007) 7(14):5, 1–10

(central Panel), and the MRM (bottom Panel). All length
changes were expressed relative to the averaged length of
the first two time frames and averaged for all seven
subjects. The greatest relaxation of the LRM was present
in the middle segment of the muscle and was greater in
the orbital apex than in the anterior segment. The
difference between the relaxation of the different muscle
segments was statistically significant, so that the muscle
segments could be distinguished after just a brief period of
eye movement. Examining the LRM (Figure 6Vtop
Panel), the anterior segment was statistically significantly
distinguishable from the other two segments from the
fourth time frame on (210 ms, see *). From the ninth time
frame on (560 ms, see **), all three segments were
statistically significantly distinguishable from each other.
The separation among the ON segments was not statisti-
cally significant in all time frames (Figure 6Vcentral
Panel). Considering the MRM (Figure 6Vbottom Panel),
the middle segment was statistically significantly stron-
ger contracting than the other two segments from the
sixth time frame on (350 ms, see *). From the seventh
time frame on (420 ms, see **), all three segments
were statistically significantly distinguishable from each
other.
In summary, the posterior and the middle segments of
the LRM were the earliest to relax, whereas the middle
segment was the earliest to contract in the MRM.
Figure 4. Change of relative tissue length of Subject 2 averaged
for the five polylines. The lateral rectus muscle (LRM, top Panel),
the optic nerve (ON, central Panel), and the contracting medial
rectus muscle (MRM, bottom Panel) are divided into three
segments. The lengths of the segment closest to the eye globe
(anterior, brown), the muscle belly or the optic nerve mid portion
(middle, blue), and the orbital apex segment (posterior, black) are
analyzed. In both horizontal rectus muscles, the middle segments
can be distinguished from the anterior and posterior segments. In
the optic nerve, however, all segments were similarly distorted
(error bars: Tone standard deviation).
Figure 5. Change of length relative to the length at the first time
frame for the relaxing lateral rectus muscle (LRM), the optic nerve
(ON), and the contracting medial rectus muscle (MRM) during 1 s
of eye movement, averaged over the seven subjects. The
standard deviation among the subjects was small. A single
asterisk (*) indicates that the LRM and MRM were statistically
significantly distinguished from the third time frame on corre-
sponding to a gaze deviation of 1.9 deg. Two asterisks (**)
indicate that all three tissues were statistically significantly
distinguished from the fourth time frame on corresponding to a
gaze deviation of less than 4.2 deg (error bars: Tone standard
deviation).
Piccirelli et al. 6Journal of Vision (2007) 7(14):5, 1–10

Discussion
Main findings
Motion-encoded MRI can be successfully used to
analyze patterns of EOM contraction during smooth
pursuit eye movement. Because the variability among
healthy subjects was small, statistically significant differ-
ences among the anterior, middle, and posterior segments
of the EOMs could be identified. Greater deformation of
the muscles was seen nearer to the orbital apex than to the
anterior segment of the muscle. As a control, the
deformation along the ON was homogeneous.
In both horizontal EOMs, the anterior segment was the
last segment that deformed. This may be due to the great
percentage of tendon within this segment. Nevertheless,
there was a slightly different distribution of the pattern
of deformation between the lateral and medial rectus
muscles. The posterior segment of the LRM relaxed
before the comparable segment in the MRM contracted.
One can speculate about the heterogeneous pattern of
deformation between the two muscles. The dynamics of
the deformation pattern of a muscle could be different
during contraction and relaxation. Future experiments,
including the analysis of the eye movement in the other
(left-to-right) horizontal direction, may find such a
difference.
Image acquisitions and postprocessing of
motion-encoded MRI
One advantage of the motion encoding technique in
comparison to conventional MRI is that tissue deforma-
tion can be resolved. Since the magnetic property of the
orbital tissues is modulated, it can be used as marker for
the tissue deformation. Moreover, this method is non-
invasive, i.e., no physical markers are necessary to track
the deformation of orbital tissue while the eye is moving.
Since motion-encoded MRI resolves motion and hence
does not need to resolve tissue borders, supramillimetric
resolution imaging was sufficient for separating the
motion of the three EOM segments during smooth pursuit
eye movement. The signal to noise ratio can be improved
by employing a small coil, as has been used for tracking
blood flow (Morse & Singer, 1970) and for orbital
imaging (Schenck et al., 1985). In addition, the relatively
small sensitivity zone of a small coil allows using a small
field of view that further shortens image acquisition time.
The signal to noise ratio drop with depth (due to the small
diameter of the coil) was more than compensated by its
greater sensitivity.
The standard deviations of the tissue deformation were
small in each subject, demonstrating the reproducibility of
the automatic tracking of the polylines. The five polylines
in each muscle did not constitute independent measures.
Figure 6. Change of relative tissue length averaged for the seven
subjects. The lateral rectus muscle (LRM, top Panel), the optic
nerve (ON, central Panel), and the contracting medial rectus
muscle (MRM, bottom Panel) are divided into three segments.
The segment closest to the eye globe (anterior, brown), the
muscle belly or the optic nerve mid portion (middle, blue), and the
orbital apex segment (posterior, black) are depicted. The three
segments can be distinguished from each other for both muscles,
but are similar for the ON (error bars: Tone standard deviation). A
single asterisk (*) indicates that, from that time frame on, one
segment was statistically significantly distinguishable from the
other two. Two asterisks (**) indicate that all three segments were
statistically significantly distinguishable from each other.
Piccirelli et al. 7Journal of Vision (2007) 7(14):5, 1–10

The reproducibility of the tissue point tracking wors-
ened in the time frames at the end of the tracking. Due to
the drop of signal to noise ratio with a greater number of
time frames, the acquisition was limited only to the right-
to-left gaze phase of the horizontal sinusoidal movement.
A possible solution to enhance the tracking quality would
be to track the five polylines as a whole and not as
separated entities. This should reduce the noise sensitivity
of the tracking and may also reduce the acquisition time or
allow a greater number of segments along the muscles to
be resolved. In this context, it is noteworthy that a good
shimming (homogenization of the static magnetic field)
prior to image acquisition is essential when using the
described technique. The gradient echo EPI sequence used
for faster acquisition in this study is sensitive to the
magnetic susceptibility artifacts (Chen, Chandna, &
Abernethy, 2005), which are prominently present in the
orbit (Herrick, Hayman, Taber, DiazMarchan, & Kuo,
1997).
The polyline tracking software TagTrack works with
hand drawn polylines on the motion-encoded image. To
ascertain that the polylines lay on the expected tissue, we
took advantage of anatomical images and realigned the
polylines if they were not on the muscles of interest or the
ON. Drawing polylines on the anatomical image, and a
subsequent automatic transfer to the motion-encoded
image, would improve the position accuracy of the
polylines.
Motion-encoded MRI acquires images during several
time frames (see Figure 1) during a single acquisition. To
maintain a reasonable scan time, the scan resolution was
ten times lower than the one used for high-resolution
conventional MRI (Kau, Tsai, Ortube, & Demer, 2007).
An evident limitation of the low resolution is the inability
to resolve diverse structures within EOMs, as, for
example, the global and orbital EOM layers. Nevertheless,
as about twenty pixels are acquired in the longitudinal
direction of the muscles, this method was capable of
resolving three muscle segments. The limited resolution
mainly affects the interface of differently moving tissues.
Since the anterior segments of the EOMs are thin, their
tracking was more difficult than the middle or posterior
segments (Figure 3). To diminish the partial volume effect
in the outer layer of the EOMs, the signal was optimized
by taking advantage of the specific T1 of muscles. Note
that the tagline distance is not important for resolution.
The tagline distance gives only the period of the phase
information. The resolution of the phase information used
for the tracking is limited by the scan resolution and the
size of the postprocessing filter.
Detection of motion of extraocular muscles
Several prior studies attempted to resolve the deforma-
tion of the orbital tissues induced by different eye
positions. CT (Simonsz, Harting, De waal, & Verbeeten,
1985) and MRI studies segmented static coronal images to
determine the contraction pattern of the extraocular recti
muscles (Miller, 1989) and the two oblique muscles
(Kono & Demer, 2003) by measuring the cross-sectional
area of the muscles. Coronal images represent a fixed
section of space through which the EOMs are shown at a
specific eye position. One should be aware that motion of
EOMs through the fixed planes of images can result in
apparent motion and thickening of a region simply due to
the geometric effect of tangentially cutting through the
curved surfaces of the EOMs as they are pulled through
the imaging plane. Although pull-through effects may
change the apparent thickness of the EOMs, they will not
change the spacing of the tagging lines as used in the
present study, permitting pull-through effects to be
distinguished from true contraction.
Another attempt to depict movement in the orbit was to
image eye positions in transversal planes (Abra`moff, Van
Gils, Jansen, & Mourits, 2000; Botha et al., 2005). Such a
method is valid for following tissue borders and not for
describing contractility within tissues, since heterogene-
ous deformation inside homogeneously contrasted tissues
cannot be assigned with certainty due to the absence of
landmarks for deformation recognition. Recognizing the
necessity of such landmarks, Miller et al. (2006)
implanted gold beads inside monkey orbits, that served
to successfully demonstrate orbital soft tissue deformation
in static CT acquisitions.
All these methods used static images. However, the
nomenclature is not standardized. Shin, Demer, and
Rosenbaum (1996) stated that “dynamic imaging refers
to viewing the EOMs during different maintained gaze
positions.” Abra`moff et al. (2000) and Bailey et al. (1993)
named the same imaging method “cine imaging.”
Recently, Kau et al. (2007) designated it as “multi
positional imaging,” which seems most appropriate to
describe multiple static imaging, as it is not associated
with motion. On the other hand, “motion-encoded MRI”
refers to imaging using a motion encoding feature and
multiple time frames acquired during eye movement. In
this paper, we restricted the analysis to the change of
muscle length. The muscle path and contraction patterns
may differ during the actual eye movement in compar-
ison to the static condition, as the force and torque
resultants do not need to be null during the eye
movement (Weber, Bockisch, Bergamin, Landau, &
Straumann, 2005).
In the future, we plan to upgrade the tracking algorithm
of the polylines in order to sample the EOMs in more than
three segments. To further enhance the spatial resolution
of the images, a specialized coil array will be needed. Our
method has the potential of improving the temporal
resolution without the need to adapt other scan parame-
ters. With a better temporal resolution, EOMs contraction
may be imaged during saccades. To reach this goal, the
accuracy of the eye movement needs to be verified with
simultaneous eye movement measurements.
Piccirelli et al. 8Journal of Vision (2007) 7(14):5, 1–10

Conclusion
Motion-encoded MRI of the orbit during eye movement
is a versatile and noninvasive method which permits
detailed studies of normal and pathologic EOMs in vivo
with good resolution of the deformation of soft orbital
tissues, even in regions where physical markers or devices
cannot easily be implanted, such as the orbital apex. This
methodology may have a number of potential clinical
benefits by identifying specific patterns of deformation in
different diseases of the orbital muscles (e.g., thyroid
orbitopathy, orbital pseudotumor) or in conditions which
alter the innervation to eye muscles (e.g., acquired or
congenital aberrant innervation).
Acknowledgments
We thank the Swiss National Science Foundation (SNF
#3100AO-102197) for grant support. This work was
rewarded by the Gesine Mohn Travel Grant Award
(ARVO 2006).
Commercial relationships: none.
Corresponding author: Oliver Bergamin.
Email: Oliver.Bergamin-remy@usz.ch.
Address: Department of Ophthalmology, University Hos-
pital of Zurich, Frauenklinikstrasse 24, CH-8091 Zurich,
Switzerland.
References
Abra`moff, M. D., Van Gils, A. P., Jansen, G. H., &
Mourits, M. P. (2000). MRI dynamic color mapping:
A new quantitative technique for imaging soft tissue
motion in the orbit. Investigative Ophthalmology &
Visual Science, 41, 3256–3260. [PubMed] [Article]
Axel, L., & Dougherty, L. (1989). MR imaging of motion
with spatial modulation of magnetization. Radiology,
171, 841–845. [PubMed] [Article]
Bailey, C. C., Kabala, J., Laitt, R., Weston, M.,
Goddard, P., Hoh, H. B., et al. (1993). Cine magnetic
resonance imaging of eye movements. Eye, 7, 691–693.
[PubMed]
Botha, C., de Graaf, T., Root, R., Wielopolski, P.,
Schutte, S., Post, F., et al. (2005). Time-varying
three-dimensional vector field visualisation for the
analysis of retrobulbar fat mobility during eye
motion. Proceedings of the 11th Annual Conference
of the Advanced School for Computing and Imaging
(pp. 1–5).
Chen, S. I., Chandna, A., & Abernethy, L. J. (2005).
Magnetic susceptibility artifact in orbital magnetic
resonance imaging. Strabismus, 13, 1–3. [PubMed]
[Article]
Fischer, S. E., Mckinnon, G. C., Maier, S. E., & Boesiger,
P. (1993). Improved myocardial tagging contrast.
Magnetic Resonance in Medicine, 30, 191–200.
[PubMed]
Herrick, R. C., Hayman, L. A., Taber, K. H., Diaz-Marchan,
P. J., & Kuo, M. D. (1997). Artifacts and pitfalls in MR
imaging of the orbit: A clinical review. Radiographics,
17, 707–724. [PubMed] [Article]
Kau, H. C., Tsai, C. C., Ortube, M. C., & Demer, J. L.
(2007). High-resolution magnetic resonance imag-
ing of the extraocular muscles and nerves demon-
strates various etiologies of third nerve palsy.
American Journal of Ophthalmology, 143, 280–287.
[PubMed]
Kaufmann, H., & Decker, W. d. (1995). Strabismus
(2nd ed., p. 37). Stuttgart: Enke.
Kono, R., & Demer, J. L. (2003). Magnetic resonance
imaging of the functional anatomy of the inferior
oblique muscle in superior oblique palsy. Ophthal-
mology, 110, 1219–1229. [PubMed]
Meyer, C. H., Lasker, A. G., & Robinson, D. A. (1985).
The upper limit of human smooth pursuit velocity.
Vision Research, 25, 561–563. [PubMed]
Miller, J. M. (1989). Functional anatomy of normal human
rectus muscles. Vision Research, 29, 223–240.
[PubMed]
Miller, J. M., Rossi, E. A., Wiesmair, M., Alexander, D. E.,
& Gallo, O. (2006). Stability of gold bead tissue
markers. Journal of Vision, 6(5):6, 616–624, http://
journalofvision.org/6/5/6/, doi:10.1167/6.5.6.
[PubMed] [Article]
Morse, O. C., & Singer, J. R. (1970). Blood velocity
measurements in intact subjects. Science, 170,
440–441. [PubMed]
Osman, N. F., Kerwin, W. S., McVeigh, E. R., &
Prince, J. L. (1999). Cardiac motion tracking using
CINE harmonic phase (HARP) magnetic resonance
imaging. Magnetic Resonance in Medicine, 42,
1048–1060. [PubMed]
Ryf, S., Tsao, J., Schwitter, J., Stuessi, A., & Boesiger, P.
(2004). Peak-combination HARP: A method to
correct for phase errors in HARP. Journal of
Magnetic Resonance Imaging, 20, 874–880.
[PubMed]
Schenck, J. F., Hart, H. R., Jr., Foster, T. H., Edelstein,
W. A., Bottomley, P. A., Redington, R.W., et al. (1985).
Improved MR imaging of the orbit at 1.5 T with
Piccirelli et al. 9Journal of Vision (2007) 7(14):5, 1–10

surface coils. American Journal of Roentgenology,
144, 1033–1036. [PubMed] [Article]
Shin, G. S., Demer, J. L., & Rosenbaum, A. L. (1996).
High resolution, dynamic, magnetic resonance imag-
ing in complicated strabismus. Journal of Pediatric
Ophthalmology and Strabismus, 33, 282–290.
[PubMed]
Simonsz, H. J., Harting, F., de Waal, B. J., & Verbeeten,
B. W. (1985). Sideways displacement and curved
path of recti eye muscles. Archives of Ophthalmology,
103, 124–128. [PubMed]
Soellinger, M., Ryf, S., Boesiger, P., & Kozerke, S.
(2007). Assessment of human brain motion using
CSPAMM. Journal of Magnetic Resonance Imaging,
25, 709–714. [PubMed]
Suryan, G. (1951). Nuclear resonance in flowing liquids.
Proceedings of the Indian Academy of Science A, 33,
107–111.
Weber, K. P., Bockisch, C. J., Bergamin, O., Landau, K.,
& Straumann, D. (2005). Modulation of saccade
curvature during Bielsehowsky head-tilt testing in
patients with unilateral trochlear nerve palsy and
healthy subjects. Journal of the Neurological Sciences,
238, S48.
Yee, R. D., Goldberg, R. A., Jones, O. W., Baloh, R. W., &
Honrubia, V. (1983). Effect of eccentric gaze on pursuit.
Investigative Ophthalmology & Visual Science, 24,
1108–1114. [PubMed] [Article]
Zerhouni, E. A., Parish, D. M., Rogers, W. J., Yang, A., &
Shapiro, E. P. (1988). Human heart: Tagging with
MR imagingVA method for noninvasive assessment
of myocardial motion. Radiology, 169, 59–63.
[PubMed] [Article]
Piccirelli et al. 10Journal of Vision (2007) 7(14):5, 1–10