Review: Magnetic resonance imaging techniques in ophthalmology

Magnetic resonance imaging techniques used to acquire high spatiotemporal resolution eye images

Surface detector coils can be used to achieve high spatial and/or temporal resolution of MRI images in general, including eye images. The best ocular imaging scenario uses a head volume coil in the transmission mode and an eye-shaped single or double loop surface coil in the reception mode [38]. In this way, uniformity of the excitation RF field is achieved with the head volume coil, while the eye-shaped detector coil is used to maximize the SNR in the images. If uniformity of the excitation RF field is sufficient over the eye volume, an eye-shaped transceiver surface coil (that is, a coil that can be used in both transmit and receive modes) may be used to acquire images with even higher SNRs. This increase in SNR can be used to improve the spatial and/or temporal resolution of the images [39]. Eye structures, such as the retina, iris, ciliary body, lens, and aqueous and vitreous humors, of a single human [14,26-28,40], mouse [41,42], and rat [7] eye or of both human [43] and rat [38] eyes can be visualized in vivo using surface coils.

Anatomic information of normal and diseased human eye visualized in vivo by T₂-weighted MRI at 1 T is shown in Figure 1 [33]. The corresponding quantified T₁ maps in this figure reveal a greater than twofold decrease in T₁ values in the vitreous and aqueous humors of the eye with a cataract. Figure 2A,B show the anatomic information of normal and diseased rat eye visualized in vivo by T₂-weighted MRI at 4.7 T [38]. The images were acquired in 30 min at a high spatial resolution of 60×60×700 μm³ and clearly show structures such as the lens, cornea, iris, ciliary body, retina, and sclera. A large human retinal detachment was visualized by Deans et al. [43] using a surface coil in the receive mode (Figure 2C). Retinal abnormalities were also assessed by measuring the thicknesses of the retina on MRI images of the human eye at 1 T [14] and at 3 T [31].

More recently, variations in eye volume, surface area, and shape, with the refractive error in the mouse at 4.7 T [44,45] and human at 3 T [46], were quantitatively evaluated. The authors showed that both the mouse and human vitreous humor are elongated in myopia and that both mouse and human eye shapes change even in the early stages of myopia. The three-dimensional enlargement of normal eyes is uniform, while that of myopic eyes takes place along the axial globe direction only, determining the prolate shape of the myopic eyes [46].

Visualization of human eye structures in vivo requires high spatiotemporal resolution [26,32]. The average diameter of the human eyeball is approximately 25 mm, while that of the rat eye is approximately 5 mm [47]. The eye contains thin eye structures; for example, the maximum retinal thicknesses of normal rat [38], mouse [42], and human [14] eyes measured on ocular MRI images acquired in vivo ranges between approximately 200 μm [38,42] and 1,200 μm [14].

A high spatial resolution clinical ocular MRI study allowed for the measurement of the mean thickness of the human retina/choroid complex, which was found to be 711 μm, and for the identification of three layers in the retina [36]. A recent high spatial resolution ex vivo animal ocular MRI study allowed for the visualization of several layers in the region of the retina/choroid complex, albeit with poor contrast [48].

The highest reported in-plane spatial resolution for in vivo ocular multislice MRI images is 23×23×620 μm³ for animal studies [7] and 100×200×2,000 μm³ for human studies [31], with acquisition times of 50 min and 8 s, respectively. Representative images from these studies are presented in Figure 3. For three-dimensional acquisitions, on the other hand, spatial resolutions as high as 42×42×84 μm³ and 20×20×57 μm³ were recently achieved in vivo in ocular MRI animal research studies [48,49].

The reduced MRI signal intensity collected from the resultant smaller voxel volumes in these high-resolution images invariably leads to a decrease in the MRI image quality. The MRI image quality may be improved to some degree via signal averaging with the consequent increase in acquisition time. A compromise between acquisition time and image quality always has to be achieved in ocular MRI, given the propensity for the eye to move over short timescales and the small structures present that need to be visualized. One can thus place a realistic upper limit of approximately 2 min on the acquisition time for human ocular MRI images.

The resolution (spatial and/or temporal) can also be improved using specially designed surface coils that can be placed over both eyes, allowing for the simultaneous acquisition of images of both eyes. Figure-eight surface receive-only coil [38] was designed and constructed for just this purpose. The resultant images are presented in Figure 2A,B and Figure 4, where retinal thicknesses as small as 100 μm were measured at 4.7 T [38]. In this study, retinal abnormalities in a rat model of experimentally induced autoimmune uveitis (EAU) were detected in T₂-weighted images, with findings confirmed ex vivo by histology.

A similar image acquisition setup was used by Deans et al. [43] to acquire rat eye images, which also benefited from the reduced time to acquire images of both eyes.

Histologically ex vivo-confirmed sites of infiltrated macrophages were also detected in vivo by MRI in the region of the retina of rat eyes at the peak of EAU [39]. The MRI images in this macrophage-specific study were acquired using a two-turn single-loop receiver coil placed over one eye of an anesthetized rat, where the second loop was designed to increase the SNR over a single-loop coil [39].

Blood-oxygenation, level-dependent, functional magnetic resonance imaging techniques

Retinal function and ocular physiology can also be evaluated using fMRI techniques [11]. fMRI can be performed with or without contrast agents and is most commonly used to evaluate brain processes ranging from sensory perception to cognitive functions. The most commonly used fMRI technique is based on blood-oxygenation level-dependent (BOLD) contrast. The BOLD contrast is determined by the intravoxel magnetic field inhomogeneities produced by the paramagnetic deoxyhemoglobin in the erythrocytes in the blood. A regional reduction in deoxyhemoglobin concentration is produced due to an increased neural activity when a specific task is performed, for example, when a visual stimulus is presented to the subject. These changes can be visualized and/or quantified on T₂ and T₂*-weighted MRI images [11]. Duong et al. [11] acquired BOLD–fMRI eye images during dark, drifting grating, and stationary grating stimuli to evaluate retinal function and physiology in the cat eye. Three different segments in the retina and differential responses in the two vascularized retinal regions were detected [11]. In a separate study, spatial-independent component analysis was applied to the data to quantify the human eye movements and to estimate the point of gaze at the time of data acquisition [50]. Information on the retinal/choroidal blood flow of normal, unanesthetized, and awake humans was also achieved at 3 T using BOLD–fMRI techniques [51].

Diffusion-weighted and diffusion-tensor magnetic resonance imaging techniques

Diffusion-weighted imaging (DWI) and diffusion-tensor imaging (DTI) can quantitatively assess the isotropic (DWI) and anisotropic (DTI) water diffusion in tissues in vivo [8,9]. DWI uses additional magnetic field gradients to encode diffusion of ¹H nuclei in water molecules in the tissue investigated. Several MRI images are acquired using T₂-weighted pulse sequences and diffusion-encoding gradients with at least two different amplitudes. Although it is possible to apply the diffusion-encoding gradients along one direction alone, in general they are applied along three orthogonal directions to sensitize the data to diffusion properties in all directions, with an average across all directions calculated. The resulting data are mathematically modeled using a mono- or bi-exponential model of the diffusion of water molecules in tissue, from which quantitative measurements of the apparent diffusion coefficient may be calculated. Fractional volumes of fast and slow diffusing water components may also be calculated, which are thought to correspond to intracellular and extracellular water diffusion. Assessment of the bi-exponential diffusion component is difficult to perform clinically because it requires acquisition of diffusion-weighted images with high SNR [52]. In DTI, the diffusion-encoding gradients have to be applied along at least six, although often more, different directions, and hence acquisition times tend to be much longer than that of a DWI experiment [52]. The quantitative information extracted from DTI images gives more complex information on the anisotropy of water diffusion, which can be evaluated quantitatively in vivo by the eigenvalues of the diffusion tensor, apparent diffusion coefficients, and fractional anisotropy calculated after mathematically modeling the data acquired in a DTI experiment [52].

Using DWI techniques, apparent diffusion coefficient values in normal cat, rat, and mouse retina were calculated at 4.7 T, 7 T, and 11.7 T [53] and the restricted diffusion in an inflammatory optic neuropathy in a 4-year-old girl with acute visual loss in the right eye [54] was detected. Mono-exponential water diffusion in the inner, middle, and outer normal rat retinal layers were also quantitatively assessed by DWI [34].

DTI techniques allowed detection of three retinal layers and one choroidal layer in normal mouse retina [41]. The internal diffusive pathways of the lens ex vivo in bovine [55] and human [56] and in vivo in mice [41] eyes were also determined using DTI techniques. In these studies, elongated fiber cells extending from the anterior to the posterior suture were detected in the lens cortex of the eyes of each species. Individual lens fiber cells were organized into concentric layers parallel to the lens surface [41,56].

The retinal cell alignment and water diffusion in mice were also assessed quantitatively by DTI, where layer-specific apparent diffusion coefficients and fractional anisotropy were detected in the mice retina [41]. The photoreceptor cells in the central retina, adjacent to the optic nerve, were found to exhibit a well organized spatial distribution that is parallel to the optic nerve axis.

Ocular relaxometry

The clinical potential of the NMR technique was demonstrated in the 1970s by Raymond Damadian who showed that T₁ and T₂ relaxation times of tumors were significantly different compared to those of normal tissue [19]. Recent studies have indicated that quantitative MRI relaxometry could help to optimize anatomic, physiologic, and functional contrast in vivo [34]. During the past 40 years, MRI techniques have developed rapidly, and the T₁, T₂, and T₂* relaxation times of the cat [57], rat [34], and mouse [42] retina and vitreous humor were calculated in vivo at 4.7 T, 7 T, and 11.74 T using MRI. More recently, T₁ [28,32,33] and T₂ [32] maps of human eyes were extracted from ocular MRI images. T₁ maps of a normal and of a disease-affected human eye are shown in Figure 1 [33]. Mean T₁ and T₂ values in the region of the aqueous humor, lens, vitreous humor, retina, and sclera in eyes of normal controls and patients at different stages of diabetic retinopathy were also calculated from data acquired at 1 T, with statistically significant differences found between the T₁ values measured in the region of the lens and aqueous humor [32].

These preliminary human studies have demonstrated the potential of MRI relaxometry for clinical ocular MRI, and it is clear that future studies should focus on the development of such techniques at higher magnetic field strengths, while clinical studies need to be performed on more refined disease stages. The use of more quantitative data in clinical imaging in general, and MRI in particular, would be helpful for disease identification, staging, and for elucidating ocular disease mechanisms.

Quantification of ocular perfusion using arterial spin-labeling techniques

Tissue perfusion can be quantitatively assessed in vivo using the technique of arterial spin labeling (ASL). ASL techniques involve the rapid acquisition of two MRI images, one of the images magnetically labeling the blood flow in the tissue of interest. Typically, adiabatic RF pulses are used to magnetically label the flowing spins (i.e., the ¹H nuclei in water molecules in the blood vessels) to ensure uniform labeling. Separation of the arterial, capillary, and venous blood flow is possible if appropriate velocity-encoding magnetic field gradients are applied to the tissue of interest. Labeling of the blood flow can be achieved using continuous or pulsed ASL techniques; the optimal technique remains the focus of much research activity. If the tissue–blood partition coefficient and the spin-labeling efficiency are known, the labeled and unlabeled MRI images and a T₁ map can be used to calculate blood flow maps of the tissue of interest. Each pixel of these maps will indicate the value of the blood flow expressed in units of ml/g/min [9].

Blood flow maps of the rat retina were obtained at 7 T using a continuous ASL technique [58]. The authors assumed that the tissue–blood partition coefficient in the retina was 0.9, while the mean T₁ value of the retina/choroid complex was found to be 1.7 s and the spin-labeling efficiency was 0.8 for the ASL technique used.

The measured blood flow in the rat retina/choroid complex and the rat ciliary body were high, whereas that in the rat cornea and rat vitreous humor were essentially absent or within noise levels. Blood flow of the normal rat retina/choroid complex (4–6 ml/g/min) was markedly higher than cerebral blood flow (0.9–1.1 ml/g/min) under essentially identical experimental conditions, including 1.1% isoflurane anesthesia. The blood flow in the rat retina/choroid complex with spontaneous retinal degeneration was markedly reduced compared to the corresponding control values. Blood flow is tightly coupled to basal metabolic activity, so the reduced metabolism of degenerated retinas of rats is expected to lead to reduced blood flow. The reduction of blood flow in the rat retina/choroid complex is in agreement with results in studies on neurodegenerative brain diseases. The reduction of the rat blood flow in the region of the optic nerve head did not confirm the theory of high vascular density in the region of the rat optic nerve head [58]. Blood flow along the retina/choroid complex was relatively uniform (as expected, since rats do not possess a fovea), contradicting the theory that the optic nerve head is densely populated by large arteries and veins in the rat [58].

It was not possible to separate the rat choroidal and retinal blood flow in vivo using the ASL MRI technique described by Li et al. [58]. Other techniques allowed separation of the rat choroidal and retinal blood flow measurement but they involved radioactively labeled microspheres. Based on results obtained using microsphere techniques, the rat choroidal blood flow was about ten times more rapid than either rat cerebral or retinal blood flow [58].

More recently, an improved ASL technique allowed for the separation of the choroidal and retinal blood flow in mice [59,60] and in rats [61]. The choroidal blood flow in the normal mice was found to be five to six times more rapid [59], while in the normal rats this was eight times [61] more rapid than the retinal blood flow. The retinal and the choroidal blood flow decreased in the mice affected by glaucoma [60] and increased in the rats with retinal degeneration [61] compared to the corresponding blood flow of normal mice and rats. In the rats with retinal degeneration, the difference between the choroidal and the retinal blood flow decreased with approximately 25% compared to the difference detected in the normal rats [61], the choroidal blood flow remaining more rapid relative to the retinal blood flow.

Assessment of the blood flow of the human retina/choroid complex is also feasible using ASL techniques. Peng et al. [62] reported for the first time a blood flow of 93 ml/(100 ml min) of normal, unanesthetized, and awake humans. Zhang et al. [63] showed recently that the human retinal/choroidal blood flow increases by 25% during brief handgrip exercises.

The ASL technique can also be set up to be more sensitive to smaller vessels, such as arterioles, capillaries, and venules, if the imaging parameters are adjusted to minimize contributions from large vessels. For example, the inclusion of a delay (for example, 200 ms) between spin labeling and image acquisition allows time for the labeled spins to leave the large arteries and move into the smaller vessels, thereby decreasing the sensitivity of the large arteries. Second, the labeled spins lose their degree of magnetic labeling by the time they reach large draining veins (since the T₁ of blood is approximately 2 s), thereby decreasing the impact of large veins. Selective labeling of the flowing spins can also be achieved based on their velocity using additionally applied, velocity-encoding, magnetic field gradients. Selectively detecting blood flow in smaller vessels is advantageous since it more accurately reflects local tissue perfusion. It is also interesting to note that, in contrast to most optically based approaches, ASL measures tissue perfusion of labeled water in the whole tissue within a voxel without the need to resolve individual vessels [9,58]. The advantages of the ocular ASL techniques over the optically based ocular imaging techniques are that they have the potential to provide detailed quantitative information on the blood flow and volume over the whole volume of an eye or both eyes simultaneously, with acquisition times on the order of minutes, and do not require transparent media for the propagation of the electromagnetic waves through the eye during image analysis. Arterial, venous, or capillary blood flow can be separated and quantitatively analyzed using ASL techniques. More complex quantitative information can also be extracted from the ASL data using compartmental modeling [64]. However, high spatiotemporal resolution is required to separate measurements of the retinal and choroidal blood flow, while the achievable SNR in in vivo ASL studies tends to be low, rendering it difficult to mathematically model the data to extract accurate quantitative information on the perfusion. In animal studies, the data have been found to be influenced by the type of anesthesia performed, although this will pose less of a problem in clinical studies (of course, problems with eye movement will then arise). Nevertheless, ASL techniques in general, and ocular ASL techniques in particular, need to be improved to boost the achievable SNR before they will become clinically available [9,58].

Intravascular contrast agents

Extracellular contrast agents based on gadolinium (Gd), manganese (Mn), and superparamagnetic iron oxide (SPIO) nanoparticles have been used in several ocular MRI studies. Gd-based agents are commonly used in clinical MRI to enhance image contrast to improve visualization of anatomic information and also for the evaluation of the diffusional pathways of plasma-derived solutes through the blood-to-ocular barriers [72-74] or in dynamic contrast-enhanced MRI studies [65].

Only one Mn-based agent has been approved for human use, and thus most of the published literature is restricted to animal imaging studies [7,75-83]. It is clear that further investigations need to be performed on these agents before widespread clinical acceptance will be achieved.

Extravascular contrast agents for ocular magnetic resonance imaging

One of the earliest applications of nanotechnology in MRI involved the use of iron-oxide particles. Iron-oxide crystals have long been used as superparamagnetic T₂* contrast agents for MRI. SPIO (particle diameter >50 nm)-based contrast agents for MRI are intravascular, while ultrasmall SPIO (USPIO, particle diameter <50 nm)-based contrast agents for MRI are extravascular. For in vivo applications, the particles have nonstoichiometric microcrystalline magnetite core(s) and are typically coated with dextran (e.g., ferumoxide) or siloxane (e.g., ferumoxsil) [21]. Chromium (Cr)-based contrast agents for MRI represent another type of extravascular contrast agent [90].

While at low doses, circulating iron oxides can decrease the T₁ time of blood and are used in MRI angiography, at the usual doses used for molecular imaging, the T₂* effects predominate, resulting in marked signal loss on T₂*-weighted images. Unfortunately, for molecular imaging applications, persistent T₂* effects from circulating iron oxide nanoparticles make it necessary to delay the MRI examination by 24 h to 72 h postinjection, complicating clinical implementation of these techniques. Iron oxides concentrated at a target site generate magnetic susceptibility artifacts that can extend over the region infiltrated by them, appearing as areas with decreased signal intensity on T₂*-weighted images [21].

Cr-based contrast agents have also been developed and used as extravascular agents in ocular MRI studies. Cr, specifically hexavalent chromium [Cr(VI)], in the form of potassium or sodium dichromate, is thought to alter heterogeneous tissue oxidation by diamagnetic dichromate [Cr(VI)], which creates paramagnetic Cr species [Cr(V) and Cr(III)] that shorten the T₁ of surrounding water and/or lipid protons. Structures in the tissue infiltrated by the Cr-based contrast agent will, therefore, appear bright on T₁-weighted MRI images compared with tissue regions without the Cr-based contrast agent. The retina has a unique lipid profile with the highest level of long-chain polyunsaturated fatty acids in the body and is unmyelinated in most mammalian species but has an uneven distribution of oxidizable lipid contents across layers. The cone and rod outer segments in the outer retina in rodents have the highest lipid density among retinal layers and have many structural similarities as in myelin. Cr-based contrast agents have, therefore, high potential in differentiating retinal layers based on the Cr-based contrast agent uptake and on the T₁ alteration produced by the contrast agent in the retinal layer [90].

Both USPIO- and Cr-based contrast agents are toxic and thus can be used only in animal research studies [90]. Further developments of these contrast agents are needed before such contrast agents for MRI can be used clinically.

Ocular ¹H functional magnetic resonance imaging using gaseous contrast agents

Rat ocular blood flow has been quantitatively evaluated by fMRI in several studies during hyperoxic (100% O₂) and hypercapnic (mixture of O₂ and CO₂ in different concentrations) conditions, with the air breathing condition used as a baseline in each case. The blood flow calculated in the avascular rat eye structures was extremely low compared to that calculated in the vascular rat eye structures. Hyperoxia was found to decrease the blood flow due to vasoconstriction, whereas hypercapnia increased blood flow due to vasodilation. Rat blood flow in the retina was significantly diminished in degenerated retina and robust hypercapnia- and hyperoxia- induced blood flow changes were observed in normal retinas. Hyperoxia/hypercapnic breathing reduced/increased the total blood flow in the retina of normal animals by 12% and 14%, respectively. Basal blood flow in disease-affected rat retinas was markedly reduced compared to those of control rat retinas. Blood flow is tightly coupled to basal metabolic activity and the reduced metabolism of degenerated retinas of diseased rats may have caused the reduction of retinal blood flow. Hypercapnia- and hyperoxia-induced absolute blood flow changes were not statistically different between normal and diseased rat retinas [58].

Two distinct laminar signals corresponding to the retinal and choroidal vascular layers bounding the retina, separated by the avascular layer in between, were detected based on blood volume measurements. The choroidal blood volume was about ten times larger than the retinal blood volume. fMRI based on blood volume measurements also detected vascular layer-specific responses to physiologic challenges and revealed that choroidal blood vessels were less responsive to physiologic challenges than retinal vessels, indicative of differential hemodynamic regulation of the two vasculatures. In an accepted animal model of photoreceptor degeneration, MRI confirmed the disappearance of the outer nuclear layer and photoreceptor segments. The baseline blood volume values in the retinal and choroidal vasculatures were elevated compared to age-matched controls [67].

The literature on hypercapnic responses and on basal blood flow in the retina is inconsistent, placing these animal studies some distance from likely clinical implementation [67-69]. Despite these inconsistencies, a more recent study on human retinal/choroidal blood flow under hypercapnic conditions (5% CO₂ and 95% O₂ inhalation) showed a 12% increase of retinal/choroidal blood flow relative to air inhalation conditions [62].

Fluorine magnetic resonance imaging

¹⁹F MR spectroscopy experiments have been performed in the eye [91] and in aqueous humor assay [92]. However, ¹⁹F techniques are not sensitive enough for imaging in vivo due to the low signal from the significantly lower concentration of ¹⁹F nuclei in the body compared to ¹H; hence, only localized ocular ¹⁹F NMR spectroscopy of large voxels can be used to detect the presence of ¹⁹F nuclei. In addition, broadband capabilities, which are necessary to detect NMR signals at frequencies other than that for ¹H, are not common features on most clinical MRI systems. Even if a clinical scanner is equipped with a broadband RF channel for nonproton MRI/magnetic resonance spectroscopy, some hardware development is generally required, such as developing a suitable RF coil [70].

Ocular magnetic resonance imaging using magnetization transfer contrast

Clinical MRI images mainly ¹H nuclei in water molecules. The ¹H NMR signals can be emitted by mobile ¹H nuclei in water molecules, by less mobile nuclei in lipids, or by rigid ¹H nuclei in the tissue macromolecules or membranes. The signal emitted by the rigid nuclei is not easily observed directly because of its large bandwidth relative to the free water ¹H nuclei. However, ¹H nuclei in the rigid pool can be selectively labeled by applying an off-resonance RF pulse. If the ¹H nuclei in the mobile and rigid pools are interacting through magnetization exchange (dipolar and/or chemical exchange), then the magnetic labeling of ¹H nuclei in the rigid pool (macromolecules, membranes) can be transferred to the ¹H nuclei in the mobile pool (water). Effective coupling between the ¹H nuclei in the two pools is visualized on the MRI image by a decrease in signal intensity [15]. It has previously been shown that this effect is dependent on the concentration, mobility, macromolecules, or membranes and remodels macromolecular structures within the tissues [93]. The magnetization exchange between the ¹H nuclei in the two pools can be quantified by calculating magnetic transfer ratios on the acquired MRI images.

Lens opacifications must be accompanied by alterations in tissue compositions, including the hydration state. MTC-enhanced MRI has been successfully used to document lens changes in longitudinal studies of galactosemic dogs during sugar cataract formation [94]. These MTC-weighted images were consistent with localized osmotic lens changes during cataract formation and consistent with the osmotic hypothesis of sugar cataract formation. MTC-weighted images were able to show lens changes sooner than standard MRI images. Measurements of the signal intensity on the MTC-enhanced MRI images revealed that significant tissue changes occurred before any clinically visible lens changes [93]. These measurements also showed that the addition of an MTC preparation pulse to a standard MRI sequence yields high contrast in the dog [94] and human [93] lens. Cortical lens changes were better detected with unenhanced MRI, whereas increased sensitivity to nuclear changes in the lens was observed with MTC enhancement.

MTC-enhanced MRI was also used to quantitatively evaluate the changes produced in the optic nerve in patients with optic neuritis. The quantitative evaluation of the optic nerve using MTC-enhanced MRI may also be useful in the investigation of multiple sclerosis. Magnetic transfer ratio values measured on MRI images in the region of the optic nerve of patients with optic neuritis were significantly different than that measured in corresponding regions of normal eyes [95]. The pathophysiological basis of MTC has not yet been elucidated, and consequently more clinically relevant conclusions cannot be drawn from these studies [95].

Magnetic resonance elastography of the eye

The mechanical properties of the eye and its anatomic components are believed to be of central importance in many pathologic conditions, such as age-related macular degeneration, glaucoma, Graves’ disease, and myopia. The accurate quantification of these mechanical properties can thus be used to improve the basic understanding of eye mechanics and to enhance patient care, diagnosis, and management of disease [96].

Historically, the mechanical properties of the eye have been a challenge to assess and have proved difficult to quantify, especially noninvasively. Ocular elasticity has been called one of the most confused areas of ophthalmology. The eye comprises multiple tissue types and various anatomic structures that exhibit characteristics of heterogeneity, nonlinearity, anisotropy, viscoelasticity, physiologic accommodation, and extreme sensitivity to hydration—mechanical characteristics that are difficult to account for with a single measurement technique [96].

Conventional mechanical testing methods, including axial strip testing and dynamic mechanical analysis, have been applied to various tissues of the eye, although these techniques are generally limited in their accuracy because testing requires dissection and flattening of doubly curved tissue samples. In addition, these methods are restricted to research applications due to their invasive nature [96].

More recently, experimental image-based approaches for studying eye mechanics have been employed with some limited success by directly measuring displacements (strain) in the eye under inflation conditions. These approaches include techniques such as two-point spot scanning, holographic interferometry, optical imaging of treated corneal surfaces, wave speed analysis using ultrasound, and MRE. Of these techniques, ultrasonography and MRE are the only viable techniques for in vivo applications [96]. MRE is a rapidly developing technology for quantitatively assessing the mechanical properties of the tissue. The technology can be considered to be an imaging-based counterpart to palpation, commonly used by physicians to diagnose and characterize diseases. The success of palpation as a diagnostic method is based on the fact that the mechanical properties of tissues are often dramatically affected by the presence of disease processes, such as cancer, inflammation, and fibrosis. MRE obtains information about the stiffness of tissue by assessing the propagation of mechanical waves through the tissue with a special MRI technique. The technique essentially involves three steps: generating 50–500 Hz shear waves in the tissue using an external driver, acquiring MRI images depicting the propagation of the induced shear waves, and processing the images of the shear waves to generate quantitative maps of tissue stiffness called elastograms [97].

The acoustic shear waves in the tissue are generated by external driver devices. The electrical signal for these devices is created by a signal generator triggered by and synchronized to the MRI pulse sequence and is amplified by an audio amplifier before being fed into the mechanical driver. Over the years, several driving mechanisms have been developed, each with their own advantages and limitations. Spin-echo, gradient-recalled echo, steady-state free precession, and echo planar imaging-based pulse sequences can be designed to be sensitive to motion with a specific frequency, a particular multiple of the specific frequency, or a broad range of frequencies [97]. The resulting tissue motion is encoded into the phase of the MRI images with a synchronous gradient field applied at the motion frequency [96]. Preprocessing directional filtering techniques can be used to minimize shear wave interference artifacts that can affect the stiffness calculations. Mathematical algorithms are then used to calculate the mechanical properties of the tissue from the phases of the images acquired. The frequency domain equation of motion for a general, homogeneous, anisotropic, viscoelastic material relates an applied mechanical stress to the resultant strain and can be expressed as a rank-4 tensor with 21 independent complex quantities. Simplifying assumptions, like isotropy and incompressibility, allow for the calculation of mechanical properties (for example, shear modulus) to be used for clinical interpretations. Making the assumption of isotropy reduces the number of independent quantities to two; these are typically the two Lamé constants that predominantly control the longitudinal and shear strains [97]. Depending on the technique used to derive the elastograms from the original MRI images, elastograms shave typically one-third to one-fifth of the spatial resolution off the resolution of the original MRI images [97].

Due to flexibility, noninvasiveness, and potential clinical applications, the field of MRE has been rapidly evolving with new applications emerging for various organs, including the eye [97]. The elastic properties of the corneoscleral shell of an intact, enucleated, bovine globe were estimated using MRE and finite element modeling, assuming linear isotropic motion. Two-dimensional axisymmetrical model geometry was derived from a segmented two-dimensional MRI image, and estimations of Young’s modulus in both the cornea and sclera were made at various intraocular pressures, using an iterative flexural wave speed-matching algorithm. This ocular MRE study showed that it is possible to estimate the intrinsic elastic properties of the corneoscleral shell in an ex vivo bovine globe, suggesting that MRE may be useful to assess the mechanical properties of the eye and its anatomy. Further development of the technique and modeling process is needed to enhance the clinical potential of ocular MRE [96].