Eye dominance in the visual cortex using functional MRI at 1.5 T: An alternative method

000-000/Liu120170

3/7/02

11:49 AM

Page 40

Eye Dominance in the Visual Cortex Using Functional MRI at 1.5 T: An Alternative Method Grant T. Liu, MD, Atsushi Miki, MD, PhD, Zachariah Goldsmith, Theo G. M. van Erp, MA, Ellie Francis, PhD, OD, Graham E. Quinn, MD, Edward J. Modestino, MLA, Gabrielle R. Bonhomme, MD, and John C. Haselgrove, PhD Purpose: To develop a functional MRI method for producing eye dominance histograms in humans at 1.5 Tesla (T). Methods: In the first set of experiments, 8 normal persons were tested. The eye dominance of each voxel within the person’s visually activated primary visual cortex was determined with Student t statistics during a left eye versus right eye contrast. Eye dominance distribution was plotted, and the mean t statistic was used to describe the histogram asymmetry. In the second set of experiments, the effect of monocular optical blur and decreased luminance via filter was studied, and eye dominance distributions were similarly determined. Results: The eye dominance histogram in each of the 8 normals was approximately symmetric; the average mean t value was +0.13. All 4 subjects with the right eye blurred had histograms approximately symmetric or slightly shifted toward the left eye (average mean t = +0.56), and all 4 subjects with the right eye filtered had histograms dramatically shifted toward the left eye (average mean t = +2.22). The average mean t for the group with the right eye filtered was significantly different from that of the other 2 groups (P < .0001). Conclusions: With noninvasive methods in normal persons, functional magnetic resonance imaging techniques at 1.5 T were able to characterize the distribution of eye dominance of voxels in primary visual cortex, based upon their t statistic in the left eye versus right eye contrast. The method is sensitive to filtering but relatively insensitive to visual blur. This approach may have a future use in the study of amblyopia in humans.(J AAPOS 2002;6:40-8)

n their landmark studies in experimental animals with amblyopia, Hubel and Wiesel1,2 used a histogram method that characterized the eye dominance of neurons in the visual cortex. During intracellular recordings in lightly anesthetized animals undergoing monocular visual stimulation, they determined whether neurons were driven primarily by the contralateral eye, the ipsilateral eye, or equally by both eyes. According to these criteria, neurons were assigned an ocular dominance num-

I

From the Children’s Hospital of Philadelphia, Functional MRI Research Unit, and the University of Pennsylvania School of Medicine, Philadelphia. This study was supported, in part, by The Children’s Hospital of Philadelphia Ethel Foerderer Fund for Excellence (Drs Liu and Haselgrove), the University of Pennsylvania Research Foundation, the Knights Templar Eye Foundation, and the F. M. Kirby Foundation (Dr Liu), and Prevent Blindness America GA98015 and PD98017 (Drs Liu and Miki). Presented, in part, at the Association for Research in Vision and Ophthalmology (ARVO) meeting, Fort Lauderdale, Fla, May 2000. Submitted December 18, 2000. Revision accepted September 17, 2001. Reprint requests: Dr Grant T. Liu, Division of Neuro-ophthalmology, Department of Neurology, Hospital of the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104 (e-mail, [email protected].) Copyright © 2002 by the American Association for Pediatric Ophthalmology and Strabismus. 1091-8531/2002/$35.00 + 0 75/1/120170 doi:10/1067/mpa.2002.120170

40

February 2002

ber from 1 (for completely contralaterally driven) to 7 (completely ipsilaterally driven), with 2 through 6 designating neurons with various degrees of binocular input. Histograms with the eye dominance along the x-axis and the numbers or percentages of total cells were plotted along the y-axis. They used this method to show, for instance, a reduction in the proportion of cells driven by an eye with amblyopia caused by suturing of the eyelid in early life. This was accompanied by an increase in the number of cells driven by the unaffected normal eye. The eye dominance histogram method was used by others3-5 and is still employed by some investigators studying experimental amblyopia in nonhuman primates.6,7 A similar approach to amblyopia in humans would require an alternative, safe, noninvasive method for studying eye dominance in the visual cortex. Cell-by-cell recordings are prohibitive in humans, and intraoperative surface and penetrating electrodes, used in epilepsy patients, for instance, are impractical in large numbers of patients because they require neurosurgery. Visual evoked potentials (VEPs) and magnetoencephalography (MEG) lack sufficient spatial resolution for such an analysis, and localization within area 17 is difficult with VEPs. Positron emission tomography (PET) and single photon emission Journal of AAPOS

000-000/Liu120170

3/7/02

11:49 AM

Page 41

Journal of AAPOS Volume 6 Number 1 February 2002

computerized tomography (SPECT) studies have adequate spatial resolution, but they require injection or inhalation of radioactive substances. Furthermore, PET and SPECT lack the temporal resolution for a meaningful intereye comparison of visual activation in single experiments. Tandem experiments, in which activation of 1 eye is performed and the other eye is tested minutes or hours later, introduce methodological confounds such as head movement. Functional magnetic resonance imaging (fMRI), which is noninvasive and nonradioactive, offers the temporal and spatial resolution for studying the eye dominance of small volumes of visual cortex. To our knowledge, only 1 fMRI study8 has previously attempted to replicate Hubel and Wiesel’s eye dominance histograms. However, the method of Menon et al uses high field strength at 4.0 Tesla (T), which is not widely available. Therefore, we developed an alternative fMRI method at 1.5 T for studying eye dominance distributions in human primary visual cortex. The methodology does not detect ocular dominance column (ODC) size. Instead, it measures the relative activation of small volumes of visual cortex by stimulating the left or the right eye.

METHODS Sixteen normal volunteers were studied, and each signed a consent form approved by the Institutional Review Board of The Children’s Hospital of Philadelphia. None had an abnormal neurologic or ophthalmic history. Each had best corrected visual acuity of at least 20/20 in each eye, normal confrontational visual fields, and normal color vision when tested with Ishihara color plates. The age range was 10 to 40 years; 9 were male, and 7 were female. Preparation of Subjects Persons using contact lenses for distance correction wore them during the testing. Otherwise, when the uncorrected vision was subnormal, the subject’s manifest refraction or present prescription, whichever allowed the person to see 20/20 with each eye, was given using a nonmetallic lens set in a plastic frame (adapted from a Titmus stereo test by Gulden Ophthalmics, Abington, Pa). Both spherical and cylindrical corrections were provided when necessary. To facilitate monocular stimulation, a red Wratten #59 filter (Kodak, Rochester, NY) was placed in the plastic frame over the right eye, and a green Wratten #23 filter (Kodak) was placed in the plastic frame over the left eye. With guidance from a photometer, to equate the luminance of the stimuli between the 2 eyes during periods of monocular stimulation, an additional 0.9 log neutral density filter was placed in the plastic frame over the left eye. A drawback of this, however, was that during binocular stimulation periods, each eye did not receive identical stimuli: the right eye stimulus was brighter than the left, and the left eye received a green stimulus and the right a

Liu et al

41

red one. This baseline setup was used for all experiments described below. Eight normal subjects were tested in this manner. Normal subjects would be expected to have relatively symmetric histograms, which unfortunately could also be produced by chance in a t test. To prove that the histogram reflected visual cortex activation and not chance, we tested whether perturbing the vision of 1 eye would shift the histogram. Therefore, the remaining volunteers had their vision degraded by 1 of 2 methods: visual blur (mimicking refractive error) or filtering with a 2.0 log neutral density filter (darkening the stimulus). In 4 persons, a phorometer was used to determine an amount of plus lens over their normal prescriptions, which blurred the right eye to 20/200 at distance. The amount ranged from +2.50 to +3.50 in the 4 subjects, and this amount was added to their prescriptions in the plastic frame for the testing. In 4 others, an additional log 2.0 neutral density filter was placed over the right eye for the fMRI testing. The visual acuity with the filter over these eyes ranged between 20/20 and 20/40. The subjects’ heads were padded firmly with foam in the quadrature head coil to discourage motion. They were instructed to keep their heads still at all times and their eyelids open during the periods of visual stimulation. A mirror was placed above the opening of the head coil and angled at 45° so that the subjects could see a groundglass screen (Resonance Technologies; Van Nuys, Calif) placed at their feet. Once the subjects were positioned in the MRI bore, dark material was placed on the sides of the bore opening to block their peripheral vision, so that only the ground-glass screen could be seen. Image Acquisition Sequences The magnet was shimmed with the Siemens automatic shimming routine that uses first- and second-order gradients. A “slice prescription procedure”9 was performed. First, a coronal scout image was obtained, and oblique axial images perpendicular to the midline of this coronal image were prescribed. Subsequently, sagittal images perpendicular to the midline of the oblique axial images were taken. Finally, the 16 oblique axial planes encompassing the visual cortex for the anatomic and functional images were positioned parallel to the calcarine fissure. All subjects were studied with the same image parameters to prevent a voxel size-dependent variation of signal to noise from influencing the results. A voxel is a small 3dimensional volume in which individual magnetic resonance measurements are taken. Each functional study was preceded by a T1 weighted spin echo sequence (TR/TE, 600/20 ms; matrix, 256*256; slice thickness, 3 mm; interslice space, 2 mm; field of view (FOV) = 240 mm × 240 mm [resolution = 0.94 × 0.94 × 3 mm3]). These images were called the anatomical images. Thereafter, functional images were acquired in identical planes and FOV with a T2* weighted echo-planar image sequence with a 64*64

000-000/Liu120170

42

3/7/02

11:49 AM

Page 42

Journal of AAPOS Volume 6 Number 1 February 2002

Liu et al

FIG 1. Arrangement of red Wratten #59 and green Wratten #23 filters to achieve monocular stimulation. Red filter was placed over right eye, and green filter over left for all conditions. During binocular stimulation, no filter was placed in front of projector lens. When checkerboard was presented, red filter was placed in front of projector lens to stimulate right eye only. Similarly, green filter was placed in front of projector lens to stimulate only left eye. A 0.9 log neutral density filter (not shown) was also placed over left eye to make stimuli equiluminant during monocular conditions. P, projector; L, left; R, right. TABLE 1. Visual stimulation paradigm. There were 16 epochs, each consisting of 10 images (33.6 sec). Experiment included 4 conditions (R, right eye stimulation; L, left eye stimulation; B, both eyes stimulated; and O, no stimulation) which were grouped, and each group was presented 4 times to each subject. Eight normal subjects were assigned 8 different randomly generated and balanced Latin squares that defined order of presentation. Table shows resulting order of presentation for each subject. Four subjects whose right eyes were optically blurred and 4 whose right eyes were filtered were assigned S1 through S4 Epoch Subject

1a

1b

1c

1d

2a

2b

2c

2d

3a

3b

3c

3d

4a

4b

4c

4d

S1 S2 S3 S4 S5 S6 S7 S8

B R B R O B O B

O L O L R L L R

L O R B L R B O

R B L O B O R L

O L O L R L L R

B R B R O B O B

R B L O B O R L

L O R B L R B O

R B L O B O R L

L O R B L R B O

O L O L R L L R

B R B R O B O B

L O R B L R B O

R B L O B O R L

B R B R O B O B

O L O L R L L R

matrix, and 5 mm thick slices without interslice gaps (resolution = 3.75 × 3.75 × 5 mm3), and TE, 29 ms; 90° flip angle; and TR, 3 seconds. For each functional experiment, 160 sets of 16 images were acquired, and stimuli were applied during 16 successive epochs, each consisting of 10 image sets. Visual Stimulation The visual stimulus consisted of a black-and-white checkerboard drawn with Adobe Photoshop (Niles, Calif)

and then displayed by a Macintosh G3 running Macstim software (David Darby, West Melbourne, Australia). The checkerboard was projected onto the ground-glass screen by a video projector (Sharp XG-NV4SU, Mahwah, NJ) with a 640 × 480 pixel resolution. Each check subtended a visual angle of 0.5° (1 cycle/degree), and the entire checkerboard subtended 10.6° × 8.0°. The black-andwhite check contrast was 97%, and the mean luminance of the entire checkerboard was 174.8 cd/m2. A white fixation cross was placed in the middle of the checkerboard to

000-000/Liu120170

3/7/02

11:49 AM

Page 43

Journal of AAPOS Volume 6 Number 1 February 2002

Liu et al

43

TABLE 2. Conversion values for t statistics to eye dominance number T statistic Ocular dominance number

> 4.0 1

4.0 ≥ t > 2.4

2.4 ≥ t > 0.8

0.8 ≥ t ≥ -0.8

-0.8 > t ≥ -2.4

-2.4 > t ≥ -4.0

2

3

4

5

6

t < -4.0 7

enhance central fixation. The checkerboard reversed at 8 Hz, a rate known to provide maximal stimulation.10 A red Wratten #59 filter was placed in front of the projector lens to stimulate the right eye only when the checkerboard was presented (Fig 1). Similarly, a green Wratten #23 filter was placed in front of the projector lens to stimulate only the left eye. During binocular stimulation and rest periods, no filter was placed in front of the projector lens. The rest stimulus was a dark screen with only the white central-fixation cross. Functional Paradigm Each subject underwent an experiment consisting of 4 conditions: L, stimulation of the left eye; R, stimulation of the right eye; B, stimulation of both eyes; and O, no stimulation. The 4 conditions were grouped and repeated 4 times (4 × 4 = 16 epochs). The order of the conditions within each grouping was randomized to minimize any possible intercondition dependence (Table 1). The order of presentation was chosen by constructing a 4 × 4 Latin square (not shown). The 8 normal (unblurred, unfiltered) subjects were assigned 8 different randomly generated and balanced Latin squares that defined the order of presentation. The columns of the original Latin square were permuted to give a new Latin square that defined the order of presentation for the first subject. This order was balanced in the order for the second subject, who had the reverse permutation of the columns. The third and fourth subjects had balanced permutations of the rows of the original Latin square. The fifth through eighth subjects had orders of presentation that were defined in the same manner but with different randomly generated permutations of columns and rows of the original Latin square. The 4 optically blurred and the 4 filtered subjects underwent the same paradigms as the first 4 normal subjects. An MRI-compatible fiberoptic pushbutton response pad (Current Designs; Inc, Philadelphia, Pa) was used to ensure that the subject stayed awake and to enhance attention to the visual stimulus. The pushbuttons were connected to an interface with light-emitting diode (LED) lights monitored by the testers. The subject was asked to push the button once quickly when the stimulus changed, such as when a red checkerboard changed to a green one. Thus, the button should have been pushed at the beginning of each condition. Only studies with at least 90% correct responses were used. Analysis—Generation of Functional Images All images were downloaded from the MRI scanner and then analyzed on a Sun SPARC workstation (Sun

FIG 2. Region of interest (ROI) for area 17, drawn on T1 template using Talairach and Tourneaux atlas 11 as guide, on 1 coronal view.

Microsystems; Palo Alto, Calif). A combination of IDL (Interactive Data Language; Research Systems Inc, Colorado) and SPM96 (Wellcome Department of Cognitive Neurology; London, UK) packages was used. The first 4 scans of each functional experiment were discarded to eliminate magnetic saturation effects. To correct for motion, functional images of each subject were realigned with SPM96 to the sixth image by a 6-parameter (3 translations and 3 rotations) rigid body transformation. The images were then spatially normalized with SPM96 into the anatomical space of Talairach and Tournoux11 by minimizing the sum-of-squares difference between the functional images and the SPM96 EPI template, with an 8-parameter (3 translations, 3 rotations, and x and y scalings) affine transformation. The voxel size was maintained at 3.75 × 3.75 × 5 mm3. The spatial normalization was not performed with a nonlinear transformation because of the relatively limited volume coverage in the z direction. Statistical Maps Spatial smoothing, a high-pass filter, and temporal smoothing with SPM96 were used during the analysis of both eyes stimulated versus off contrast. Global normalization was not performed because of the large areas of

000-000/Liu120170

44

3/7/02

11:49 AM

Page 44

Journal of AAPOS Volume 6 Number 1 February 2002

Liu et al

A

B FIG 3. Signal intensity vs. image set (time) from individual voxels in 2 subjects. Data sets had been motion corrected and transformed to Talairach space. A, From subject S1. This voxel’s activation correlated roughly equally with right and left eye stimulation (t = +0.46). B, From subject S3, with log 2.0 filter placed over right eye. This voxel’s activation correlated more with left than with right eye stimulation (t = +5.89). B, both eye stimulation; O, off (dark); R, right eye stimulation; and L, left eye stimulation. Epochs consisted of 10 image sets of each condition.

activation. These manipulations were not performed in the analysis of the left versus right contrast to preserve maximally the integrity of the signal characteristics of each voxel and maintain spatial resolution. Therefore, the statistics portion of SPM96, which requires spatially smoothed data sets, could not be used in this part of the analysis, and, instead, a statistical analysis with programs in IDL was implemented. The first 2 images of each 10image epoch were discarded to remove data taken while the physiologic state of the brain was expected to be still changing. Statistical parametric maps for the left versus right contrast were constructed with Student’s t statistic. A region of interest (ROI) for area 17 (striate cortex) was drawn on the T1 template using the Talairach and Tournoux atlas11 as a guide (Fig 2). This area 17 ROI was used for all subjects. In each subject, a left eye-versus-right eye contrast map in area 17 was performed in 3 steps: (1) Visually activated voxels in the occipital lobe were identified as those having a z score more than 4.0 during both versus off contrast.

(2) A functional area 17 was then defined as voxels that were activated in the occipital lobe (step 1) but also contained in the area 17 ROI. (3) The left-versus-right contrast map was determined for the whole brain, but only voxels in the functional area 17 from step 2 were used for further analysis. Histograms Only voxels from step 3 above were considered. An eye dominance number from 1 (left eye dominant) through 7 (right eye dominant) was assigned to each voxel, based linearly on their t statistic (Table 2). Any voxel with a t statistic between +0.5 and -0.5 was considered equally influenced by both eyes and assigned an eye dominance number of 4. Voxels with t statistics greater than +0.8 were considered to be dominated primarily by the left eye (eye dominance number 1 for t > 4.0, 2 for t ≤ 4.0 and > 2.4, 3 for t ≤ 2.4 and > 0.8). Voxels with t statistics less than -0.5 were considered to be dominated primarily by the right eye (eye dominance number 7 for t < -4.0, 6 for t ≥ -4.0 and

000-000/Liu120170

3/7/02

11:50 AM

Page 45

Journal of AAPOS Volume 6 Number 1 February 2002

Liu et al

45

A

B

C FIG 4. Eye dominance histograms for A, Unblurred, unfiltered subjects; B, subjects with right eye blurred optically to 20/200, and C, subjects with 2.0 log neutral density filter placed over right eye. Note in A the histograms are relatively symmetric; in B they are shifted slightly toward left eye; but in C there is dramatic shift in histogram toward left eye (S1 to S8, subject number; n, number of voxels analyzed; m.t., mean t statistic; y-axis, percentage of n; x-axis, eye dominance number (1, left eye dominant; 7, right eye dominant)).

< -2.4, 5 for t ≥ -2.4 and < -0.8). For example, a voxel with a t statistic of +6.5 was assigned an eye dominance number of 1, and a voxel with a t statistic of -0.033 was assigned an eye dominance number of 4. Eye dominance histograms for each subject and for the unblurred/unfiltered, optically blurred, and filtered groups were plotted with y = the percentage of activated voxels, and x = eye dominance number. The mean t statistic for each subject was used to describe the histogram asymmetry. Group Comparisons The group mean t statistics were compared with 1-way analysis of variance (ANOVA). The data consisted of a mean t statistic for each of the 16 subjects, where each subject was in 1 of the 3 groups (normal, blurred, filtered). A post-hoc comparison among the group means with the Scheffé method was performed.

RESULTS Examples of plots of signal intensity versus image set (time) are shown in Figure 3. Depicted are a voxel activated symmetrically (between eyes) in a unblurred, unfiltered normal subject and a voxel asymmetrically activated in a filtered subject. The eye dominance histogram in each of the 8 normals was approximately symmetric; the range of mean t statistics was -0.48 to +0.50 (Fig 4, A). The average mean t was +0.13 (SD, +0.34). All 4 subjects with right eye blurred had histograms approximately symmetric or slightly shifted toward the left eye (range of mean t values, +0.26 to +0.81; average mean t, +0.56; SD, +0.27) (Fig 4, B), while all 4 subjects with the right eye filtered had histograms dramatically shifted to the left eye (range of mean t, +1.35 to +2.86; average mean t, +2.22; SD, +0.66) (Fig 4, C).

000-000/Liu120170

46

3/7/02

11:50 AM

Page 46

Liu et al

FIG 5. Scatterplot of mean of t values versus group for 16 subjects. For each subject, t statistics were computed at each voxel and then averaged in regions of interest. T statistics were for left versus right contrast. Mean t is significantly greater in filtered group.

We found strong evidence for differences among the group means (F2,13 = 33.6, P < .0001). The post-hoc analysis revealed that the significant ANOVA F was due to significant differences between the group with the right eye filtered and each of the other 2 groups. There was a trend for the histograms of the group with the right eye blurred to be shifted toward the left eye, but the difference between the normal and the blurred groups was not significant. The scatterplot of mean of t values versus group for the 8 normal, 4 blurred, and 4 filtered subjects is shown in Figure 5.

DISCUSSION We have devised an fMRI method at 1.5 T which characterizes the eye dominance of small volumes in the primary visual cortex of normal human subjects, based on their t statistics in the left eye-versus-right eye contrast. In simpler terms, for each voxel in the visual cortex analyzed, we asked the question: is its visual activation influenced more by stimulation of the left or the right eye? Optical blurring of 1 eye shifts the histogram slightly, albeit not in a statistically significantly way, toward the normal eye. This is evidenced by the difference in the mean t statistics of the normal versus blurred groups. However, decreased luminance of the stimulation in 1 eye, created by a log 2.0 filter, causes a statistically significant large shift of the histogram toward the unfiltered eye. Menon et al8 used fMRI methods at 4 T to produce eye dominance histograms in primary visual cortex. As we did, they used blood oxygenation level dependent (BOLD) techniques, but they had the theoretical advantage of

Journal of AAPOS Volume 6 Number 1 February 2002

smaller voxel sizes (0.55 × 0.55 × 4 mm3), higher field strength, and the use of a surface coil. However, 4 T is not widely available, and there were several stimulation and postprocessing drawbacks of their technique: (1) Visual stimulation in their study was provided by a single round flashing red LED. Flash stimulation is probably inferior to pattern stimuli to study activation in visual cortex. (2) The subjects were instructed by headset to close the right or the left eye gently during the scanning to provide monocular stimulation. However, bright LEDs can still activate the visual cortex through closed eyelids,12 so this may not guarantee monocular stimulation. (3) They chose only 5 oblique 4-mm slices parallel to the calcarine fissure in which to obtain images. Because, medially, the distance from the dorsal edge of the upper bank to the ventral edge of the lower bank of the calcarine cortex may measure approximately 2.0 cm,13 and the calcarine fissure typically bends at least once, their slices may not capture the dorsal-ventral extent of the primary visual cortex. (4) The primary visual cortex V1 was defined by selecting statistically significant voxels from both eyes stimulated versus dark contrast. However, the result of this contrast would include striate and extrastriate cortices. For instance, visual areas V5 and V3a might be activated by the flashing aspect of the light. (5) No motion correction was used. It would have been desirable because their method analyzed signal intensities of small voxels, and this analysis would be confounded by even small amounts of movement. (6) They selected voxels for sorting by first determining which voxels had a significant difference between the right and the left stimulation periods. Then they sorted these voxels into dominance categories according to differences in signal between 1.5% and 4.5% for partial dominance, and more than 4.5% for monocularity. These small differences would be extremely sensitive to noise. Subsequently, Menon and Goodyear14 published a report in which they identified 2 different populations of voxels spaced 700 µm apart. These voxels were believed to be consistent with ODCs. Their stimulation paradigm was improved to include monocular and binocular checkerboard stimulation by a pair of liquid crystal shutter glasses, and they defined V1 by determining cortex that was sensitive to changes in luminance contrast. However, no histograms were constructed. Our method is an attractive alternative because it groups the dominance of voxels according to the correlation of their visual activation with either left or right stimulation, in the same spirit as the Hubel and Wiesel 7-point scale.1,2 Also, 1.5 T MRIs are more widely available. There are other advantages to our method which warrant discussion: Refractive error was corrected when necessary, because the subject’s current refractive correction or best manifest refraction was given, including sphere and cylinder, during the fMRI testing. The check size was purposely set at 1 cycle per degree, to ensure that a person with 20/200 visu-

000-000/Liu120170

3/7/02

11:50 AM

Page 47

Journal of AAPOS Volume 6 Number 1 February 2002

al acuity could still resolve the checkerboard. Smaller check sizes might have caused visual cortex activation to be more susceptible to the effects of visual blur or filtering; this could be investigated in future studies. Subjects viewing high-contrast large checks through +3.00 or +4.00 lenses have only mild attenuation of their visual evoked responses.15 Although our stimulus size was limited to 10.6° × 8.0°, primate studies alluded to earlier similarly sampled the visual cortex primarily in the foveal or macular regions.7 The combination of the red filter over the right eye, the green filter over the left, and alternating red and green filters over the projector lens (Fig 1) provided alternating, complete, monocular stimulation during a single testing session. This allowed us to avoid testing 1 eye with the other patched, taking the subject out of the scanner, switching the patch, and then testing the other eye. Testing in 2 sessions would have made comparison of activation within single voxels problematic because of difficulty with head repositioning and signal drift, for instance. Although using colored filters has the disadvantage of decreasing the total luminance of the visual stimulus to each eye, compared with having no filters in front of the eye or projector, we found this was an optimal method for obtaining monocular stimulation with a video projector. Because the histograms were fairly symmetric, with only slight shifts in the direction of either the left or the right eye, we believe the stimulation to each eye was relatively similar for the purposes of this study. Furthermore, qualitatively, we found little difference in the signal intensity during right and left eye stimulation in normal unblurred, unfiltered subjects (Fig 3, A). We considered many other options for monocular stimulation. For instance, we had attempted to use polarized filters in front of the eyes and the projector but had 2 problems: complete orthogonality of the filters was difficult to achieve, and a large filter in front of the projector had to be placed near the projector screen and then alternated with 1 orthogonal to it. Smooth transitions were difficult. We had also tried to achieve monocular stimulation using red-and-black and green-and-black checkerboards, and then matching the red and green of these checkerboards to the filters in front of the eyes. However, the check edges were often still visible. We thought mechanical shutters in front of the eyes to cover 1 eye at a time would be too cumbersome and disturbing to the subject. Fiberoptic video projection goggles are commercially available, but the most popular ones (Resonance Technologies, for instance) are designed for binocular viewing and do not allow the flexibility of spherical and cylindrical refractive correction. The order of the stimulus conditions was randomized in a balanced fashion to reduce any theoretical intercondition dependence and a subject’s possible ability to anticipate the stimulus. We also used a pushbutton response pad to monitor the subjects’ wakefulness and attention to the

Liu et al

47

stimulus. It has been well established that attention to visual stimuli enhances activation in the visual cortex, in part because of top-down influences.16,17 Motion correction was performed on all data sets. A common ROI for the primary visual cortex for all subjects was drawn on a standard template in Talairach space, and each subject’s study was transformed to the same space. The drawn ROI was most likely liberal and probably included some of area V2 as well as V1. We recognize that retinotopic mapping of the primary visual cortex has become standard in many visual fMRI paradigms.18-22 In addition, unlike in extrastriate areas where the responses saturate at 6%, responses in the primary visual cortex remain relatively low at low contrasts but increase gradually at contrasts greater than 6%.23 Therefore, the primary visual cortex can also be delineated by alternating stimuli with 6% and 95% contrast and identifying cortical areas that exhibit contrast modulation.23,24 However, combining retinotopic mapping or contrast modulation techniques with our technique would require either longer or multiple studies. This would be problematic in many respects. First, in long study sessions, cortical activation varies and sometimes diminishes,25 and subjects fatigue and lose attention. Combining the results from serial studies, with rest periods in between, is confounded by errors in head realignment and intertest variability,26 as suggested above. In future studies, an alternative ROI can be created by retinotopic or contrast modulation mapping of V1 in a separate cohort of normal subjects and then spatially transforming their studies to Talairach space. The results can be averaged to find a common V1. However, with the number of computational manipulations required to do this, these types of ROIs may be inferior to a drawn one. Our fMRI method for studying eye dominance is modelled after single-cell techniques, but there are clearly limitations to comparing the 2, especially with our relatively large voxel size. Firstly, fMRI cannot achieve the spatial or temporal resolution of single-cell recordings, even at higher resolution. Second, our definition of the eye dominance of a voxel is not equivalent to the eye dominance of a neuron. The visual cortex consists of 6 layers, and only layer IVc contains clusters of monocular neurons in OCDs. Cells in other layers have varying degrees of binocularity. Since there are approximately 150,000 neurons per cubic millimeter of visual cortex, the 3.75 × 3.75 × 5 mm3 voxel in our study would contain about 10.5 million neurons. The eye dominance of a voxel presumably would reflect an aggregate of the neuronal population’s eye dominance. It is possible that if half of the cells in a voxel are left-eye dominated and the other half right-eye dominated, then the activation would be categorized as 4, even if there are no binocular neurons. In other words, our method does not assess the monocularity or binocularity of single neurons but evaluates the eye dominance of a population in a voxel. Finally, small changes in the widths of the OCDs (normally 0.5 × 0.5 × 1.0 mm3) seen in animals with amblyopia, for

000-000/Liu120170

48

3/7/02

11:50 AM

Page 48

Journal of AAPOS Volume 6 Number 1 February 2002

Liu et al

instance,1,2 most likely are not detectable with our, or any other, fMRI method.14 Despite these limitations, our method may prove useful for studying the eye dominance of voxels in the human visual cortex. The addition of an infrared eye-movement tracker to monitor eye fixation to the center of the checkerboard and a surface coil to improve signal to noise may enhance our methodology. We hope that patients with abnormal sensory visual processing, such as strabismus with alternate fixation, suppression, and different forms of amblyopia, can be studied in the future with our fMRI methodology. For instance, by varying the psychophysical attributes of the visual stimulus, such as check size and contrast, one can study whether anisometropic, deprivational, and strabismic amblyopia affect eye dominance in the visual cortex in similar or different ways. Our group has already begun investigating the relative contribution of magnocellular or parvocellular defects in anisometropic amblyopia.27 The authors would like to thank Drs David Darby for all his help and advice regarding the Macstim software, Maureen Maguire for her advice and encouragement, John Detre for his helpful suggestions, Jill Hunter for reviewing the anatomical MRI studies, and Jonathan Raz for his tremendous help with the statistical analysis.

References 1. Hubel DH, Wiesel TN. The period of susceptibility to the physiological effects of unilateral eye closure in kittens. J Physiol 1970;206:419-36. 2. Wiesel TN. Postnatal development of the visual cortex and the influence of environment. Nature 1982;299:583-91. 3. Baker FH, Grigg P, von Noorden GK. Effect of visual deprivation and strabismus on the response of neurons in the visual cortex of the monkey, including studies on the striate and prestriate cortex in the normal animal. Brain Res 1974;66:185-208. 4. Crawford MLJ, Blake R, Cool SJ, von Noorden GK. Physiological consequences of unilateral and bilateral eye closure in macaque monkeys: some further observations. Brain Res 1975;84:150-4. 5. von Noorden GK, Crawford MLJ. Form vision deprivation without light deprivation produces the visual deprivation syndrome in Macaca mulatta. Brain Res 1977;129:37-44. 6. Smith EL, Chino YM, Ni J, Cheng J, Crawford MLJ, Harwerth RS. Residual binocular interactions in the striate cortex of monkeys reared with abnormal binocular vision. J Neurophysiol 1997;78:1353-62. 7. Kiorpes L, Kiper DC, O’Keefe LP, Cavanaugh JR, Movshon JA. Neuronal correlates of amblyopia in the visual cortex of macaque monkeys with experimental strabismus and anisometropia. J Neurosci 1998;18:6411-24. 8. Menon RS, Ogawa S, Strupp JP, Ugurbil K. Ocular dominance in human V1 demonstrated by functional magnetic resonance imaging. J Neurophysiol 1997;77:2780-7.

9. Noll DC, Genovese CR, Nystrom LE, Vasquez AL, Forman DS, Eddy WF, et al. Estimating test-retest reliability in functional MR imaging 2: application to motor and cognitive activation studies. Magn Reson Med 1997;38:508-17. 10. Fox PT, Raichle ME. Stimulus rate determines regional brain blood flow in striate cortex. Ann Neurol 1985;17:303-5. 11. Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain. Stuttgart: Thieme; 1988. 12. Liu GT, Hunter J, Miki A, Fletcher D. Brown L, Haselgrove JC. Functional MRI in children with congenital structural abnormalities of occipital cortex. Neuropediatrics 2000;31:13-5. 13. Stensaas SS, Eddington DK, Dobelle WH. The topography and variability of the primary visual cortex in man. J Neurosurg 1974;40:747-55. 14. Menon RS, Goodyear BG. Submillimeter functional localization in human striate cortex using BOLD contrast at 4 Tesla: implications for the vascular point-spread function. Magn Reson Med 1999;41:230-5. 15. Katsumi O, Hirose T, Sakaue H, Mehta M, Rosenstein RB. Effect of optical defocus on the steady state pattern reversal visual-evoked response. Ophthalmic Res 1990;22:383-90. 16. Watanabe T, Sasaki Y, Miyauchi S, Putz B, Fukimaki N, Nielsen M, et al. Attention-regulated activity in human primary visual cortex. J Neurophysiol 1998;79:2218-21. 17. Somers DC, Dale AM, Seiffert AE, Tootell RB. Functional MRI reveals spatially specific attentional modulation in human primary visual cortex. Proc Natl Acad Sci USA 1999;96:1663-8. 18. Sereno MI, Dale AM, Reppas JB, Kwong KK, Belliveau JW, Brady TJ, et al. Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging. Science 1995;268:889-93. 19. DeYoe EA, Carman GJ, Bandettini P, Glickman S, Wieser J, Cox R, et al. Mapping striate and extrastriate visual areas in human cerebral cortex. Proc Natl Acad Sci USA 1996;93:2382-6. 20. Engel SA, Glover GH, Wandell BA. Retinotopic organization in human visual cortex and the spatial precision of functional MRI. Cereb Cortex 1997;7:181-92. 21. Tootell RBH, Hadjikhani NK, Mendola JD, Marrett S, Dale AM. From retinotopy to recognition: fMRI in human visual cortex. Trends Cog Sci 1998;2:174-83. 22. Wandell BA. Computational neuroimaging of human visual cortex. Annu Rev Neurosci 1999;22:145-71. 23. Tootell RBH, Hadjikhani NK, Vanduffel W, Liu AK, Mendola JD, Serono MI, et al. Functional analysis of primary visual cortex (V1) in humans. Proc Natl Acad Sci USA 1998;95:811-7. 24. Goodyear BG, Menon RS. Effect of luminance contrast on BOLD fMRI response in human primary visual areas. J Neurophysiol 1998;79:2204-7. 25. Liu GT, Fletcher DW, Bishop RJ, Maguire MG, Quinn GE, Hendy P, et al. Variability in visual cortex activation during prolonged functional magnetic resonance imaging (fMRI). J Neuroophthalmol 1998;18:258-62. 26. Miki A, Raz J, van Erp TGM, Liu CSJ, Haselgrove JC, Liu GT. Reproducibility of visual activation in functional magnetic resonance imaging and effects of post-processing. Am J Neuroradiol 2000;21:910-5. 27. Miki A, Siegfried JB, Liu GT, Liu C-SJ, Chen HK, Modestino EJ, et al. Magno- and parvocellular visual cortex activation in anisometropic amblyopia studied with functional MRI. Invest Ophthalmol Vis Sci 2001;42:S52.