Advances in neuroimaging of the visual pathways and their use in glaucoma

Introduction

Recently developed neuroimaging techniques such as diffusion tensor (DT) magnetic resonance (MR) imaging and functional MR imaging (fMRI) can be used to evaluate the microstructural integrity of white-matter fibers and the functional activity of gray matter. Their introduction has allowed a reexploration of the normal anatomy of

Corresponding author. Tel.: +39 06 2090 2401; Fax: +39 06 2090 2404; E-mail: garaci@gmail.com

white-matter tracts in the living human brain and the elaboration of connectional models of brain function. During the last 10 years, these techniques have also been used for the in vivo study of a variety of brain pathologies. First used in the investigation of ischemic stroke, they are now becoming increasingly important in the evaluation of intracranial neoplasms, inflammatory disorders, developmental anomalies of the central nervous system (CNS), and neurodegenerative diseases.

Both techniques can also be utilized to investigate the integrity and functional connections of the visual pathways, including areas that play key

166

roles in glaucoma: preand postgeniculate whitematter fibers, the lateral geniculate nuclei (LGNs), and various areas of the visual cortex. This article provides an overview of some of the newer MRI techniques being used in neuro-ophthalmology and their specific applications in the study of patients with glaucoma.

Conventional MR imaging and the visual pathways

The macroscopic anatomy of the visual system can be demonstrated and assessed in vivo with conventional MR imaging. The optic nerve is a whitematter tract that is sheathed by the leptomeninges and dura mater, and its subarachnoid space is continuous with the intracranial subarachnoid space. On oblique coronal MR images, the intraorbital segment of the optic nerve ranges in thickness from 3.1 (anteriorly) to 2.5 mm (posteriorly), whereas the mean dural diameter measures between 5.1 (anteriorly) and 3.8 mm (posteriorly). The width of the subarachnoid space between the pial and dural sheaths ranges from 0.4 to 0.6 mm.

The optic nerve itself can be divided into intraocular, intraorbital, intracanalicular, and intracranial segments. The intraorbital tract, which is approximately 3 cm long, follows a sinuous course in both the horizontal and vertical planes. The intracanalicular portion is about 8 mm long. It passes through the optic canal (the channel at the apex of the orbital cavity) and the anterior end of the optic groove, which lies just below the lesser wing of the sphenoid bone. The ophthalmic artery runs through the same canal. The two optic nerves emerge from the optic foramina, ascend at an angle of approximately 451, and join to form the optic chiasm, which lies beneath the floor of the third ventricle, approximately 10 mm above the diaphragma sella.

The orbit is lined with adipose tissue, which is well organized and divided by fibrovascular septa. The contrast that makes the orbital structures stand out on an MR image is produced by this fat, which is associated with high signal intensity on T1-weighted images and high signal intensity on conventional T2-weighted images.

The anatomy of the optic nerve can be studied on an inversion recovery T1-weighted volume set,

which has a high signal-to-noise ratio, provides excellent contrast resolution, and can be reformatted in several planes. Use of this sequence facilitates differentiation of the optic nerve from the adipose tissue that surrounds it. The macroscopic anatomy can be better visualized on axial, coronal, and oblique–sagittal (along the long axis of the optic nerve) reconstructions with thin thickness and no gaps between the sections (Fig. 1). T2-weighted images can also be used to study the normal anatomy of the optic nerve (Fig. 2), but they are more useful for excluding the presence of lesions and/or atrophy. In the presence of optic nerve atrophy, the volume of the cerebrospinal fluid surrounding the nerve (which produces high-intensity signals on T2-weighted images) is increased. Again, T1-weighted images are helpful for visualizing the anatomy of the chiasm and for assessment of post-contrast enhancement on fat-suppressed images (Fig. 3).

The optic tracts are visualized using the same method. A three-dimensional T2-weighted sequence can also be used to image the structures in and around the optic chiasm, thanks to the natural contrast furnished by the abundant cerebrospinal fluid in the chiasmatic cistern (Fig. 4). The optic tract segments close to the LGN are difficult to visualize on morphological sequences owing to the partial volume effect produced by the intraaxial brain structures. T1-weighted inversion recovery images are also used to identify the LGNs based on the contrast between the MR signals generated by gray and white matter. Postgeniculate fibers are not visualized with conventional morphological imaging because these sequences can only distinguish brain structures characterized by ‘‘natural contrast’’ (e.g., white vs. gray matter, nerves vs. CSF) (Jacobs and Galetta, 2007).

For distinguishing the white-matter fibers and mapping the visual gray matter areas (LGN and cortex), the functional neuroimaging modalities discussed below (diffusion-weighted MR and fMRI) are valuable tools.

Diffusion MR imaging

Diffusion is a random process that results from the thermal translational motion of molecules. In

Fig. 1. Conventional axial (A) and coronal (B) T1-weighted images demonstrating the macroscopic anatomy of the intraorbital portion of the optic nerve. Note the hyperintense signal of intraorbital adipose tissue lining the nerves and the hypointense signal from the extraocular muscles.

biological tissues, water diffuses inside, outside, around, and through cellular structures. Cellular membranes hinder these movements and force the water to take more tortuous paths. Diffusion is restricted by the presence of cellular swelling or increased cellularity, whereas necrosis, which involves the breakdown of cellular membranes,

167

Fig. 2. Conventional morphological axial (A) and coronal (B) T2-weighted images obtained by using a 3 T magnet showing the intraorbital segment of optic nerves.

decreases diffusion-path tortuosity and increases apparent diffusivity (Le Bihan, 1995). Studies of water diffusion can thus provide information on cellular integrity and pathology. In conventional MR imaging, the effect of diffusion contributes very little to the overall signal intensity. As the diffusing protons move through intrinsic and extrinsic magnetic field gradients, they experience phase shifts that result in a loss of transverse

168

Fig. 3. Sagittal T1-weighted image (A) showing the chiasmatic region and the optic tract (arrow); (B) Post-contrast axial T1weighted image with fat-suppression technique, note the normal enhancement of the extraocular muscles (long arrow) and the absence of enhancement of the optic nerve (thick arrow).

magnetization. This phenomenon is exploited to create diffusion maps, which represent the spatial distribution of diffusion coefficients for the imaged tissue. An early study on the diffusivity of whitematter fibers in normal subjects showed that the apparent diffusion coefficients (ADC) are dependent on the orientation of the diffusion gradients

with respect to the tracts being examined (Hajnal et al., 1991). Fiber tracts lying parallel to an applied gradient have the highest ADC, and the lowest coefficients are observed with tracts that lie transverse or oblique to a sensitizing gradient. This observation highlighted the importance of directionality in the assessment of diffusing molecules. The diffusion of water molecules (especially in white matter) does not proceed equally in all directions: it is anisotropic, i.e., characterized by the predominance of movement in one direction (Basser, 1995).

Diffusion-weighted MR imaging has found numerous applications in the field of neuroscience, including the evaluation of stroke, brain development, tumors, and demyelinating disorders. The most common approach is to obtain an echoplanar sequence with supplementary diffusion-sensitive gradients in at least three main axes in space (xx, yy, zz). Two strong gradients of opposite polarity are applied with a short interval in between. All stationary spins are equally dephased and rephased by the action of these gradients. Spins moving between the two gradients are subjected to the effects of a different field at the time of the second application. Their phase recovery is thus incomplete, and the result is signal attenuation. The sensitivity of the sequence to microscopic motion such as the diffusion of water is related to the strength and the duration of the applied gradients, which are collectively expressed by the b-value. In general, a b-value of about 800–1000 s/mm2 is used. Acquisition of a diffusion-weighted image is always accompanied by the acquisition of a reference image with a b-value of 0 (no applied diffusion gradients). The latter consists of an echoplanar image weighted in T2 (Fig. 5A).

The final signal produced by the water molecules depends on their ability to diffuse within a tissue: increased diffusion is reflected by a reduction in the intensity of the tissue signal. This phenomenon can be exploited to quantify the diffusivity of water molecules in different directions. The prevalence of diffusion in one direction, for example, along the course of white-matter fibers — diffusion anisotropy — can be quantified by calculation of a second-order symmetric tensor of six elements (or diffusivities) (Moseley et al., 1990; Conturo et al.,

1999). The information derived from this DT can be used to elaborate quantitative maps (Fig. 5B–D) showing mean diffusivity (MD) and fractional anisotropy (FA), which are useful for comparing individuals or populations. Anisotropy data can also be used to deduce the pathways of major nervefiber tracts, a technique known as tractography.

Diffusion tensor imaging (DTI) has been used to assess axonal degeneration within the visual pathways. Trip et al. (2006) found that MD is significantly increased and FA is significantly reduced in optic nerves affected by optic neuritis (compared with unaffected contralateral nerves and with those of healthy controls). These findings, which appeared compatible with the presence of axonal disruption, were similar to

those reported in studies of chronic brain lesions in multiple sclerosis (Werring et al., 1999; Filippi and Inglese, 2001; Dong et al., 2004). Correlation between clinical and electrophysiological parameters and diffusion-weighted MR data has been observed in optic nerves with chronic postinflammatory lesions (Hickman et al., 2005).

Postmortem histological studies have demonstrated that a lesion located between the retina and lateral geniculate body can cause full-length degeneration of the retinal ganglion cell (RGC) axons that develops in both the anterograde and retrograde directions (Hoyt and Luis, 1962). Ueki et al. (2006) used a PROPELLER (periodically rotated overlapping parallel lines with enhanced reconstruction) sequence (Pipe et al., 2002) to

obtain diffusion-weighted images of the visual pathway and found that increased diffusivity was the main alteration associated with RGC degeneration. Recently, Wu et al. (2007) demonstrated that diffusion changes are closely correlated with the total axolemmal cross-sectional area of the prechiasmatic segment of the murine optic nerve. These quantitative histological data confirm that optic nerve diffusivity is a reliable quantitative index of ultrastrucural axonal degeneration.

Irreversible degeneration of the optic nerve axons has been demonstrated in glaucoma (Nickells, 1996; Artes and Chauhan, 2005). DTI of the optic nerve in rats has revealed that FA decreases and MD increases with time after the experimental induction of glaucoma (Hui et al., 2007). The changes observed on DTI in this study

were associated with histological evidence of reductions in the number of RGC axons in the glaucomatous optic nerve. (The loss of these axons results in an expansion of the extracellular space, which enhances diffusion.) In a study being conducted by our group, DTI changes are also being found in the optic nerves of patients with different degrees of glaucoma, and these changes seem to be significantly correlated with those of the Humphrey field analysis. These data suggest that MR imaging could be a reliable, noninvasive tool for monitoring the progression of human glaucoma based on quantitative assessments of axonal loss.

DT tractography is a more recently developed technique, which can disclose the complex arrangement of fiber tracts and provide

fundamental information on cerebral connectivity (Fig. 6). It requires calculation of various parameters of the DT ellipsoid, a graphical representation of the DT within each voxel that highlights the three-dimensional character of diffusion directionality. One of the most important is the direction of the largest eigenvector (l1), i.e., the direction of greatest diffusivity, which is generally assumed to coincide with that of the fiber bundles (Mori et al., 1999; Masutani et al., 2003). In DT ellipsoid-based fiber tractography, the path of a reconstructed fiber is determined by the l1 in each voxel. Tractography usually models the water diffusion characteristics within each voxel using the DT and assumes that the tensor field within and across voxels reflects the underlying axonal architecture (Basser and Pierpaoli, 1996). In this manner, white matter can be parcellated into fiber

171

structures containing similarly oriented axonal fascicles (Mori and van Zijl, 2002).

DT fiber tractography can be used to visualize the longitudinal direction of the optic nerve fibers (Figs. 6 and 7), to depict the optic radiation (also known as the geniculocalcarine tract) (Yamamoto et al., 2007), and to assess the axonal density of postgeniculate fibers (Figs. 8 and 9).

The optic radiation consists of a broad, thin layer of fiber tracts that extend from the LGN to the primary visual cortex (which corresponds to Brodmann area 17) and are organized into anterior, central, and posterior bundles. The anterior bundle runs in an anterolateral direction and then sweeps forward and courses along the inferior horn of the lateral ventricle in the temporal lobe, thus forming the Meyer loop. It then runs posteriorly, along the lateral wall of the

172

ventricle, and projects into the lower lip of the calcarine fissure. The central bundle runs in a lateral direction and then curves posteriorly, continuing along the lateral wall of the ventricle and converging on the calcarine cortex. The

Fig. 7. Visual pathways tractography with superimposed morphological T2-weighted image at the level of the chiasm.

posterior bundle runs posteriorly along the roof of the lateral ventricle to the upper lip of the calcarine fissure. There are numerous anatomical variations in the course of the optic radiation, especially in the Meyer loop (Yamamoto et al., 2005). Knowledge of the exact course of the optic radiation in individual patients is important for predicting the outcome of visual field defects caused by temporal lobe damage and for planning neurosurgical procedures that will minimize injury to this critical tract.

Using the fiber-tracking method, Yamamoto et al. (2005) clearly depicted the three bundles of the optic radiation, from their origins in the LGN all the way to the calcarine fissure. Their findings show good agreement with those of classic anatomic studies of visual pathway topography, especially those regarding the courses of the central and posterior bundles (Ebeling and Reulen, 1988). On T2-weighted coronal sections passing through the trigonum, the optic radiation lay 2–3 mm lateral to the wall of the ventricle, which is consistent with previous reports indicating that the