Anterior visual pathway assessment by magnetic resonance imaging in normal-pressure glaucoma

Authors


  • Both authors equally contributed to the study.

Liang Xu, MD and Jost B. Jonas, MD
Beijing Institute of Ophthalmology
17 Hougou Lane
Chong Wen Men
100005 Beijing
China
Tel.: + 86 10 58265918
Fax: + 86 10 6512 5617
Email: xlbio1@163.com

Abstract.

Purpose:  To analyze the axonal architecture of the optic nerve in patients with normal-pressure glaucoma and determine whether these parameters correlate with the disease severity.

Methods:  Using magnetic resonance (MRI) imaging (1.5-Tesla unit) and diffusion tensor (DT) MRI, we measured the optic nerve diameter, optic chiasm height and lateral geniculate nucleus (LGN) volume in patients with normal-pressure glaucoma and an age-matched control group. The retinal nerve fibre layer thickness (RNFL) was determined by optical coherence tomography (OCT).

Results:  The study included 30 patients with normal-pressure glaucoma and 30 age-matched control subjects. Optic nerve diameter (p < 0.001), optic chiasm height (p < 0.001) and LGN volume (p = 0.02) were significantly smaller in the glaucoma group than in the control group and were significantly correlated with RNFL thickness and perimetric loss. In the control group, the parameters significantly (p < 0.05) decreased with age. The DT-MRI-derived fractional anisotropy for the optic nerve was significantly lower (p < 0.001), and the DT-MRI-derived mean diffusivity (p < 0.001), radial diffusivity (λ; p < 0.001) and axial diffusivity (λ||; p = 0.009) for the optic nerve were significantly higher in the glaucoma group and significantly correlated with RNFL thickness and mean perimetric defect.

Conclusions:  Patients with normal-pressure glaucoma show an age-adjusted reduced optic nerve diameter, optic chiasm height and LGN volume as measured by MRI, correlating with a reduced RNFL thickness and increased perimetric loss. MRI may be applied to examine the optic nerve in patients with glaucoma with opaque optic media.

Introduction

Using magnetic resonance imaging (MRI) of the anterior visual pathway, previous studies have shown that glaucomatous optic neuropathy affected the anterior visual pathway at least up to the optic chiasm and that these morphologic changes in the anterior visual pathway were correlated with glaucomatous optic nerve damage (Iwata et al. 1997; Yücel et al. 2001; Kashiwagi et al. 2004). As compared with the vertical cup/disc diameter ratio, the assessment of the height of the chiasm as measured by MRI was a better predictor of visual field defects. Consequently, in patients with advanced glaucomatous visual field loss, the optic chiasm was markedly atrophic (Iwata et al. 1997). In a similar manner, experimental investigations on monkeys with induced glaucoma have shown glaucoma-related changes of the lateral geniculate nucleus (LGN) and visual cortex (Weber et al. 2000; Yücel et al. 2001; Gupta et al. 2007; Ito et al. 2009). In a clinicopathological report of a patient with glaucoma, Gupta et al. (2006)were the first to demonstrate in patient glaucoma-related degenerative changes of the brain involving the intracranial optic nerves, the LGN and the visual cortex. Consequently, it has been proposed that glaucoma may be considered a central nervous system (CNS)-based degenerative disease that is influenced by ocular factors such as intraocular pressure (Gupta & Yücel 2007).

Diffusion is the random motion of molecules in any fluid environment such as biological systems. Water diffusion in biological tissues such as white matter occurs preferentially parallel to the orientation of axons (Takahashi et al. 2000). Such diffusion has been known as anisotropic diffusion that depends on the structural environment of white matters. Diffusion tensor imaging (DTI) was therefore developed to provide a complete description of water diffusion in three dimensions (Basser et al. 1994). DTI has found its applications in the study of pathological changes in the peripheral nervous systems and CNSs such as the optic nerve of a mouse model of retinal ischaemia (Beaulieu et al. 1996; Song et al. 2003; Sun et al. 2006), and the spinal cord of mice with an experimental autoimmune encephalomyelitis (Kim et al. 2006). DTI has thus become a widely recognized imaging modality used to study the connectivity and integrity of white matter in CNS tissues (Le Bihan et al. 2001; Huppi & Dubois 2006; Mori & Zhang 2006). The DT can be represented by an anisotropic diffusion displacement-probability ellipsoid, characterized by three eigenvalues (λ1, λ2, and λ3) and three eigenvectors in a local frame of each image voxel after matrix diagonalization. The average of the three eigenvalues is referred to as the mean diffusivity (MD; with MD = (λ1 + λ2 + λ3)/3). In the CNS, white matter, λ1, the largest eigenvalue, represents the water diffusivity parallel to the axonal fibres. It is referred to as λ||, the axial diffusivity. The water diffusivities perpendicular to the axonal fibres, λ2 and λ3, are averaged and referred to as λ = (λ2 + λ3)/2, the radial diffusivity (Song et al. 2002, 2003).

Based on the findings of the previous studies on glaucoma-related changes in the CNS and considering the potentials of the new imaging technologies, the aims of our study were to assess the optic nerve diameter, the height of optic chiasm and the volume of LGN in patients with normal-pressure glaucoma, to analyze the axonal architecture of the optic nerves by DTI-MRI, and to correlate the measurements of the CNS structures with measurements of the retinal nerve fibre layer (RNFL) thickness using optical coherence tomography (OCT).

Materials and Methods

The prospective observational study included patients with normal-pressure glaucoma as study group and a control group of normal subjects. The Medical Ethics Committee of the Beijing Tongren Hospital approved the study protocol, and all participants gave informed consent according to the Declaration of Helsinki. The rationale for choosing normal-pressure glaucoma for the study was that patients with normal-pressure glaucoma had undergone brain imaging to exclude an intracranial tumour as potential cause for their disease (Qu et al. 2011). Normal-pressure glaucoma was defined by a glaucomatous appearance of the optic nerve head characterized by a neuroretinal rim shape not fulfilling the inferior–superior–nasal–temporal (ISNT) rule (Jonas et al. 1988a,b) usually showing a clear notch in the neuroretinal rim; an optic cup too large for the size of the optic disc, disc haemorrhages and localized or diffuse defects of the RNFL (Jonas & Schiro 1994); and an open anterior chamber angle upon gonioscopy. If there was no clear neuroretinal rim notch, an abnormal rim configuration had to be present in association with an abnormal RNFL; the presence of a disc haemorrhage was optional. Presence of visual field defects was not a necessary inclusion criterion for the diagnosis of normal-pressure glaucoma. The intraocular pressure measurements corrected for central corneal thickness (measured by Lenstar Biometry, LS900; Haag-Streit Co, Berne, Switzerland) did not exceed 21 mmHg in 24-hr intraocular pressure profiles (Suzuki et al. 2005). Exclusion criterion was any other ocular disease except of mild cataract, and any history of extensive blood loss and head trauma. The control group consisted of healthy individuals who were matched in age with the patients in the glaucoma group. All control subjects underwent an ophthalmologic examination, which included measurement of visual acuity and fundus examination including fundus photography. Inclusion criteria were a best corrected visual acuity >20/30 and a normal appearance of the fundus. Any diseases of the retina or optic nerve were exclusion criteria.

All study participants underwent a detailed ocular examinations including best corrected visual acuity, applanation tonometry, slit-lamp-based biomicroscopy of the anterior and posterior segment of the eye, gonioscopy, ocular biometry (Lenstar Biometry), 45° fundus photography using a non-mydriatic digital fundus camera (CR-DGI camera; Canon Inc., Tokyo, Japan), computerized perimetry [Humphrey field analyzer (HFA) 30-2 (Carl Zeiss Meditec AG, Jena, Germany); SITA Fast strategy] and spectral domain OCT of the peripapillary RNFL (Spectralis®; Heidelberg Engineering, Heidelberg, Germany). For the OCT examination, the RNFL was scanned in a peripapillary circle with a diameter of 3.4 mm. Reliability criteria for the automated visual field testing were fixation losses of <20% and a rate of false-negative errors and false-positive errors of <33%. The criteria for acceptable SD-OCT fundus images were absence of large ocular movements (seen as no abrupt shift of large retinal vessels in the projection image) with a quality factor exceeding 60%.

In addition to the ophthalmological examination, all patients underwent MRI using a 1.5-T magnet (General Electric Co., Milwaukee, WI, USA). The MRI scans were evaluated by two experienced radiologists (ZCW, JL) masked to the diagnosis and status of the optic nerve. Each subject was additionally scanned using a high resolution T2-weighted fluid-attenuated inversion recovery (FLAIR sequences) sequence (TR = 9000 ms, TE = 120 ms, TI = 2125 ms, slices = 32, slice thickness = 4 mm, slice gap = 0.8 mm, field of view (FOV) = 240 × 210 mm2, matrix size = 256 × 222) to detect any brain abnormalities. Coronal optic nerve images were obtained with an eight-channel head coil using spin-echo EPI sequence with parallel acquisition. The coronal-oblique slices were set orthogonal to the nerves (Fig. 1). The slices were set at eight locations equidistant to each other and covering the whole region between optic nerve head and the apex of the orbit. The diameter of the retrobulbar optic nerve was measured at 5, 10 and 15 mm behind the eye. The bright liquor ring was well delineated and exhibited a high contrast to its surrounding tissue. The optic nerve parenchyma was the hypointense area inside this hyperintense ring of cerebrospinal fluid (Fig. 1; Lagrèze et al. 2009; Weigel et al. 2006). For assessment of the dimensions of the optic chiasm, the section located immediately anterior to the pituitary stalk was examined. The vertical height of the right and left sides of the chiasm in this section were measured (Fig. 3). In our analyses, we related the right eye to the right side of the chiasm and the left eye to the left side of the chiasm. The measurement of the diffusivity changes in the optic nerve and optic radiation was described in a previous study by Li et al. (2011).

Figure 1.

 Magnetic resonance imaging to determine the diameter of the optic nerve in the orbit; the position of the slices was viewed on an axial localizer view of the optic nerve. There were eight slices between the anterior part of the optic nerve (adjacent to the optic nerve head) to the posterior part (near the orbital apex).

We used the following acquisition parameters for the optic nerve: one b0 and six non-collinear gradient directions with b = 600 seconds/mm2, slices = 8, thickness = 5 mm, skip = 0, FOV = 22 × 22 cm2, acquisition matrix size = 128 × 128, NEX = 16. The diffusion acquisition parameters of the optic radiation were the following: one b0 and 15 non-collinear directions with b = 1000 seconds/mm2, 22 contiguous axial slices with 4 mm thickness and no gap; repetition time/echo time = 6000/71.1 ms, FOV = 24 × 24 cm2, acquisition matrix = 128 × 128, NEX = 6. In addition, a whole-brain 3D T1 SPGR sequence (TR = 10 ms, TE = 4.4 ms, TI = 600 ms, slices = 152, flip angle = 15°, NEX = 1, slice thickness = 1 mm, FOV = 260 × 260 mm2, matrix size = 256 × 256) was used as a source image for the subsequent co-registration of the optic radiation and the measurement of LGN. The eddy current distortions and motion artefacts in the DT-MRI data were corrected by applying an affine alignment, using FMRIBs diffusion toolbox (FSL, Oxford, UK; Behrens et al. 2003; Kamali et al. 2010). The MD, fractional anisotropy (FA) and eigenvalues (λ1, λ2, λ3) of optic nerve were calculated on a voxel-by-voxel basis using DTIStudio (mri Studio Software, Johns Hopkins University, Baltimore, MD, USA). The intraorbital 4th layer of the nerve (about 2.0 cm after the eye globe) was used for the following analysis. The regions of interests (ROIs) were defined manually on the b0-template (average non-diffusion-weighted) consisting of two square 2 × 2 voxels (Fig. 2). The ROIs were placed in the centre of the nerve. Then mean values of MD, FA and eigenvalues from these ROIs were obtained.

Figure 2.

 Creation of a region of interest.

The boundaries of the LGN were defined in accordance with the procedure adapted from Korsholm’s method (Korsholm et al. 2007). The volume of LGN was estimated to compare the differences between the glaucoma and controls briefly: first, to highlight the landmarks for LGN localization by taking advantage of intensity threshold adjusting and colour mapping; second, to get the candidate voxels of the LGN with a region growing algorithm; third, to isolate the LGN with a ventral diencephalon (VDC) mask that accounted for the spatial information, the MR scanner parameters and morphological differences between the atlas and the each individual; finally, to discriminate the LGN from surrounding white- and grey-matter structures as the intersection of volume of interest and VDC mask. The procedure was performed on all relevant slices with a mouse-controlled cursor, with boundaries displayed in real time. This routine required around 6–8 min to generate the LGN result per subject (for two hemispheres). The section located immediately anterior to the pituitary stalk, which best showed the optic chiasm, was selected. The vertical heights of the right and left sides of the chiasm in this section were measured (Fig. 3). In our analyses, we related the right eye to the right side of the chiasm and the left eye to the left side of the chiasm.

Figure 3.

 Example of a measurement of the height of optic chiasm.

All the measurements on the MRI images were independently carried out by two investigators (ZCW, JL) to determine the inter-observer reliability (Table 6). These measurements were then repeated by the same investigators after two weeks to assess intra-observer reliability. Each investigator performed the measurements three times to get the mean value, with the final results being the average of the measurements by the two investigators.

Statistical analysis was performed using a commercially available statistical software package (spss for Windows, version 17.0; SPSS, Chicago, IL, USA). The data of only one randomly chosen eye per study participant were included into the statistical analysis. Data were reported as mean values ± standard deviations. Differences between the groups were evaluated with two-tailed paired-samples T test. A p < 0.05 was considered to be statistically significant.

Results

The study group included 30 patients (15 men) with normal-pressure glaucoma and a mean age of 54.8 ± 11.9 years (range: 26–69 years), and the control group consisted of 30 healthy age-matched control subjects (16 men; age: 53.9 ± 11.2 years; range: 27–67 years). In the glaucoma group, 26 (87%) of the 30 patients showed visual field defects. The patients of the control group and the subjects of the study group did not differ significantly in age (p = 0.29).

In the study group and in the control group taken separately, the diameters of the optic nerve were significantly smaller in the orbital apex than in regions closer to the eye globe (1.21 ± 0.25 mm versus 1.31 ± 0.33 mm (p < 0.001) and 2.22 ± 0.25 mm versus 2.55 ± 0.37 mm, (p < 0.001), respectively). Comparing study group and control group with each other revealed that the optic nerve diameter was significantly smaller in patients with glaucoma (Table 1, Fig. 4).

Table 1.   Optic nerve diameter (mm; mean ± standard deviation) as measured by magnetic resonance imaging on orbital cross-sections of the optic nerve in 30 patients with normal-pressure glaucoma and 30 age-matched control subjects. p-value: statistical significance of the difference between both groups.
Distance to the optic nerve head (mm)Glaucoma groupHealthy control groupp-value
 51.31 ± 0.332.55 ± 0.37<0.001
101.25 ± 0.272.32 ± 0.32<0.001
151.21 ± 0.252.22 ± 0.25<0.001
Figure 4.

 Plot showing the distribution of the optic nerve diameter (mm) as measured by magnetic resonance imaging on orbital cross-sections of the optic nerve in 30 patients with normal-pressure glaucoma and 30 age-matched control subjects at locations of 5, 10 and 15 mm behind the optic nerve head.

To analyze the correlations among different parameters, we calculated a partial correlation matrix (controlled for age; Table 2). According to the r-values presented in Table 2, the structure–function relationship showed the highest correlations coefficients for the relationship between the mean perimetric loss [mean deviation (dB)] and the RNFL thickness measured by OCT (= 0.80). Correlation coefficients between the mean perimetric loss and optic nerve diameter measurements as determined by MRI were significantly (p = 0.03; Fisher r-to-z transformation) lower (rmax = 0.52; Table 2). The correlation coefficients were higher, however, not significantly (p > 0.05) higher, for the optic nerve diameter measurements obtained in the orbital apex than at 5 mm behind the eye. The optic nerve diameter at 15 mm behind the eye correlated best with the RNFL thickness (rmax = 0.74; Fig. 5). Again, correlation coefficients were higher and, however, not significantly (p > 0.05) higher, for the optic nerve diameter measurements obtained at 15 mm behind the eye in the orbital apex than closer to the globe.

Table 2.   Partial correlation matrix (controlled for age) between the perimetric loss, retinal nerve fibre layer thickness as measured by optical coherence tomography (RNFL-OCT), and optic nerve diameter at 5–15 mm behind the eye, as measured by magnetic resonance imaging (MRI).
 Perimetric defectMRI 5 mm behind the globeMRI 10 mm behind the globeMRI 15 mm behind the globeRNFL-OCT
Perimetric defect1    
MRI 5 mm behind the globe0.50 (0.005)1   
MRI 10 mm behind the globe0.43 (0.02)0.86 (<0.001)1  
MRI 15 mm behind the globe0.52 (0.004)0.89 (<0.001)0.97 (<0.001)1 
RNFL-OCT0.80 (<0.001)0.69 (<0.001)0.66 (<0.001)0.74 (<0.001)1
Figure 5.

 Scatterplot showing the correlation between the retinal nerve fibre layer thickness as measured by optical coherence tomography and the optic nerve diameter (mm) as measured by magnetic resonance imaging on orbital cross-sections of the optic nerve 15 mm behind the optic nerve head in 30 patients with normal-pressure glaucoma.

The height of the optic chiasm was significantly lower in the study group than in the control group (1.60 ± 0.33 mm versus 2.86 ± 0.39 mm; p < 0.001). The height of optic chiasm was significantly correlated with the mean perimetric loss (= 0.44, p = 0.02) and the RNFL thickness measurements (= 0.72, p < 0.001; Table 3).

Table 3.   Partial correlation matrix (controlled for age) between the optic chiasm height, mean perimetric defect and the retinal nerve fibre layer (RNFL) thickness as measured by optical coherence tomography.
 Optic chiasm heightMean perimetric defectRNFL thickness
Optic chiasm height1  
Mean perimetric defect0.44 (0.02)1 
RNFL thickness0.72 (<0.001)0.80 (<0.001)1

Volume of lateral geniculate nucleus

For the measurement of the volume of the LGN, the images of 28 patients with glaucoma and 28 control subjects were examined. The volume of the LGN could not be measured for two patients of the glaucoma group. One of these patients had a bilateral mean perimetric defect of more than 30 dB, and the volume of LGN was too small for a reliable measurement. The images of the other patient (mean perimetric defect of −11 dB in the right eye and of −10 dB in the left eye) showed a motion artefact. The volume of LGN was significantly smaller in the study group than in the control group (143.2 ± 22.2 versus 154.2 ± 16.5; p = 0.02).

Diffusion tensor imaging (DTI) of the optic nerve

For the DTI of the optic nerve, the study group included 17 patients with glaucoma and the control group consisted of 17 age-matched control subjects. The DT-MRI-derived FA for the optic nerve was significantly lower in glaucoma group (p < 0.001). The DT-MRI-derived mean MD values (p < 0.001), radial diffusivity (λ; p < 0.001) and the axial diffusivity (λ||; p = 0.009) for optic nerve were significantly higher in the glaucoma group (Table 4). The mean DT-MRI-FA values of the optic nerves were significantly and positively correlated with the mean perimetric defect (= 0.858, p < 0.001) (Table 5). As a corollary, the mean DT-MRI-MD of the optic nerves was significantly and negatively correlated with the mean perimetric loss (r = −0.77, p = 0.001).

Table 4.   Diffusion tensor imaging (DTI) of the optic nerve.
 Glaucoma groupControl groupp-value
  1. DTI-FA = diffusion tension imaging-derived fractional anisotropy value; λ1 value = DTI-eigenvalue representing the axial diffusivity; λ value = DTI-eigenvalue representing the radial diffusivity; DTI-MD = diffusion tension imaging-derived mean diffusivity value for the optic nerve.

DTI-FA value0.39 ± 0.090.54 ± 0.08<0.001
λ|| value2.83 ± 0.33 × 10−32.48 ± 0.30 × 10−30.009
λ value1.50 ± 0.34 × 10−30.99 ± 0.17 × 10−3<0.001
DTI-MD1.94 ± 0.31 × 10−31.48 ± 0.17 × 10−3<0.001
Table 5.   Partial correlation matrix between the mean perimetric loss and measurements obtained by diffusion tension imaging (correlation coefficients; p-values in brackets).
 Mean perimetric lossRNFL-OCTDTI-FAλ||λDTI-MD
  1. RNFL-OCT = retinal nerve fibre layer thickness as measured by optical coherence tomography; DTI-FA = diffusion tension imaging-derived fractional anisotropy value; λ1 value = DTI-eigenvalue representing the axial diffusivity; λ value = DTI-eigenvalue representing the radial diffusivity; DTI-MD = diffusion tension imaging-derived mean diffusivity value for the optic nerve

Mean perimetric loss1     
RNFL-OCT0.79 (<0.001)1    
DTI-FA0.86 (<0.001)0.73 (0.002)1   
λ||−0.38 (0.15)−0.28 (0.30)1  
λ−0.85 (<0.001)−0.68 (0.003)1 
DTI-MD−0.77 (0.001)−0.62 (0.010)1

Correlation between age and MRI data in normal control group

In the control group, the optic nerve diameter as measured at 5 mm (r = −0.39; p = 0.04), at 10 mm (r = −0.50; p = 0.005) and at 15 mm (r = −0.43; p = 0.02) behind the eye, the height of the optic chiasm (r = −0.42; p = 0.02) and the volume of the LGN (r = −0.55; p = 0.003) decreased significantly with higher age.

If the four patients from the glaucoma groups without visual field defects were excluded from the statistical analysis, the results of the analysis grossly remained unchanged.

The ROIs were selected by an experienced radiologist masked to all other data of the patient or subject. The ROIs of individual cases were mapped from the template b0 using an inverse transformation. These ROIs were then overlaid onto the data maps of the MD, FA and eigenvalues, so that the mean values from the four voxels could be obtained. This procedure was performed by two investigators independently of each other. The correlation coefficients for the data obtained by the two investigators ranged between 0.69 (for lFA4j-lFA4m) and 0.76 (for rMD4j-rMD4m). When the results obtained by the two investigators were compared with each other, the t-test for paired samples did not reveal statistically significant differences (Table 6).

Table 6.   Results of the comparison between the assessments of the MRI images by two masked investigators independently of each other.
 MeannStandard deviationp-value
Pair 1
 rFA4j0.55200.140.60
 rFA4m0.54200.11
Pair 2
 rMD4j0.0047200.00090.83
 rMD4m0.0046200.0010
Pair 3
 lFA4j0.58200.090.23
 lFA4m0.60200.09
Pair 4
 lMD4j0.0044200.00080.35
 lMD4m0.0043200.0006

Discussion

Examining patients with normal-pressure glaucoma and comparing them with healthy control subjects, we observed that the optic nerve diameter and the height of the optic chiasm as measured by MRI significantly correlated with the psychophysical parameter of visual field damage (rmax = 0.52) and with the structural parameter of RNFL thickness as measured by OCT (rmax = 0.74). The correlation coefficients were generally higher for optic nerve diameter measurements taken at the orbital apex than closer to the globe. It confirms a previous study by Lagreze et al. (2009), and recent investigation by Hernowo et al. (2011). It is complemented by a recent study by Engelhorn et al. (2011), who found a significant rarefaction of the optic radiation in patients with glaucoma. Several studies using histomorphometric techniques and enucleated globes as study material (Jonas et al. 1995) as well as in vivo non-invasive imaging techniques (Beatty et al. 1988; Dichtl & Jonas 1996; Boles et al. 2002;Kashiwagi et al. 2004) demonstrated that glaucomatous optic nerves were thinner than those of normal subjects. Correspondingly, Kashiwagi et al. (2004) reported that glaucoma affected the anterior visual pathway at least up to the optic chiasm and that these morphologic changes in the anterior visual pathway correlated with the amount of glaucomatous optic nerve damage. Parallel to the finding of significant associations between the dimensions of the optic nerve and of the optic chiasm with the thickness of the RNFL and the perimetric loss, we detected a significant reduction in the dimensions of the optic nerve and optic chiasm with increasing age in the control group. In agreement with a histomorphometrically shown age-related optic nerve fibre loss of about 0.3%/year (Jonas et al. 1990), and considering ageing as a non-glaucomatous type of optic nerve atrophy, it suggests that the observed changes in our glaucoma group versus the control group were not specific for glaucoma but that they generally reflected the loss of optic nerve fibres.

Diffusion tension imaging has become a widely applied imaging modality to study the connectivity and integrity of the white matter in CNS tissues (Le Bihan et al. 2001; Huppi & Dubois 2006; Mori & Zhang 2006). An increased DTI-MD value and decreased FA (DTI-FA) value have been reported to reflect axonal disruption (Wheeler-Kingshott et al. 2006). Hui’s study showed that radial diffusivity (λ) and FA of glaucomatous optic nerves increased and decreased, respectively, with time elapsed after the induction of glaucoma, whereas significant changes in the axial diffusivity (λ||) were not detected (Hui et al. 2007). Supported by the histological examinations of the optic nerve, such changes in the two DTI-derived parameters were attributed to a 10% decrease in the axon density of the glaucomatous optic nerve (Hui et al. 2007). Directional diffusivities (λ||, λ) have been shown to be specific to axonal injury and myelin damage in mouse models of optic nerve injury. On the basis of the successful application of directional diffusivities in mouse models of white matter injury, axial and radial diffusivities were potential biomarkers of axonal and myelin damage to be tested in human optic nerve disorders (Xu et al. 2008). The optic nerves of patients with glaucoma, as compared with control subjects, had significantly higher DTI-MD values and significantly lower DTI-FA values. Correspondingly, the mean DTI-MD values of the optic nerve were significantly and positively (= 0.81; p < 0.0001), and the mean DTI-FA values of the optic nerves were significantly and negatively (= 0.75; p < 0.0001) correlated with the amount of glaucomatous optic neuropathy (Garaci et al. 2009). In a similar manner, our study showed that the DT-MRI-derived FA for the optic nerve was significantly lower in the glaucoma group and that the DT-MRI-derived mean MD values, radial diffusivity (λ) and the axial diffusivity (λ||) of the optic nerve, were significantly higher in the glaucoma group. Correspondingly, the mean perimetric defect was significantly correlated with the DTI-FA values of the optic nerve (= 0.86; p < 0.001) and the DTI-MD values of the optic nerve (r = −0.77, p = 0.001).

The decreased diameter of the optic nerve in its orbital part in patients with glaucoma is of importance if the width of the orbital cerebrospinal fluid space is taken as surrogate for the orbital cerebrospinal fluid pressure (Kimberly & Noble 2008; Watanabe et al. 2008; Hansen et al. 2011), as the cerebrospinal fluid space width depends on the optic nerve diameter. This may potentially also be of importance for the discussion of the pathogenesis of normal-pressure glaucoma, in which besides other factors, a low orbital cerebrospinal fluid pressure has been considered to be involved (Jonas 2011; Orgül 2011; Ren et al. 2011).

As the diameter of the optic nerve and the height of the optic chiasm, the volume of the right and left LGN was significantly lower in the glaucoma group than in the control group in our study. It suggests that the atrophy of the LGN may be a biomarker of an atrophy of the visual system (Gupta et al. 2009).

Potential limitations of our study should be mentioned. Firstly, as for any hospital-based study, the question arises about the representativeness of the study population. Secondly, the sample sizes were relatively small and limited the conclusiveness of the statistical power of the study to draw conclusions. Thirdly, the choice of normal-pressure glaucoma compromised the generalizability and representativeness of the study population. Fourthly, the SITA fast strategy programme used in our study is less precise than the SITA standard programme to quantify the amount of glaucomatous visual field loss. Fifthly, last but not least, the definition of normal-pressure glaucoma did not include presence of visual field defects so that some patients classified as normal-tension glaucoma may in fact have been false positives with suspect optic nerve heads or with large but physiological optic cupping. In the glaucoma group, however, only 4 (13%) of the 30 patients showed no visual field defects. If these four subjects had falsely been diagnosed as glaucomatous, it would have reduced the difference between the study group and the glaucoma group. Despite of this potential limitation of the study, however, the differences between the glaucoma group and the control group were statistically significant, only serving to strengthen the results and conclusions of the study.

In conclusion, patients with normal-pressure glaucoma showed a reduction in the optic nerve diameter, height of the optic chiasm height and of volume of the LGN. These changes as measured by MRI were significantly correlated with RNFL thickness measurements by OCT and perimetric loss. In view of the relatively high correlation coefficients, it opens the possibility to examine the status of the optic nerve in patients with glaucoma with opaque optic media. The finding that the dimensions of the optic nerve, optic chiasm and LGN decreased with higher age suggested that the observed changes in the normal-pressure glaucoma group may not have been specific for glaucoma.

Ancillary