Introduction

Amblyopia is characterized by the developmental impairment of spatial vision, which cannot be corrected by eye refractive instruments. Largely a cortical disorder, it may result from abnormal visual experience in early childhood [1, 2]. It is accepted that a central rather than a retinal deficit is the major factor in the pathogenesis of amblyopia. Animal studies suggest that amblyopia represents functional and morphological influences of visual deprivation on the lateral geniculate nucleus (LGN) and visual cortex [3, 4]. However, the neuroanatomical information on structural changes in the human brain remains limited and inconclusive. Previous neuroimaging studies commonly focused on adults with amblyopia [5–7] and studies concerning the development of the visual system in children with amblyopia are scarce.

That the visual cortex plays a critical role in the pathogenesis of amblyopia has been confirmed using various techniques, and reduced functional activation in the calcarine cortex has been found [8]. The reduced contrast sensitivity (pattern vision) in amblyopia is coupled with deactivation in identified areas of the occipital visual system [9], and magnetoencephalographic responses in the occipital cortex triggered through the amblyopic eye have longer latencies and smaller amplitudes [10]. Electrophysiological studies have also shown that the visual cortical response to stimulation at higher spatial frequencies is reduced in anisometropic amblyopia [11]. Global brain gray matter [12], including the visual cortex [13], has been investigated physiologically and psychopathologically using functional magnetic resonance imaging (fMRI). Furthermore, recent fMRI studies have shown that amblyopia is associated with deactivation in the striate cortex (Brodmann area 17), the occipitotemporal cortex, and the LGN [14, 15]. In addition, the gray matter volume of the visual cortex is reduced in adults and children with amblyopia, as revealed by voxel-based morphometry (VBM) [16]. However, it has been pointed out that VBM may cause inaccurate assessment in regions of cortical overlap [17]. Surface-based morphometry is feasible to precisely quantify and characterize the cortical thickness in different regions [18–20]. By using this method, comparison of the cortical thickness in the four lateral occipital regions (the medial lingual cortex, the cuneus, the superior occipital cortex and the lunate cortex) showed that these regions are significantly thinner in amblyopic patients than in controls [7].

Diffusion tensor imaging (DTI) reflects water diffusion anisotropy in axons to provide insight into the geometric organization of the white matter. DTI-based diffusion tensor tractography (DTT) can non-invasively provide quantitative structural and functional information on the white matter in vivo and is widely used to evaluate white matter integrity [21]. During brain maturation, the total water content declines as myelination develops. Because these maturational processes result in changes of water diffusion, information on the magnitude and severity of any alterations in anisotropy may be important for understanding the pathophysiological features of developmental abnormalities [22, 23]. The integrity of the optic radiation (OR) fibers that are directly connected with the visual cortex also plays an important role in amblyopia. Fractional anisotropy (FA), an imaging measure derived from DTI data, can be used to assess the directionality of molecular diffusion. It has been applied to investigate the maturity and plasticity of white matter fiber tracts [24]. A recent study on amblyopia combined the DTI and DTT techniques and found clear reductions in the FA values of visual fiber tracts in amblyopic patients [25, 26]. Furthermore, the integrity of visual fiber tracts is compromised in amblyopia [5, 6]. The depicted three major bundles of ORs (the anterior, central, and posterior bundles) by the fiber-tracking method were in good agreement with the classic anatomic topography of the visual pathway [27]. In light of the results from previous studies, it may be hypothesized that changes in the visual fiber tracts elicited by amblyopia are associated with abnormalities of the visual cortex. It has been found that the anatomical connections of the OR constrains the fMRI response of the visual cortex [28]. However, the relationship between the thickness of the visual cortex and the integrity of the OR in children with anisometropic amblyopia has not been studied.

In this study, we used surface-based morphometry and DTT to investigate the relationship between cortical thickness and OR integrity in amblyopia. Since the OR fibers are directly connected with the visual cortex, clarifying this correlation may help understand the developmental mechanism of amblyopia.

Participants and Methods

Participants

All participants were recruited from the Ophthalmology Clinic at Xijing Hospital, Fourth Military Medical University, Xi’an, China. Anisometropic amblyopia was defined as 1.5 diopter (D) spherical and/or 1.0 D cylindrical refractive error difference and a best-corrected visual acuity <16/20 according to the established standard of prevention and control for amblyopia and strabismus [29]. Normal acuity was defined as 20/20 or better. The exclusion criteria were: (1) a known organic brain disorder, (2) specific clinical evidence of neurological dysfunction, (3) congenital or acquired organic pathology of the eye, or (4) long-term treatment of the eye. Fifteen children with anisometropic amblyopia and fifteen gender- and age-matched healthy controls were recruited; all were right-handed. In the patient group, the best-corrected visual acuity of the amblyopic eye ranged from 2/20 to 10/20 and that of the other eye from 16/20 to 20/20. In the control group (children with normal vision), the visual acuity of both eyes was 20/20 or better. The controls were recruited from primary schools based in the surrounding area. They all underwent comprehensive examination of the eye to ensure normal acuity and no neurological problems.

All participants were given MRI and DTI scans. This study was approved by the Ethics Committee of the Fourth Military Medical University. Written informed consent was given by all participants before entering the study.

MRI Data Acquisition

All participants wore headphones with their eyes closed and lay in a supine position. They were imaged using the 12-channel head coil of a 3.0-Tesla MR scanner (Magnetom Trio, Siemens AG, Erlangen, Germany) in a dim scanning room. Form padding was used to limit head movement. Each participant received two types of scan. A high-resolution three-dimensional magnetization-prepared rapid acquisition gradient-echo (MPRAGE) T1-weighted sequence was used to acquire MRI data from the whole brain (176 sagittal slices). The MRI parameters were as follows: TR = 1900 ms, TE = 2.26 ms, TI = 900 ms, flip angle = 9°, acquisition matrix = 256 × 256, FOV = 220 mm, slice thickness = 1.00 mm, inter-slice gap = 0 mm, isotropic resolution = 1.0 × 1.0 × 1.0 mm3, total time = 7 min and 37 s. DTI was acquired using single-shot diffusion-weighted echo-planar imaging (DW-EPI) with the following acquisition parameters: TR = 6100 ms, TE = 71.1 ms, matrix = 128 × 128, FOV = 240 mm × 240 mm, slice thickness = 3 mm, gradient directions = 20, b values = 0 s/mm2 and 1000 s/mm2, total time = 7 min and 51 s.

Cortical Thickness Estimation

Cortical reconstruction and parcellation were performed with the FreeSurfer V4 package (v 4.0.2) image analysis suite [30], which is documented and available at http://surfer.nmr.mgh.harvard.edu/. This method has been fully described [31] and validated via both histological and manual measurements [32, 33]. For cortical analysis, the gray/white matter interface in each hemisphere was segmented, tessellated, corrected for topological errors, and inflated to unfold the cortical surface. Then, each individual hemispheric surface was registered to a template based on a surface with its sulco-gyral pattern to allow inter-individual analysis [30]. Cortical thickness was calculated from both intensity gradients across tissue class and continuity information, as the closest distance between the gray/white matter border and the gray/cerebrospinal fluid interface at each vertex of the tessellated surface [30]. Automated topology correction was performed as necessary [34]. The cortical thickness values were obtained from the statistical output of FreeSurfer cortical parcellation, called “aparc.a2005s.stats”, using the Desikan-Killiany atlas and containing 75 regions. The data from each participant were resampled to an average participant, and a ‘qcache’ command and 10-mm full-width half-maximum Gaussian kernel were used to perform surface smoothing prior to statistical analysis.

DTI Data Processing

DTT was processed using TrackVis software (http://www.trackvis.org/), which can visualize and analyze fiber tract data from DTI [35]. The mean FA values of the ORs were calculated. The OR fiber bundles were identified on the RGB (red–green–blue) map, using anatomical atlases obtained from children with anisometropic amblyopia and controls. In the RGB map, five ROIs were selected from the bilateral ORs to measure the mean FA values. Each ROI area was designed for 9 pixels uniformly and placed symmetrically in both ORs as described in previous studies [36–38]. Fibers were clustered into bundles using the normalized cut algorithm. A pairwise distance matrix between all fibers was required for clustering. The distance between fibers was defined as the mean of the closest point distances of all points along each fiber. A mean 3D curve attributed with tensors representing average tensors calculated within cross-sections was used to model each bundle. The number of tracks of ORs longer than the mean length (43.24 mm) was used as the coordinate system for statistical analysis.

Statistical Analysis

Cortical Thickness Analysis

QDEC software (https://surfer.nmr.mgh.harvard.edu/fswiki/Qdec), which is a FreeSurfer application, was used to explore the differences in cortical thickness between groups. Group differences were shown via statistical parametric maps of the entire cortical mantle. A general linear model was estimated at each vertex across the cortical surface, with cortical thickness as dependent variable, gender as a categorical predictor, and age as a continuous predictor. The mean regional cortical thickness value was obtained for further Pearson correlation analysis between cortical thickness and the FA values of the ORs.

Correlations between DTI and Cortical Thickness

To investigate the correlations of mean FA values for ORs with cortical thickness in amblyopic children, the extracted mean regional cortical thickness and FA value were imported into SPSS 17.0 software (SPSS Inc., Chicago, IL) for Pearson correlation analyses, and the critical level for statistical significance was set at P < 0.05 adjusted for multiple comparisons using false discovery rate correction according to the nine clusters with reduced cortical thickness for which correlations were performed (P level adjusted to <0.028) [39].

Results

Clinical Data

There were no significant differences between groups in gender (children with anisometropic amblyopia: male/female, 9/6; normal children: 8/7; χ2 = 0.136, P = 0.7125), age (8.0 ± 1.3; 7.9 ± 0.8; P = 0.8176), or handedness (right/left: 15/0; 15/0) (nonparametric tests). The detailed demographical characteristics and clinical data are listed in Table 1.

Table 1 Clinical characteristics of amblyopic children
Full size table

Brain Imaging Results

Cortical Thickness

There was no significant group difference in the whole-brain average cortical thickness. But the children with anisometropic amblyopia showed significantly lower cortical thickness than controls in the following sub-regions of both cerebral hemispheres: medial lingual cortex, lateral occipitotemporal gyrus, lateral occipital cortex, cuneus, superior occipital cortex, inferior parietal lobe, medial occipital cortex, and lunate cortex. The calcarine gyrus was thicker in these children than in controls (Fig. 1, Table 2).

Fig. 1
figure 1

Regions with alterations of cortical thickness in children with anisometropic amblyopia relative to normal controls. A Regions with significant cortical thickness changes in a lateral view of the left hemisphere with the inflated model. B Regions in a lateral view of the right hemisphere. C Regions in a medial view of the left hemisphere. D Regions in a medial view of the right hemisphere. The color map overlays the reconstruction image, age and gender effects were removed by regression, and thresholded at P < 0.05 corrected for multiple comparison. The color-code for T value is on a logarithmic scale of −5.1–5.1, shown below. Warmer colors (positive values) represent cortical thinning; cooler colors (negative values) represent cortical thickening.

Full size image
Table 2 Differences in cortical thickness between children with anisometropic amblyopia and controls
Full size table

Diffusivity

The FA values of ORs in amblyopia were significantly lower than in controls (Fig. 2). The number of reconstructed OR fibers did not differ significantly between the two groups (Table 3). Also, there was no significant difference in the mean FA value, voxel value, and number of ORs between the bilateral ORs in amblyopic children and controls (Table 3).

Fig. 2
figure 2

Representative images showing high-resolution reconstruction of DTI fiber tracks of ORs. Left panel representative tracts of ORs of a healthy child, right panel representative tracts of ORs in a child with anisometropic amblyopia. Inserts show significant differences between the DTT reconstructed images in amblyopia relative to controls.

Full size image
Table 3 FAs, voxel values, and numbers of ORs from amblyopic children and controls, including ipsilateral and contralateral ORs
Full size table

Correlations of OR Integrity with Cortical Thickness in Regions of Interest

There were positive correlations between cortical thickness and the FA values of ORs in the medial lingual cortex (MLP) (r 2 = 0.331, P = 0.0253), superior occipital gyrus (SO) (r 2 = 0.361, P = 0.0181), and occipital temporal medial and lingual (OTML) cortex (r 2 = 0.395, P = 0.0163) in the left hemisphere; and in the lateral occipitotemporal gyrus (LOTG) (r 2 = 0.448, P = 0.0091), MLP (r 2 = 0.365, P = 0.0174), SO (r 2 = 0.343, P = 0.0225), and OTML (r 2 = 0.492, P = 0.0079) in the right hemisphere. There was a trend of positive correlations in the left LOTG (r 2 = 0.302, P = 0.0341) and middle occipital and lunate cortex (OML) (r 2 = 0.272, P = 0.0469) and the right OML (r 2 = 0.273, P = 0.0457). However, there was no significant correlation in the following cortical regions: the inferior parietal lobe, calcarine gyrus, cuneus, lateral occipital cortex, and occipital temporal-parietal cortex (Fig. 3).

Fig. 3
figure 3

Correlations of OR fractional anisotropy with mean cortical thickness in regions with significant cortical changes in patients with anisometropic amblyopia. X-axes mean cortical thickness, Y-axis mean FAs of ORs (Pearson’s correlation analysis; r and p values at lower right). MLP, medial lingual cortex; LOTG, lateral occipital temporal gyrus; SO, superior occipital gyrus; OML, middle occipital and lunate cortex; OTML medial occipital temporal cortex.

Full size image

Discussion

Our main findings were that (1) the FA values of ORs in children with anisometropic amblyopia were significantly lower than in controls, but there was no significant difference in FA values between the bilateral ORs; (2) the cortical thickness was significantly lower in the bilateral occipital lobe, lingual gyrus, cuneus, and lunate cortex, and significantly higher in the calcarine cortex, in amblyopic children than in controls; and (3) there were significant correlations between FA value and average cortical thickness in the medial lingual cortex, lateral occipitotemporal gyrus, lateral, superior, and middle occipital cortex, and lunate cortex. All these results suggest that the reduced thickness in the visual cortex is due to OR damage in children with anisometropic amblyopia, reflecting projection impairments in their functional networks.

Decreased gray matter volume and density in different regions have been revealed by VBM [16, 40], but this method does not have specificity for highly convoluted structures of the cortex [41]. SBM is a recently-developed and highly reliable MRI-derived neuroanatomical measure, which overcomes the shortcomings of volume-based analysis, and is unaffected by voxel geometry [42]. In this study, we found thinning of the bilateral occipital lobe, lingual gyrus, cuneus, and lunate cortex in amblyopia. fMRI has also indicated that adults and children with amblyopia have functional deficits of the visual cortex [7]. From the electrophysiological perspective, in senescent individuals, the age-associated decline of selectivity of visual cortical neurons to spatial frequency might contribute to the decreased visual acuity [43]. Vision is relatively poorly developed at birth; normal acuity is achieved at about three years of age, given proper visual stimulation in the early months and years after birth [44]. If the stimulation is anomalous for the receptors, vision does not develop properly, leading to amblyopia [2]. It is well established that amblyopia is a failure of the cortical developmental process [45–48], although the exact extent of the visual cortex deficit is largely unknown. Previous studies have shown that the thickness of visual cortex in blind patients can manifest compensatory thickening [41, 49]. The increased thickness of the calcarine sulcus in this study could reflect more visual stimulation in the primary visual cortex, and an increase in connectivity in this region.

It is believed that FA in DTI can provide microstructural information about white matter development [22]. Furthermore, DTT enables extraction of the trajectories of certain fiber tracts in vivo. Current DTI results show that in children with anisometropic amblyopia, the FA values of ORs are significantly lower than in controls. This has been verified in a similar investigation in the previous study, which also found more voxels in the posterior ORs of controls than in the amblyopic children [50]. Sampling of high-spatial-frequency components of visual stimuli in visual cortex revealed the main dysfunction in anisometropic amblyopia [51]. A reduction of FA implies either a difference of such fiber architecture such as axon density, size, orientation, and internal structure, or a difference in myelination of the fibers. In our study, tractography showed a reduction of FA in the LGN of bilateral ORs of amblyopic children, which indicated that the controls had better OR development. It is known that children’s brains undergo an extended period of postnatal maturation, with increasing white matter as well as progression of myelination. Recent advances in neuroscience have provided evidence of an effect of training on white matter development [52]. It has been routinely reported that visual-evoked potentials in human amblyopes are reduced and distorted, indicating dysfunction of the visual pathway [53, 54]. In animal models, it has been demonstrated that neuronal activity triggers the induction of myelination and that myelination of the optic nerve is decelerated when animals are reared in darkness [55]. As such, the anomalous development of ORs found in our study may result from abnormal neuronal signals and weak connections. Another explanation for the current results may be that the myelination of ORs was affected by the low acuity of the amblyopic eye; this has also been found in another study [56]. It might be associated with the fact that the ORs receive visual inputs from both eyes. Anisotropic diffusion in white matter may be due to the effect of myelination, axonal thickness, amount of parallel organization of axons, or a combination of these factors [57]. In a previous study, the fiber tract organization was found to be positively correlated with practice, and the increased myelination induced by neuronal activity in fiber tracts during training contributes to the increase in FA [24]. Therefore, well-developed ORs would be expected to have higher FA values, and lower FAs in bilateral ORs indicate that children with anisometropic amblyopia exhibit abnormalities in both ORs, which has been demonstrated previously [50].

Because the ORs are bidirectional, we believe that their underdevelopment in amblyopia is probably related to changes in both the visual cortex and LGN. DTI and fMRI have revealed that the FA values of ORs are positively correlated with functional activity of the visual cortex [28], and the functional deficit in amblyopia is accompanied by changes in the cortical volume of the occipital lobe [58]. These results imply that the impairment of white matter integrity associated with dysfunction-related cerebral cortex, i.e., the visual cortex, may accompany structural deficit in the visual cortex. The ability of amblyopia patients to process visual impulses from the LGN is reduced, which directly affects OR white matter development, leading to a decreased FA value, meanwhile causing dysfunction of the visual cortex. However, the extent to which OR integrity changes with the thickness of the visual cortex in amblyopia was unclear. This study is the first to investigate the relationship between deficits in the integrity of ORs and cortical thinning in anisometropic amblyopia. We found that the OR-related visual cortex changes were all in the V1 area (LOTG, OML, SO, OTML, and MLP). But cortical thinning in other regions (IFL, cuneus, and LOTP) was not associated with the impairment of OR integrity. In patients with anisometropic amblyopia, the activation by higher-spatial-frequency stimuli in visual cortex is reduced [51]. Visual input can be processed separately by different types of visual cortical neurons [59], so the related OR fibers may experience dysfunction, causing the subregional cortical thinning in the related visual cortex. The detailed mechanism still needs further studies investigation. In this study, the reduction in OR FA value was not correlated with reduced cortical thickness in some regions. A possible explanation is that the cortical thickness changes in those regions may be secondary to deficits in the primary visual cortex, in other words, they are compensative effects for neuron agenesis. It is also possible that cortical thickness changes in the inferior parietal lobe are caused not only by the reduction of visual information from the retina, but also by changes in cognitive activities, such as the basic visual sensation and perception [28].

The present work has several limitations, which we plan to address in future studies. First, the sample size is small. We had only 15 children with anisometropic amblyopia and 15 controls, which may hinder the detection of subtle cortical thinning. Second, we only studied patients with anisometropic amblyopia but not those with strabismic amblyopia. Comparison between the two types of amblyopia will help better understand their pathogenic mechanisms. Third, the ROI approach is not optimal to map the whole-brain white matter. Partial volume effects on gray matter and/or CSF in the white matter may blur DTI measures [55]. A histogram-based or voxel-based statistical approach may overcome the partial volume effects. However, these methods, which require a smoothing step and a normalization scheme, might limit the sensitivity of anisotropic measurements. Fourth, we did not investigate the correlations among OR integrity, cortical thickness, and visual acuity loss or duration of amblyopia; we will do these in further work.

In conclusion, we report impaired OR integrity and cortical thickness reduction within the visual cortex in anisometropic amblyopia. The reductions of thickness in subregions of visual cortex were correlated with impairment of OR fiber integrity. These findings indicate that amblyopia can cause anomalous development in both ORs and visual cortex.