Abstract
We employed high-resolution fMRI to distinguish the impacts of anisometropia and strabismus amblyopia on the evoked ocular dominance (OD) response. Sixteen amblyopic participants (eight females) plus eight individuals with normal vision (one female), participated in this study for whom we measured the difference between the response to stimulation of the two eyes, across areas V1–V4. In controls, the evoked OD response formed the expected striped pattern within V1. Compared to controls, the OD response in amblyopic participants formed larger fused patches that extended into downstream visual areas. Moreover, both anisometropic and strabismic participants showed elevated OD responses across V1–V4. Beyond these common effects, and despite similar densities of amblyopia between the two groups, strabismus, and anisometropia had differential impacts on the OD bias, binocular response, and correlation between V1 depth levels. Specifically, we found a greater increase in the size of the V1 portion that responded preferentially to fellow eye stimulation in anisometropic compared with strabismic individuals. We also found a greater difference between the amplitudes of the response to binocular stimulation, in those regions that responded preferentially to the fellow versus amblyopic eye, in anisometropic compared with strabismic participants. In contrast, strabismic participants demonstrated an increased correlation between the OD responses evoked within V1 superficial and deep depths, whereas anisometropic individuals did not. These results provide the primary direct functional evidence for distinct impacts of strabismus and anisometropia on the mesoscale functional organization of the human visual system, thus extending what was inferred previously about amblyopia from animal models.
- amblyopia
- columnar organization
- high-resolution fMRI
- interocular visual acuity difference
- ocular dominance response
Significance Statement
Amblyopia is a developmental disorder caused by perturbations to normal binocular visual experience during the critical period impact. Despite its high prevalence, our current understanding of amblyopia impacts on the mesoscale functional organization of the primary visual area (V1) is mostly based on invasive techniques in animals. In this study, we showed the primary direct evidence for the distinct impacts of anisometropia versus strabismus (two major causes of amblyopia) on the fMRI activity evoked within the human visual cortex. Our findings also confirmed the hypothesized link between the evoked OD response and the interocular visual acuity difference in amblyopia.
Introduction
Ocular dominance (OD), the preference for responding to stimulation of one eye over the other, is a prominent characteristic of most neurons in the primary visual cortex (V1; Hubel and Wiesel, 1962). In humans and many nonhuman mammals, neurons with similar OD preferences are grouped together in ocular dominance columns (ODCs; LeVay et al., 1975; Tootell et al., 1988; Sincich et al., 2003; Adams et al., 2007). The development of ODCs depends on balanced binocular visual input at early life stages (Hubel et al., 1977; LeVay et al., 1980; Horton and Hocking, 1997). Perturbations to normal binocular visual experience during this period impact the selectivity and distribution of ODCs and are associated with amblyopia, a prevalent neurodevelopmental disorder affecting a range of visual functions in one or both eyes (McKee et al., 2003; Maurer and McKee, 2018).
Our understanding of amblyopia impacts on ODCs is mainly based on electrophysiological and anatomical studies in animals (Fig. 1). According to these studies, asymmetric binocular vision in early life stages, caused either by misalignment of the eyes (strabismus), differential optics of the eyes (anisometropia), or monocular deprivation, leads to a reduction in the number of V1 neurons that respond binocularly (Crawford and Von Noorden, 1979; Movshon et al., 1987; Crawford et al., 1996; Smith et al., 1997a; Kiorpes et al., 1998; Bi et al., 2011). Beyond this common effect, anisometropia, even in milder forms, is associated with a decrease in the number of neurons that respond preferentially to the amblyopic eye. Whereas such a bias is only detectable in strabismic participants with severe amblyopia (Crawford et al., 1996; Kiorpes et al., 1998; Bi et al., 2011). Moreover, strabismus (but not anisometropia) increases the segregation between ODCs with opposing ocular preference (Lowel, 1994; Tychsen et al., 2004).
Schematic representation of the relative impact of anisometropic and strabismic amblyopia on the ocular preference of V1 neurons in nonhuman primates. Individuals with normal binocular vision and no amblyopia (left) have a uniform preference for either eye, with some neurons favoring the dominant or nondominant eye and others showing varying degrees of binocular preference. Amblyopic individuals (regardless of cause) show a decrease in the total number of binocular neurons in V1 (Crawford and Von Noorden, 1979; Crawford et al., 1996; Smith et al., 1997a; Kiorpes et al., 1998; Bi et al., 2011), while the distribution varies with type: In anisometropic amblyopia (middle), this effect is accompanied with a decrease in the number of V1 neurons that respond preferentially to the amblyopic eye, even in those with milder forms of amblyopia. In strabismic individuals with milder forms of amblyopia (right, dashed line), amblyopic eye-preferring neurons remain frequently detectable across V1, whereas in more severe forms (right, solid line), these neurons are less frequently observed.
In humans, fMRI has been used to localize V1 OD bands noninvasively (Menon et al., 1997; Cheng et al., 2001; Yacoub et al., 2007; Nasr et al., 2016). Using this technique, further studies suggest amblyopia is associated with a greater number of voxels responding preferentially to the fellow eye compared with the amblyopic eye (Algaze et al., 2002; Goodyear et al., 2002; Liu et al., 2004) and that the OD activity was stronger in amblyopic participants compared with controls (Conner et al., 2007). It was also suggested that amblyopia changes the mechanism of binocular interaction from excitation to suppression (Farivar et al., 2011; Thompson et al., 2019) and affects the receptive field size of V1 neurons, when measured based on stimuli presented to the amblyopic eye (Clavagnier et al., 2015). However, these studies did not distinguish the impacts of anisometropia versus strabismus on the evoked OD response and/or the mesoscale functional organization of V1, presumably due to the limited spatial resolution and contrast-to-noise ratio of the neuroimaging techniques available at the time.
To address these knowledge gaps, this study used higher spatial resolution fMRI, conducted in a 7 T MR scanner. The contrast-to-noise ratio was improved by minimizing the level of unwanted signal blurring without applying any spatial smoothing within cortical layers (Blazejewska et al., 2019; Wang et al., 2022). Using these methods, we compared the impact(s) of strabismus and anisometropia on the spatial distribution and columnar organization of the evoked OD response in human V1. We also measured the impact of amblyopia on the amplitude of OD responses in V1–V4 and compared the correlation between the evoked OD response and the interocular visual acuity difference as a measure of amblyopia severity across these areas. Lastly, we aimed to compare the evoked activity across V1 regions to binocular stimulation to test whether the binocular response varies between regions that respond preferentially to the fellow versus amblyopic eye.
Materials and Methods
Participants
Twenty-five human participants (10 females), aged 19–56 years old, participated in this study (Table 1). This included seven anisometropic, one deprivational, and eight strabismic participants with amblyopia. We also included eight individuals with normal (n = 6) or correct-to-normal (n = 2) visual acuity, as controls. One extra participant with mild strabismus (but no amblyopia) also participated in our study. The data from this individual are demonstrated separately. All participants had radiologically intact brains and no history of neuropsychological disorders.
Demography and ophthalmologic assessment of the participants
During the main experiments, three amblyopic individuals could not wear their prescribed eyeglasses due to safety concerns with MRI compatibility (Table 1). To test the impact of this, one control participant underwent an additional control experiment during which the participant was tested without their contact lenses.
All experimental procedures conformed to NIH guidelines and were approved by Massachusetts General Hospital protocols. Written informed consent was obtained from all participants prior to all experiments.
Ophthalmological assessment
Outside the scanner, participants were tested by an optometrist (J.S.) with extensive experience with amblyopic individuals. During these tests, participants’ visual acuity [ETDRS retro luminant chart (Precision Vision)] was measured with pinhole (i.e., best corrected) and without pinhole (as in fMRI scans). The stereoacuity was measured using the Randot stereo test (Stereo Optical). We identified the participant's dominant eye (Miles test) and tested for suppression or diplopia (Worth 4 Dot).
MRI experiments
Participants were scanned in an ultrahigh field 7 T scanner (whole-body system, Siemens Healthcare) for the functional experiments. All participants were also scanned in a 3 T scanner (Tim Trio, Siemens Healthcare) for structural imaging.
During the fMRI experiments, stimuli were presented via an LCD projector (1,024 × 768 pixel resolution, 60 Hz refresh rate) onto a rear-projection screen, viewed through a mirror mounted on the receive coil array. MATLAB 2021a (MathWorks) and the Psychophysics Toolbox (Brainard, 1997; Pelli, 1997) were used to control stimulus presentation. The participants were instructed to look at a centrally presented fixation object (radius, 0.15°) and to do either a shape change for the fixation target (circle-to-square or vice versa) during the OD measurements or a random dot detection during the retinotopic mapping. These tasks were conducted without any significant difference across experimental conditions (p > 0.10).
Notably, due to the small size of the head coil and strong magnetic field, we were not able to conduct eye monitoring during the scans. However, by keeping the fixation task identical across the experimental conditions, we reduced the potential impact of fixation instability caused by amblyopia. This point and the discrepancies between the findings and the expected impacts of fixation instability are further discussed thoroughly in Discussion.
Response to monocular and binocular visual stimulation based on moving random dots
All participants completed two separate scan sessions. In each session, we stimulated the participant's fellow (dominant) and amblyopic (nondominant) eyes in different blocks (i.e., block design; 24 s per block). The stimuli were sparse (5%) moving random red (50% of blocks) and green (the rest of blocks) dots (0.09° × 0.09°; 56 cd/m2), presented against a black background. In separate blocks, we also measured the response to binocular presentation of the simultaneous stimulation of both eyes (with zero disparity) in all participants except for one control.
Participants viewed the stimuli through custom-made anaglyph spectacles with red (cutoff wavelength, >575 nm) and green–blue (cutoff wavelength, <560 nm) filters mounted to the head coil. Considering the wavelength of the red (∼700 nm) and green (495–570 nm) light and the low sparseness of the stimuli, we expected the level of cross talk to be negligible. The absence of cross talk was further confirmed during our pilot studies (Nasr et al., 2016) during which we asked the participants with an intact vision to verbally confirm that they could not see the nonpreferred stimulus through the color filters (e.g., the green random dots through the red filter). The absence of cross talk was also confirmed in amblyopic individuals in the same way.
During the blocks, dots were oscillating horizontally (−0.22° to 0.22°; 0.3 Hz). Stimuli extended 20° × 26° in the visual field. Each experimental run began and ended with 12 s of uniform black. The sequence of blocks was pseudorandomized across runs (14 blocks per run), and each participant participated in 12 runs. Filter laterality (i.e., red-left vs red-right) was counterbalanced between sessions and across participants. The results of this experiment are reported below (see Results, Age and interocular visual acuity difference to The impact of amblyopia on the evoked response to binocular visual stimulation).
Response to monocular visual stimulation based on moving gratings
To test whether the strabismic amblyopia impact on the columnarity of the OD response was detectable based on stimuli other than random dots, in this experiment participants (including strabismic and control but not anisometropic individuals; Table 1) were presented with gratings (2.25 cycle/degree). Red and green gratings were presented in different blocks (24 s per block), and participants viewed the stimuli through custom anaglyph spectacles mounted on the head coil. To avoid adaptation, gratings were oscillated left to right [−0.22° to 0.22° (0.3 Hz)]. Stimuli were presented against a black background, extending 20°×26° in the visual field. The orientation of gratings varied randomly between blocks.
Each experimental run began and ended with 12 s of uniform black. The sequence of blocks was pseudorandomized across runs (seven blocks per run), and each participant participated in two runs. Filter laterality (i.e., red-left vs red-right) was counterbalanced across participants. The results of this experiment are reported below (see Results, The impact of amblyopia on the columnarity of OD response).
Retinotopic mapping
For all participants, the border of retinotopic areas was defined retinotopically (Sereno et al., 1995). Stimuli were based on a flashing radial checkerboard, presented within retinotopically limited apertures, against a gray background. These retinotopic apertures included wedge-shaped apertures radially centered along the horizontal and vertical meridians (polar angle, 30°). These stimuli were presented to participants in different blocks (24 s per block). The sequence of blocks was pseudorandomized across runs (eight blocks per run), and each participant participated in at least four runs.
Imaging
Functional experiments (see above) were conducted in a 7T Siemens whole-body scanner (Siemens Healthcare) equipped with SC72 body gradients (70 mT/m maximum gradient strength and 200 T/m/s maximum slew rate) using a custom-built 32-channel helmet receive coil array and a birdcage volume transmit coil. Voxel dimensions were nominally 1.0 mm. We used single-shot gradient-echo EPI to acquire functional images with the following protocol parameter values: TR, 3,000 ms; TE, 28 ms; flip angle, 78°; matrix, 192 × 192; BW, 1,184 Hz/pix; echo-spacing, 1 ms; 7/8 phase partial Fourier; FOV, 192 × 192 mm; 44 oblique-coronal slices; and acceleration factor R = 4 with GRAPPA reconstruction and FLEET-ACS data (Polimeni et al., 2015) with 10° flip angle. The field of view included occipital cortical areas V1–V4.
Structural (anatomical) data were acquired in a 3 T Siemens Tim Trio whole-body scanner, with the standard vendor-supplied 32-channel head coil array, using a 3D T1-weighted MPRAGE sequence with protocol parameter values: TR, 2,530 ms; TE, 3.39 ms; TI, 1,100 ms; flip angle, 7°; BW, 200 Hz/pix; echo-spacing, 8.2 ms; voxel size, 1.0 × 1.0 × 1.33 mm3; FOV, 256 × 256 × 170 mm3.
General data analysis
Functional and anatomical MRI data were preprocessed and analyzed using FreeSurfer and FsFast (version 7.11; http://surfer.nmr.mgh.harvard.edu/; Fischl, 2012).
Structural analysis
For each participant, inflated and flattened cortical surfaces were reconstructed based on the high-resolution anatomical data (Dale et al., 1999; Fischl et al., 1999, 2002). Then, during this reconstruction process, the standard pial surface was generated as the gray matter border with the surrounding cerebrospinal fluid or CSF (i.e., GM-CSF interface). The white matter surface was also generated as the interface between white and gray matter (i.e., WM-GM interface).
Definition of superficial, middle, and deep layers
To enable intracortical smoothing (see below), for each subject, we also generated a family of nine intermediated equidistant surfaces, spaced at intervals of 10% of the cortical thickness, between WM-GM and the GM-CSF interface surfaces (Polimeni et al., 2010; Blazejewska et al., 2019). To improve the coregistration of functional and structural scans, all surfaces were unsampled to 0.5 mm resolution (Wang et al., 2022).
Using these surfaces, GM was divided into three depth levels: (1) “deep cortical depth” included the bottom 30% of the GM thickness, starting at the WM-GM interface, (2) “superficial cortical depth” included GM-CSF interface and the adjacent two surfaces below it, and (3) “middle cortical depth” was limited to the middle three reconstructed cortical surfaces. Notably, although it is correct to assume that superficial and deep depths coincide with laminar layers 1–2 and 6, respectively, the relationship between the laminar layers 3–5 and these three depth levels is not clear. This is mainly because the width of laminar layers is not constant and may vary based on eccentricity and cortical folding.
Functional analysis
The collected functional data were first unsampled (to 0.5 mm isotropic) and then corrected for motion artifacts. For each participant, functional data from each run were rigidly aligned (six DOF) relative to their own structural scan using rigid boundary-based registration (Greve and Fischl, 2009). This procedure enabled us to compare data collected for each participant across multiple scan sessions.
To retain the spatial resolution, no tangential spatial smoothing was applied to the imaging data acquired at 7 T (i.e., 0 mm FWHM). Rather we used the more advanced method of radial (intracortical) smoothing (Blazejewska et al., 2019)—i.e., perpendicular to the cortex and within the cortical columns. For deep cortical depths, the extent of this radial smoothing was limited to the WM-GM interface and the adjacent two surfaces right above it (see above)—i.e., the bottom 30% of the gray matter thickness starting from the WM-GM interface. For the superficial cortical depths, the extent of this procedure was limited to the GM-CSF interface and the adjacent two surfaces right below it. For the middle cortical layers, used only for presentation (Fig. 2), the extent of this procedure was limited to the three middle reconstructed cortical surfaces.
The OD response evoked by contrasting the response, evoked within the left hemisphere (LH), to stimulation of dominant/fellow (red to yellow) versus nondominant/amblyopic (blue to cyan) eye across cortical depth levels (see Materials and Methods). Panels A–C show the unthresholded activity maps detected within deep (top) and middle and superficial (bottom) cortical depths, in the left hemisphere of a control (Participant #1; Table 1), a strabismic (Participant #13), and an anisometropic (Participant #17) individual, respectively. In the control participants, the OD activity formed mostly parallel stripes that were mostly confined to V1 borders. In the amblyopic participants, especially the anisometropic individual, OD stripes were less pronounced, and the evoked activity extended well beyond the V1 border. This phenomenon was comparably detectable across cortical depths. In all panels, activity maps are overlaid on the person's own reconstructed cortical surface. The V1–V2 border (black dashed line) is also defined for each individual based on their own retinotopic mapping. The foveal direction is shown with the letter F in the top-left panel.
A standard hemodynamic model based on a gamma function was fitted to the fMRI signal to estimate the amplitude of the BOLD response. For each individual participant, the average BOLD response maps were calculated for each condition (Friston et al., 1999). Finally, voxelwise statistical tests were conducted by computing contrasts based on a univariate general linear model, and the resultant significance maps were projected onto the participant's anatomical volumes and reconstructed cortical surfaces. Notably, unless otherwise mentioned, activity maps were only shown at deeper cortical depth levels in which the blurring impact of pial veins is weaker (Koopmans et al., 2010; Polimeni et al., 2010; De Martino et al., 2013; Nasr et al., 2016).
Region of interest (ROI) analysis
To test the impacts of amblyopia on the OD response, ROIs including deep and superficial depths of areas V1, V2, V3, V3A, and V4 were defined for each participant based on their own structural and retinotopic mapping (see above).
To test the impact of amblyopia on the evoked response to binocular stimulation, the V1 surface was divided into two ROIs based on the ocular preference of the vertices, defined during the monocular tests. These ROIs were defined independently for deep and superficial cortical depths.
Notably, no hemisphere was excluded from any ROI analyses, and all vertices within each ROI were used in the analyses.
Statistical data analysis
Three independent parameters included group (anisometropic vs strabismic vs control participants), hemisphere (ipsilateral vs contralateral relative to the dominant/fellow eye), and cortical depth level (deep vs superficial). To test the impact of these parameters, we used either one-way or two-way repeated-measures ANOVA with a group factor. Since this analysis is particularly susceptible to the violation of the sphericity assumption, caused by the correlation between measured values, when necessary (determined using a Mauchly test), results were corrected for violation of the sphericity assumption, using the Greenhouse–Geisser method. All post hoc analyses were conducted after Bonferroni’s correction for multiple comparisons.
Data availability statement
Data and codes will be shared upon request.
Results
The OD response was measured in 24 human participants, 16 with amblyopia caused either by strabismus (n = 8), anisometropia (n = 7), or deprivational amblyopia (n = 1) and 8 control participants with normal or corrected-to-normal vision. In addition to data from these individuals, we also measured the OD response in one strabismic (but nonamblyopic) participant whose data are presented separately. Each participant was scanned twice on different days. On each day, both eyes were stimulated using moving random dots that were presented either monocularly or binocularly (using anaglyphic goggles) in a blocked-design paradigm (see Materials and Methods). The OD response was measured for each participant by averaging the activity evoked across these two sessions and calculating the (absolute) difference between the response to stimulation of dominant/fellow versus nondominant/amblyopic eye. A subset of participants (Table 1) also participated in a control test to measure responses to dichoptically presented grating stimuli. Outside the scanner, all participants were tested to measure their visual acuity and stereoacuity, to identify their dominant eye, and to test for suppression and/or diplopia (see Materials and Methods).
Age and interocular visual acuity difference
Table 1 shows the participant's demographics and visual testing results. One-way ANOVA (anisometropic vs strabismic vs control) did not yield any significant age differences across the three groups (F(2,23) = 1.11, p = 0.35). As expected, a similar analysis applied to the interocular visual acuity difference showed a significant effect of group (F(2,23) = 8.08, p < 0.01) driven by the increased interocular visual acuity difference in both anisometropic and strabismic individuals relative to controls (p < 0.01; Bonferroni corrected for multiple comparisons). The interocular visual acuity difference was similar between the anisometropic and strabismic individuals in our participants (p = 0.89). Visual acuities of amblyopic (p = 0.29) and fellow (p = 0.83) eyes were not different between anisometropic and strabismic participants. Thus, age, interocular visual acuity difference, and visual acuity in the amblyopic and fellow eyes were comparable between anisometropic and strabismic individuals.
Monocular suppression and diplopia were more common in strabismic compared with anisometropic participants (Table 1). Also, as expected based on previous studies (Levi et al., 2015), more strabismic individuals demonstrated severely impaired nonmeasurable stereoacuity (>500 arc seconds) than anisometropic individuals. All amblyopic individuals had a history of either patching or atropine therapy in childhood.
Head position stability during the fMRI tests
Head motion has a strong impact on the fMRI signal and may influence the level and pattern of evoked fMRI responses which might in turn confound between-group comparisons. Thus, as the first step, we compared the level of head motion between control, strabismic, and anisometropic participants. Since all individuals were scanned at least two times on different days, we also tested the consistency of head motion between sessions. One-way repeated-measures ANOVA [session (first vs second)], with a group factor (control vs strabismic vs anisometropic individuals), to the measured level of head motion (see Materials and Methods) did not yield a significant effect of group (F(2,21) = 0.08, p = 0.92) or group × session interaction (F(2,21) = 2.57, p = 0.10) on the degree of head motion. Thus, across the two scan sessions, head motion appears to be comparable across the groups. Head motion was nevertheless included as a nuisance covariate in all analyses to reduce any residual impact of head motion on our findings.
OD activity mapping
We measured the evoked OD activity for all participants in both deep, middle, and superficial cortical depth levels across visual areas V1–V4 by subtracting the response of the nondominant eye from the response of the dominant eye.
OD activity mapping in controls
Figure 2A shows the evoked OD activity in a control participant (Participant #1) across deep, middle, and superficial layers. Consistent with postmortem anatomical studies in humans (Adams et al., 2007) and nonhuman primates (Hubel et al., 1976; Tootell et al., 1988; Sincich et al., 2003) with normal vision, the cortical topography of the evoked OD response was organized into mostly parallel stripes. These striped patterns were similarly detected across cortical depths, reflecting the columnar organization of V1 ODCs (Tootell et al., 1988). In both hemispheres, these stripes were predominantly limited to the regions of V1 (r < 10°), representing the central retinotopic visual field that was stimulated during the scans. This pattern was consistently observed in all control participants in each hemisphere (Fig. 3).
The OD activity mapping in eight control participants, collected from deep cortical depths. In all participants, the striped pattern was apparent within V1. The response amplitude decreased sharply outside the V1 border. For each participant, the white and green box indicates the hemisphere ipsilateral and contralateral relative to the dominant eye, respectively. The white arrowheads show the large activity patch along the dorsal portion of V1–V2 that responded preferentially to the contralateral eye. This patch was detectable in almost all participants except for Participant #6. Other details are the same as in Figure 2.
In all controls, we detected a fused activity patch close to the dorsal portion of the V1–V2 border that responded preferentially to the contralateral eye (Fig. 3). Notably, this cortical region represents the inferonasal visual field occluded by the head coil resulting in monocular representation by the other eye. Also as expected, we did not detect representation of the blind spot and/or temporal monocular crescent, because these regions are represented more peripherally (r > 15°) outside the stimulus borders and scan coverage (Tootell et al., 1998; Awater et al., 2005; Adams et al., 2007; Nasr et al., 2020).
OD activity mapping in strabismic participants
Figure 2B illustrates the evoked OD activity in a strabismic (Participant #13; interocular visual acuity differences, 0.50 logMAR). Compared to controls, OD activity was stronger and formed larger, fused patches at all three cortical depth levels. This increase in the level of OD activity extended to area V2 (Figs. 4, 5) and the downstream visual areas including V3, V3A, and V4 (Fig. 7).
The OD activity mapping in eight strabismic participants, collected from deep cortical depths. Compared to controls (Fig. 3), the amplitude of OD response was larger. Moreover, the OD response extended beyond the V1–V2 borders into downstream visual areas. This effect was accompanied by an extension of those regions that responded preferentially to the fellow eye. This overrepresentation was more pronounced in the hemisphere contralateral relative to the fellow eye. Other details are the same as in Figures 2 and 3.
The OD activity mapping in eight anisometropic participants, collected from the deep cortical depths. As in the strabismic participants (Fig. 4), the amplitude of the OD response is larger relative to controls, and the OD response extended beyond the V1–V2 borders. There was an overrepresentation of the fellow eye, as seen in strabismic participants. However, in contrast to the strabismic participants, this phenomenon was detected bilaterally without any apparent difference between the two hemispheres. Other details are the same as in Figures 2–4.
As demonstrated in Figure 4, in most strabismic individuals, we found larger regions that responded preferentially to the fellow eye within the hemisphere contralateral relative to the fellow eye. This hemispheric laterality was also detectable in our single exotropic participant (Participant #9) suggesting that esotropia could not be the sole reason for this phenomenon. However, Participant #12 was the exception to this trend. In this participant, the representation of the two eyes appeared to be more balanced in the hemisphere ipsilateral relative to the fellow eye (see below, Overrepresentation of the fellow eye in amblyopic participants, for further analysis). The fused activity patch close to the dorsal portion of the V1–V2 border was readily apparent in four strabismic individuals (Participants #10, #12, #13, and #16), as in controls.
Among the participants categorized as having strabismic amblyopia, Participants #9 and #14 had only a small difference in interocular visual acuity at the time of testing (≤0.12 logMAR; Table 1). Both individuals had a history of strabismus surgery and patching in childhood. Despite the small interocular visual acuity difference, both individuals showed signs of diplopia on Worth 4 Dot testing, with reduced stereoacuity in Participant #9 but not in Participant #14. In both cases, we found an elevated OD response in V1, especially in the hemisphere contralateral to the fellow eye, similar to the other strabismic individuals. This result suggests that imbalanced ocular dominance may persist despite recovery of monocular visual acuity in the amblyopic eye, consistent with behavioral evidence for impaired “dichoptic” amblyopic eye visual acuity despite resolved interocular visual acuity differences after patching treatment (Birch, 2013; Birch et al., 2022).
OD activity mapping in anisometropic participants
Figure 2C illustrates the evoked OD activity in an anisometropic participant (Participant #17; interocular visual acuity differences, 0.42 logMR). As in strabismic individuals, OD activity was stronger and formed larger, fused patches at all three cortical depth levels compared with controls. In anisometropic individuals (Fig. 5), OD activity bias in favor of the fellow eye was detectable bilaterally in almost all participants. Among the four anisometropic individuals who did not show monocular suppression (Table 1), Participants #17 and #22 showed a strong bias in favor of the fellow eye, but in Participants #18 and #20, this bias was comparatively weaker. In contrast to strabismic individuals and controls, the activity patch along the V1–V2 border was less apparent in anisometropic individuals, likely due to a strong bias in favor of the fellow eye.
Notably, the individual with deprivational amblyopia (Participant #21) showed strong OD activity bias in favor of the fellow eye in both hemispheres, as in anisometropic individuals, even though the (unilateral; left eye) cataract was removed when the participant was a child, and the stimuli were perceived with best correction. Here again, this activity bias propagated into downstream visual areas. Considering the similarity between this individual's OD pattern and those of anisometropic participants, we included this participant in the anisometropic group in the following analyses. Importantly, the exclusion of this individual from the anisometropic group did not change the overall pattern of findings, nor did it change the results of between-group analyses.
Reproducibility of the OD response maps across sessions
To compare the reproducibility of these maps across the three groups, we measured the correlation between OD activity maps evoked during the first and second scan sessions. This measurement was conducted separately for the activity evoked within the deep and superficial cortical layers and for the ipsilateral and contralateral hemispheres relative to the fellow eye. As we have shown previously (Nasr et al., 2016), OD activity maps remained highly correlated across sessions (Fig. 6). Two-way repeated-measures ANOVA (hemisphere and cortical depth, with a group factor) did not yield an effect of group (F(2,21) = 0.42, p = 0.66) or an interaction between the group and other independent variables (p > 0.14). The same result was found in areas V2–V4, suggesting that activity maps were reproducible to the same extent for the three groups across visual areas.
Reproducibility of the OD maps across scan sessions. The activity maps show the OD response evoked within the left hemisphere of the same control (top), strabismic (middle), and anisometropic (bottom) participants, as in Figure 2, across two separate sessions (see Materials and Methods). The scatter plots highlight the correlation (p < 10−3) between the OD response evoked with V1 across the two sessions. Each data point represents activity in one vertex from the reconstructed cortical surface mesh.
Overrepresentation of the fellow eye in amblyopic participants
Previous studies in humans (Goodyear et al., 2002; Liu et al., 2004) and nonhuman primates (Movshon et al., 1987; Smith et al., 1997a; Kiorpes et al., 1998) have suggested an increased representation of the fellow eye within V1 in amblyopic compared with control participants. This increased representation is expected to be larger within the hemisphere contralateral relative to the dominant eye (Movshon et al., 1987) likely due to the asymmetric distribution of ganglion cells between the nasal versus the temporal portion of the retina (Perry et al., 1984; Perry and Cowey, 1985).
Consistent with these reports, we found an increase in the size of the V1 region that responded preferentially to the fellow eye in amblyopic participants compared with controls across deep and superficial cortical depth levels (Table 2). This effect also tended to be larger in anisometropic compared with strabismic individuals. Two-way repeated-measures ANOVA (hemisphere and cortical depth, with a group factor) yielded a significant effect of group (F(2,21) = 5.74, p = 0.01), and a significant group × hemisphere interaction (F(2,21) = 3.86; p = 0.04), but no group × cortical depth interaction (F(2,21) = 0.64; p = 0.53) on the size of the V1 portion that responded preferentially to the fellow eye. Post hoc analyses showed that in strabismic participants, this effect was significantly larger in the hemisphere contralateral compared with ipsilateral relative to the fellow eye (p = 0.03). We did not find such a difference in either anisometropic (p = 0.35) or control (p = 0.56) participants. Notably, all measurements were normalized relative to the size of the V1 area that was stimulated. Moreover, the exclusion of the individual with deprivational amblyopia (Participant #21) did not change the overall pattern (or the significance) of the findings.
The size of the V1 portion that responded preferentially to the fellow/dominant eye (mean ± SD)
The impact of amblyopia on the amplitude of the OD response
In addition to the change in the size of the V1 portion that responded preferentially to the fellow eye, there was an increase in the amplitude of the evoked OD response in amblyopic compared with control participants (Fig. 7). So, we tested whether there is a difference between the level of evoked OD response across the participants and if the evoked response is correlated with the level of interocular visual acuity difference. Accordingly, a two-way repeated-measures ANOVA (as above) revealed a significant effect of group (F(2,21) = 11.91, p < 10−3). Post hoc analysis further showed that the evoked OD response in V1 was significantly larger in strabismic (p < 10−3) and anisometropic (p < 10−5) participants compared with controls without a significant difference between strabismic and anisometropic participants (p = 0.22). Here again, the exclusion of the individual with deprivational amblyopia (Participant #21) did not change the significance of the findings. Thus, in line with previous studies in humans (Conner et al., 2007) and nonhuman primates (Crawford and Von Noorden, 1979; Crawford et al., 1996; Smith et al., 1997a; Kiorpes et al., 1998; Bi et al., 2011), amblyopia increased the wclamplitude of the OD response in human V1 in our fMRI data.
The amplitude of the OD response was measured in both deep (A) and superficial (B) cortical depths of V1–V4. Across all areas, the level of OD response was higher in the amblyopic participants compared with the controls, without a significant difference between the anisometropic and the strabismic individuals. To avoid signal cancelation, the ROI analysis was applied to the absolute value of OD response. Panels C and D show that, in both deep and superficial depths, the average OD response decreased in downstream visual areas relative to V1. However, the correlation between OD response and the interocular visual acuity difference increased from V1 to V2 to V3. Each point in these panels represents the average data from both hemispheres. Notably, the correlation values were calculated based on all participants. However, exclusion of controls did not change the overall results.
Importantly, the heightened OD response extended beyond V1 into downstream visual areas V2, V3, V3A, and V4 (Fig. 7A,B). Despite a gradual decrease in the OD response amplitude from V1 through V4, the significantly stronger OD response in amblyopic individuals compared with controls was preserved across all tested areas (p < 0.01). As in V1, the amplitude of the OD response remained comparable between strabismic versus anisometropic participants (p > 0.10). These results suggest that the impact of amblyopia on the OD response amplitude propagated to downstream visual areas, irrespective of the amblyopia subtype.
Moreover, we found a moderate correlation between the interocular visual acuity difference (as in the scans; Table 1) and OD response amplitudes across visual areas V1–V4 (r > 0.43; p < 0.01). This correlation was considerably stronger in areas V2–V4 (r = 0.55–0.70) compared with V1 (r = 0.47), especially in deeper cortical depth levels, despite the decrease in the overall level of evoked OD response (Fig. 7C,D; Table 3). This correlation was similarly detected in contralateral and ipsilateral hemispheres and across superficial and deep cortical depth levels.
The correlation between the interocular visual acuity difference and OD activity evoked across V1–V4 (measured across all participants)
To compare these correlation values more directly, we generated a linear multiple regression model using the interocular visual acuity difference as the dependent parameter and the OD activity evoked within V1–V4 (averaged between the two hemispheres) as the independent parameter. As demonstrated in Table 3, we found a stronger standardized beta value for V4 compared with V1 activity in both superficial and deep cortical depth levels. Thus, while the impact of amblyopia on the amplitude of the OD response was stronger in V1 compared with downstream visual areas, the correlation between OD response and interocular visual acuity difference was stronger in higher-level visual cortical areas such as V3 and V4.
Contributions of residual strabismus
As reported in Table 1, the strabismic participants show some residual misalignments between the two eyes, despite prior surgical correction on all of them except for one (Participant #10). To test whether mild strabismus, in the absence of amblyopia, may lead to the stronger OD responses we observed in individuals with strabismic amblyopia, we scanned a nonamblyopic individual with mild strabismus (separate from the other eight controls (Participant #25; Table 1). This participant showed normal, balanced visual acuities, no evidence of suppression or diplopia, and measurable stereoacuity (70 s of arc).
As demonstrated in Figure 8, in this participant, the overall pattern of the OD response in V1 and downstream visual areas was distinguishable from that in individuals with strabismic amblyopia; instead, it more closely resembled the results in the controls (Figs. 2, 3). Specifically, the size of the region that showed response preference to dominant eye stimulation within the contralateral (46.72%) and ipsilateral (43.03%) hemisphere (relative to the dominant eye) remained small compared with the individuals with strabismic amblyopia (Table 2). Similar results were detected within the superficial cortical levels. Thus, strabismus per se, in the absence of amblyopia, is not the main cause of increased OD response in our participants. Notably, data from this participant were not used in any other analyses.
The OD activity mapping in one nonamblyopic strabismic participant (Participant #25; Table 1), collected from the deep cortical depth levels. The size of the region that showed an OD bias in favor of the dominant eye remained close to what we found in control individuals (Table 2) in the contralateral (46.72%) and ipsilateral (43.03%) hemisphere (relative to the dominant eye). Other details are the same as in Figures 2–5.
Contributions of uncorrected visual acuity
Among the participants, one anisometropic (Participant #22) and two strabismic (Participants #12 and #15) individuals could not be tested with their best optical correction. Even though this deviation had a relatively small impact on the level of interocular visual acuity difference (Table 1; <0.11 logMAR), we tested whether deviation from the best corrected visual acuity per se might increase the level of OD activity. In separate scan sessions, one control individual (Participant #6) was tested again with increased visual acuity difference by instructing the participant not to wear their prescribed contact lenses. Visual acuity worsened without correction by 0.76/1.00 (right/eye visual acuity), and the interocular visual acuity difference increased from 0.06 to 0.24 logMAR, as in the three participants with amblyopia who could not be tested with their best corrected visual acuity.
Figure 9, A and B, shows the evoked OD response in this individual, measured within the deep cortical depth levels with and without corrected visual acuity, respectively. While the level of bias increased in favor of the dominant eye, the level of evoked OD activity only increased by 0.04% (fMRI signal level) and 0.12% in the contralateral and ipsilateral hemispheres, respectively. Moreover, OD activity in this participant was comparably weaker, compared with the OD activity detected on average in amblyopic individuals, in both contralateral (0.99%) and ipsilateral (0.82%) hemispheres. Similar results were detected in the superficial cortical depth levels. Thus, the increased OD activity in the three individuals with amblyopia who were unable to wear their best correction is only marginally attributable to the absence of optical correction.
The OD activity mapping in one nonamblyopic anisometropic participant (Participant #6). Panels A and B, respectively show the OD activity with corrected visual acuity and after increasing the interocular visual acuity difference in favor of their dominant eye (from 0.06 to inducing 0.24 logMAR) by instructing the participant not to wear their contact lenses. Despite the increased level of interocular visual acuity difference, the evoked OD activity remained weaker compared with those detected in the amblyopic anisometropic individuals (Fig. 5). Other details are the same as in Figures 2–5.
The impact of amblyopia on the evoked response to binocular visual stimulation
At behavioral levels, amblyopic subjects receive less benefits from binocular summation (Lema and Blake, 1977; Pardhan and Gilchrist, 1992). The decrease in binocular summation is expected to be stronger in anisometropic compared with strabismic individuals (Baker et al., 2007). But, at the neuronal levels, it remains unclear whether this effect is more selective to those neurons that respond preferentially to the amblyopic eye or if it also affects those neurons that respond preferentially to the fellow eye. To clarify this point, in separate blocks, we measured the evoked response to concurrent stimulation of both eyes. We compared binocular response amplitudes between regions preferring the dominant/fellow eye with those of the nondominant/amblyopic eye for each group. As demonstrated in Figure 10, the results of this test revealed two important phenomena: First, there was no apparent difference between the level of response evoked by the binocularly presented stimuli, within the V1 regions that responded preferentially to the fellow eye in amblyopic individuals compared with the V1 region that responded preferentially to the dominant eye in the controls.
Activity evoked during binocular stimulation in V1 regions that responded preferentially to the dominant/fellow versus nondominant/amblyopic eye. Panels A and B show the activity evoked in deep and superficial cortical depth levels, respectively. In both depth levels and hemispheres, the level of activity evoked in V1 regions that responded preferentially to the dominant eye remained comparable across the three groups. In the V1 region that responded preferentially to nondominant eye, binocular stimulation evoked a weaker response in anisometropic compared with strabismic and controls. This effect was more apparent in more superficial rather than deep cortical depths and in contralateral rather than ipsilateral hemispheres (relative to the dominant eye). In all panels, each dot pair represents one individual participant.
Second, in controls, binocular responses were comparable between V1 regions that responded preferentially to the dominant versus nondominant eye, whereas in amblyopic participants, evoked responses to binocular visual stimulation were stronger in V1 regions that responded preferentially to the fellow eye than those for the amblyopic eye. This effect appeared to be stronger at the superficial cortical depth, in the hemisphere contralateral to the dominant/fellow eye, and in anisometropic compared with strabismic individuals. Three-way repeated-measures ANOVA (hemisphere, preferred eye, and cortical depth, with a group factor) yielded significant group × preferred eye (F(2,20) = 9.99, p < 10−3), group × preferred eye × cortical depth (F(2,20) = 6.17, p < 0.01), and group × preferred eye × cortical depth × hemisphere (F(2,20) = 8.41, p < 0.01) interactions for evoked binocular responses. A post hoc test to compare the response in anisometropic versus strabismic individuals directly also showed a significant group × preferred eye × cortical depth (F(1,14) = 33.47, p < 10−3) interaction whereas the other effects remained nonsignificant after correction for multiple comparisons.
We also tested whether there is a larger binocular summation effect, indexed as the response to binocular stimulation—the response to the preferred eye (Pardhan and Gilchrist, 1992), within different ocular dominance preferring sites in V1. Here again, we witnessed that in anisometropic (but not strabismic and control) individuals, the binocular summation index was stronger in those sites that responded preferentially to the fellow rather than the amblyopic eye (Fig. 11). This phenomenon was stronger in superficial compared with deeper cortical depth levels. Consistently, a three-way ANOVA applied to the measured binocular index showed significant effects of group × preferred eye (F(2,20) = 4.94, p = 0.01) and group × preferred eye × cortical depth (F(2,20) = 4.59, p = 0.02). A post hoc test to compare the response in anisometropic versus strabismic individuals also showed a significant group × preferred eye × cortical depth interaction (F(1,14) = 7.25, p = 0.01). Together, these results suggest that anisometropic, compared with strabismic, amblyopia is associated with a stronger decrease in the level of binocular activity within V1 regions that respond preferentially to the amblyopic eye.
Index of binocular summation in V1 across the three groups. The index was measured as the response to binocular stimulation—the response to the preferred eye (Pardhan and Gilchrist, 1992). Panels A and B show the measured index in deep and superficial cortical depth levels, respectively. The anisometropic (but not strabismic and control) individuals showed a stronger binocular summation index in those sites that responded preferentially to the fellow rather than the amblyopic eye. This effect was stronger in superficial compared with deeper cortical depth levels. In all panels, each dot pair represents one individual participant.
The impact of amblyopia on the columnarity of OD response
While it is known that amblyopia changes the selectivity level of the “horizontal” (i.e., surface parallel) connection between ODCs (Tychsen et al., 2004), the impact of amblyopia on “radial” (i.e., perpendicular to the surface) connections between cortical layers remains mostly unknown even in animal models (Horton and Hocking, 1997). Here, we tested the extent to which amblyopia affects the functional link between deep and superficial cortical layers by comparing the correlation between activity maps evoked within deep and superficial cortical depths across the three groups.
As demonstrated in Figure 12, we found an increased correlation between OD activity maps in deep and superficial cortical depth levels (i.e., interlevel correlation) of V1 in strabismic individuals compared with the two other groups. One-way repeated-measures ANOVA (hemisphere with a group factor) showed a significant effect of group (F(2,21) = 8.32, p < 0.01) without any group × hemisphere interaction (F(2,21) = 0.29, p = 0.75) for interlevel correlation values in V1. A post hoc test showed that the magnitude of interlevel correlation was stronger in strabismic compared with anisometropic (p < 0.01) and control (p < 0.01) participants. Despite the extension of the OD response into the downstream visual areas (see above), application of this analysis to the evoked activity within V2–V4 did not yield a significant effect of group in any of those regions (p > 0.17). This suggests that the impact of amblyopia on the columnarity of OD response is limited to the primary visual cortex using these methods.
The level of correlation between the pattern of OD response evoked within deep and superficial cortical depths, across areas V1–V4. In area V1, but not the other visual areas, strabismic participants show a higher correlation compared with controls and anisometropic individuals. In each graph, each data point shows the data from one individual participant.
To test the reproducibility of this effect, first, we repeated our tests for individual scan sessions, conducted on separate days, rather than the averaged activity maps. Again, two-way repeated-measures ANOVA (hemisphere and session, with a group factor) showed a significant effect of group (F(2,21) = 7.74, p < 0.01) without any significant group × session interaction (F(2,21) = 1.21, p = 0.32) for interlevel correlation values, suggesting that the impact of amblyopia on the columnarity of the OD response was reproducible across scan sessions.
Second, in a subset of our participants, consisting of five control and four strabismic individuals (Table 1) who accepted to participate in an extra scan session, we tested whether this enhanced interlevel correlation is also seen in responses to dichoptically presented drifting gratings (rather than random dots). Briefly, we measured the level of OD activity evoked by gratings presented to fellow/dominant versus amblyopic/nondominant eye and then measured the correlation between the evoked OD activity across V1 at deep and superficial cortical depths. Despite fewer individuals participating in this test, we found a significantly stronger interlevel correlation in strabismic compared with control participants (F(1,7) = 11.09; p = 0.01) and a significant group × hemisphere interaction (F(1,7) = 11.12, p = 0.01). Thus, the enhanced interlevel correlation in the strabismic individuals was reproducible across stimulus types.
For the same subset of subjects, we also checked if the level of between-group difference varies between the two experiments. A two-way repeated-measures ANOVA (hemisphere and experiment, with a group factor) showed a significant effect of group (F(1,7) = 15.21, p < 0.01) without any significant group × experiment (F(1,7) < 0.01, p = 0.97), group × hemisphere (F(1,7) = 1.61, p = 0.24), and group × experiment × hemisphere (F(1,7) = 0.07, p = 0.80) for interlevel correlation values. This result indicated that the impact of strabismic amblyopia on the columnarity of the OD response did not depend on the type of dichoptic stimuli.
Discussion
According to our findings, the expanded representation of the fellow eye is more pronounced in anisometropic compared with strabismic participants, especially in the hemisphere ipsilateral relative to the fellow eye. Moreover, compared with strabismus, anisometropia has a stronger impact on the activity evoked during binocular stimulation within V1 regions that respond preferentially to the amblyopic eye. In comparison, strabismic amblyopia has a stronger impact on the columnarity of OD response evoked within V1.
Consistency with findings based on animal models
Pronounced expansion of fellow eye representation in anisometropic compared with strabismic participants is consistent with single-unit recordings in nonhuman primate V1 (Kiorpes et al., 1998; Bi et al., 2011). According to these studies, the number of neurons that respond preferentially to the fellow and amblyopic eye remains comparable in milder forms of strabismic amblyopia, whereas there is a relative increase of neurons responding preferentially to the fellow eye even in milder forms of anisometropic amblyopia.
The decreased binocular responses in our amblyopic participants are also consistent with previous reports that amblyopia may change the mechanism of interaction between the input from the two eyes (Smith et al., 1997b; Kumagami et al., 2000; Bi et al., 2011; Farivar et al., 2011; Thompson et al., 2019). Here, we showed that this decreased binocular activity is limited to V1 regions that respond preferentially to the amblyopic eye, at least by fMRI, suggesting that binocular integration is differentially impaired in V1 regions according to the ocular preference.
Our finding that strabismus is associated with an increase in the level of correlation between the OD activity in deep versus superficial cortical depths is novel, but in line with anatomical studies in V1 of strabismic animals suggesting increased segregation between ODCs with opposite ocular preference (Shatz et al., 1977; Lowel, 1994; Tychsen et al., 2004). Moreover, according to animal studies, shrinkage of ODCs in layer 4 after monocular deprivation is associated with decreased cytochrome oxidase activity of blobs that fall in register with the shrunken columns, suggesting a change in vertical connections spanning cortical layers (Horton and Hocking, 1997). However, this effect has never been tested in vivo in anisometropic and/or strabismic participants.
Amblyopia impacts beyond V1
Most amblyopia studies in animals have focused their efforts on understanding the impact of amblyopia on the primary visual cortex. While this impact is expected to extend to downstream areas, only a few studies have examined this phenomenon in the extrastriate visual cortex. Among them, Bi et al. reported that the increased OD response caused by strabismus extends to V2 (Bi et al., 2011). However, this extension was only detected in animals with severe amblyopia, suggesting a link between downstream extrastriate extension and visual impairment.
Consistent with that report, we showed that the correlation between the level of OD response and the interocular visual acuity difference, a functional measure correlated with ocular dominance shift, increased from V1 to downstream visual areas such as V3 and V4. This increase in correlation was detected despite the decrease in the OD response amplitude, suggesting that canonical propagation of OD deficits in amblyopia reflects functional visual impairment and highlighting the clinical relevance of downstream visual areas for future studies of evoked activity in the amblyopic brain.
Amblyopia impacts on fixation stability and visual attention
Previous studies have shown that fixation instability is significantly larger in the amblyopic eye compared with the fellow eye (Schor and Hallmark, 1978; Subramanian et al., 2013; Chung et al., 2015; Kelly et al., 2019; Ghasia and Wang, 2022) and in strabismic compared with anisometropic individuals (Chung et al., 2015; Kelly et al., 2019). This pattern is opposite to our reported impact of anisometropia and strabismus on the level of OD response. Specifically, we have shown that, consistent with the animal model (Fig. 1), the impact of amblyopia on the spatial distribution of OD activity is stronger in anisometropic compared with strabismic individuals (Fig. 2). Moreover, in strabismic individuals, we found an apparent interhemispheric difference in the distribution of OD activity. Considering that both hemispheres receive input from the amblyopic eye, this hemispheric laterality could not contribute strongly to the fixation instability in strabismic individuals. Considering these points and that fixation target remained visible to both eyes across all experimental conditions (see Materials and Methods), fixation instability caused by amblyopia is unlikely to contribute to our major findings.
It could be argued that the reported correlation between the OD response and the interocular visual acuity difference is a result of amblyopia's impact on the participant's attention control mechanism, influencing both measurements concurrently. Degraded visual attention in amblyopia has been previously reported (Ho et al., 2006; Hou et al., 2016; Verghese et al., 2019). To reduce the influence of attentional bias that may confound OD responses and their correlation with visual acuity, two separate steps were taken: First, the OD response was measured while the participant's attention was directed to an orthogonal task (i.e., shape change detection for the fixation object) separate from the stimuli used to elicit the OD measurement. Second, the fixation stimuli were presented dichoptically to reduce the potential impacts of biased attention in favor of the fellow eye. Thus, altered visual attention is unlikely to solely account for the strong OD response correlation with the visual acuity deficit in amblyopia.
The potential underlying mechanism for the increased OD response
Convergent evidence from both humans and nonhuman primates shows that amblyopia is associated with an increase in the level of OD response in early visual areas. However, the mechanism underlying this phenomenon remains unclear. In several mammalian species, it is widely accepted that monocular deprivation in the first few weeks of life leads to a drastic decrease in afferent input originating from the amblyopic eye to V1 (Hubel et al., 1976; LeVay et al., 1980; Horton and Hocking, 1997). However, the median age at diagnosis for amblyopia in humans is >3 years old in strabismic and >6 in anisometropic individuals (Shaw et al., 1988; Keech and Kutschke, 1995; Table 1). As demonstrated in anatomical studies in humans (Horton and Stryker, 1993; Horton and Hocking, 1996) and animals (Horton et al., 1997) with naturally occurring amblyopia, at this age, disruption of binocular vision would not be expected to change the number of thalamocortical afferent inputs to V1.
Amblyopia is also linked to changes in connections between ODCs. Anatomical studies have shown that strabismus and anisometropia are, respectively, associated with stronger and weaker segregation between ODCs (Shatz et al., 1977; Lowel, 1994; Tychsen et al., 2004). Horton and Hocking also showed evidence of a change in connections between layers 4 and 2/3 (the site of binocular convergence) after monocular deprivation (Horton and Hocking, 1997). However, the direction of this change (i.e., increased or decreased connectivity) remains unclear. Altered horizontal and/or radial connection between the ocular dominance columns may influence the ocular preference of V1 neurons and increase the OD response in amblyopic compared with nonamblyopic individuals. Longitudinal developmental studies are required to clarify the critical period for these effects and to test their correlation with the severity and distinct visual deficits of amblyopia.
Limitations
Despite recent advances in neuroimaging technologies (Polimeni et al., 2015; Blazejewska et al., 2019; Wang et al., 2022) that enabled us to map the OD response with relatively high spatial resolution, our techniques may still have missed even smaller OD patches, especially within the more peripheral portions of V1 (Adams et al., 2007). This caveat limits the interpretation of OD maps (Figs. 2–5). For instance, a relatively large patch that shows a uniform preference for one eye may contain small patches that are inaccessible due to limitations in spatial resolution.
To the best of our knowledge, no previous study has shown evidence that amblyopia impacts vascularization of the visual cortex. Nevertheless, the existence of diving veins (Duvernoy et al., 1983) may have influenced our estimation of the impact of amblyopia on the columnarity of OD response. Moreover, proximity to the pial vessels may contribute to the detected stronger fMRI signal in superficial versus deep layers (Koopmans et al., 2010; Polimeni et al., 2010; De Martino et al., 2013; Nasr et al., 2016). Thus, any interaction between amblyopia and cortical depth must be assessed carefully and reexamined using fMRI sequences less sensitive to vascularization (Yacoub et al., 2007; Huber et al., 2015; Akbari et al., 2023). Unfortunately, these methods [e.g., spin echo and/or vascular space occupancy (VASO)] have low contrast-to-noise sensitivity that limits their application for assessing the mesoscale organization of the human brain.
Due to the small size of the head coil used in 7 T scanners, we could not use the goggles designed for lower field (e.g., 3T) scanners to avoid using anaglyphic goggles that rely on colorful stimuli and to expand the stimulation field (Nasr et al., 2020). The anaglyph approach remains often used in studies of monocular visual stimulation and disparity encoding. Here, the potential effect of colored stimuli was reduced by counterbalancing the color filters across the sessions.
For the same reason, we could not use accessories designed to correct visual acuity in those individuals who exclusively wore glasses, to stimulate more of the peripheral visual field (r < 10°), and/or to monitor eye movements. While the impact of microsaccades and/or fixation instability on the fMRI signal is expected to be small, and our control experiments suggested that lack of optic correction and strabismus are unlikely to be responsible for the increased OD response in amblyopic individuals, these limitations prevented us from including individuals who required high degrees of optical correction and/or those who showed larger eye misalignments.
Conclusion
Despite its high prevalence in humans, our understanding of how amblyopia impacts the mesoscale organization of the visual system has been based primarily on animal models. In this study, high-resolution fMRI has documented the impact of amblyopia on the evoked OD response with functional correlates and has drawn distinctions between the impact of anisometropia and strabismus on cortical responses.
Footnotes
This work was supported by NIH NEI (R01 EY030434, EY029713, and K08 EY030164) and by the MGH/HST Athinoula A. Martinos Center for Biomedical Imaging. Crucial resources were made available by an NIH Shared Instrumentation Grant S10-RR019371. We thank Drs. Jonathan Polimeni, Jason Stockmann, and Ms. Azma Mareyam for helping with developing data acquisition sequences and hardware maintenance during this study. We thank Ms. Amanda Nabasaliza for her help with the recruitment. We also thank Drs. Daniel Tso and Jonathan Horton for their helpful comments.
The authors declare no competing financial interests.
- Correspondence should be addressed to Shahin Nasr at shahin.nasr{at}mgh.harvard.edu.