Advertisement

MR-EYE: High-Resolution MRI of the Human Eye and Orbit at Ultrahigh Field (7T)

      Keywords

      Key points

      • Dedicated eye imaging can be implemented at 7T to acquire high-resolution, high contrast-to-noise, and high signal-to-noise images in a feasible imaging time suitable for clinical use.
      • Simple, reproducible participant preparation techniques can be adopted to reduce the motion of the eye leading to a reduction in subsequent artefacts.
      • Sequences available at ultrahigh field can be used in current 7T clinical applications to visualize ocular structures otherwise impossible with other ophthalmic imaging.
      Video content accompanies this article at http://www.mri.theclinics.com.

      Introduction

      The adult eye is approximately 2.3 cm in length, comprising tissue structures that are often less than a millimeter in size. Clinically, posterior eye imaging of the retina is frequently conducted using optical coherence tomography (OCT), which uses low-coherence interferometry of light to determine the relative depth of structures. OCT only enables imaging of biological tissues within approximately a millimeter of the surface, however, because light can penetrate the eye through the pupil to reach the retina; the technique enables imaging of retinal structures at the posterior eye. Compared with MRI, OCT has a better axial resolution (∼3.5 μm) and allows visualization of the distinct retinal layers and the underlying vascular choroid (chor). However, the lateral field of view is limited (typically <1 cm of the retina is imaged at a time, although newer wide-field imaging exists), and there is limited penetration (approximately 1–1.5 mm) into the retina (Fig. 1A, B ). Furthermore, for eyes that are abnormally elongated (such as highly myopic eyes that tend to have a steepening of the retina posteriorly
      • Jonas J.
      • Xu L.
      Histological changes of high axial myopia.
      ), the OCT-displayed retinal curvature does not readily represent true anatomy as the device algorithm artificially flattens the 2D OCT image.
      • Kuo A.N.
      • Verkicharla P.K.
      • McNabb R.P.
      • et al.
      Posterior Eye Shape Measurement With Retinal OCT Compared to MRI.
      Figure thumbnail gr1
      Fig. 1(A) En-face image of the posterior eye (ocular fundus) showing the horizontal position of the optic nerve and macula (red arrows, corresponding to labels in panel B). The green box indicates the field of view currently possible with a wide-field lens (enabling 55° diameter view) and optical coherence. (B) An example of wide-field OCT b-scan showing the axial resolution of OCT in distinguishing the retinal layers and underlying vascular choroid. The horizontal position of the optic nerve and macula (red arrows) are labeled. (C) A 3D gradient-echo MRI image of the eye (TR/TE = 10/4 ms, 0.2 × 0.2 × 0.4 mm resolution) with superimposed green box as a schematic to indicate the maximum field of view currently possible with OCT compared with the whole eye/orbit imaging possible with MRI.
      On the other hand, MRI can noninvasively image the entire eyeball, orbit, and retrobulbar structures, without optical distortion
      • Smith G.
      • Atchison D.A.
      • Iskander D.R.
      • et al.
      Mathematical models for describing the shape of the in vitro unstretched human crystalline lens.
      (Fig. 1C). For some time now, high-field MRI has been used to investigate the etiology of cataract
      • Moffat B.A.
      • Landman K.A.
      • Truscott R.J.
      • et al.
      Age-related changes in the kinetics of water transport in normal human lenses.
      and presbyopia,
      • Moffat B.A.
      • Atchison D.A.
      • Pope J.M.
      Explanation of the lens paradox.
      albeit in vitro. With increased whole-body MRI field strengths and the development of small-surface radiofrequency (RF) coils,
      • Christoforidis J.B.
      • Wassenaar P.A.
      • Christoforidis G.A.
      • et al.
      Retrobulbar vasculature using 7-T magnetic resonance imaging with dedicated eye surface coil.
      ,
      • Jones C.E.
      • Atchison D.A.
      • Pope J.M.
      Changes in lens dimensions and refractive index with age and accommodation.
      it has become possible to visualize and study the whole eye in vivo. As such, applications have expanded to studying other ocular conditions such as myopia, glaucoma, and intraocular tumors.
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      ,
      • Liebo G.B.
      • Fuller M.L.
      • Witte R.
      • et al.
      High-Resolution Volumetric MR Ocular Imaging at 3T: A Pictorial Review With Ophthalmic and Sonographic Correlation.
      Accordingly, clinical MRI referrals to image the eye and orbit for diagnosis and management of ocular masses are ever-increasing, although such imaging can be hampered by the image resolution, and motion and susceptibility artifacts.
      • Khurana A.
      • Eisenhut C.A.
      • Wan W.
      • et al.
      Comparison of the diagnostic value of MR imaging and ophthalmoscopy for the staging of retinoblastoma.
      For the most part, clinical MRI of the eye and surrounding orbit has been conducted with a head coil or orbital coil at conventional field strengths (3T or 1.5 T).
      • Harrigan R.L.
      • Smith A.K.
      • Mawn L.A.
      • et al.
      Short Term Reproducibility of a High Contrast 3-D Isotropic Optic Nerve Imaging Sequence in Healthy Controls.
      ,
      • Kasthurirangan S.
      • Markwell E.L.
      • Atchison D.A.
      • et al.
      In vivo study of changes in refractive index distribution in the human crystalline lens with age and accommodation.
      The increase in signal-to-noise ratio (SNR) at 7T compared with conventional clinical scanners of 1.5 T and 3T
      • Pohmann R.
      • Speck O.
      • Scheffler K.
      Signal-to-noise ratio and MR tissue parameters in human brain imaging at 3, 7, and 9.4 tesla using current receive coil arrays.
      has allowed for a discussion
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      on the implementation of 7T imaging to benefit neuro-ophthalmology work. Here, we describe the use of a dedicated commercial multichannel eye coil at ultrahigh-field (7 T) magnetic field strength to acquire 2D and 3D images of ocular anatomy. With careful consideration of participant preparation and sequence selection to maximize the image quality, the first aim of this study was to demonstrate clinically feasible ultrahigh-field ocular and orbital MRI with minimal motion
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      and susceptibility artifacts. The hypothesis was that the phased array eye coil would offer the ability to reduce the scanning times of high-resolution and high-contrast
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      sequences such that a very simple fixation setup would be sufficient and successful to reduce artifact from the eye movement and position. Furthermore, we aimed to remove the need for more time-intensive techniques such as blink detection to remove motion artifacts,
      • Wezel J.
      • Garpebring A.
      • Webb A.G.
      • et al.
      Automated eye blink detection and correction method for clinical MR eye imaging.
      ,
      • Zhang Y.
      • Peng Q.
      • Kiel J.W.
      • et al.
      Magnetic Resonance Imaging of Vascular Oxygenation Changes during Hyperoxia and Carbogen Challenges in the Human Retina.
      in order to assist high-resolution MRI to be transferred into a clinical setting. The second study aim was to demonstrate the application of these techniques for quantitative studies of the ocular tissue dimension and shape, retrobulbar ocular anatomy, and the effect of gaze on the optic nerve.
      • Harrigan R.L.
      • Smith A.K.
      • Mawn L.A.
      • et al.
      Short Term Reproducibility of a High Contrast 3-D Isotropic Optic Nerve Imaging Sequence in Healthy Controls.
      ,
      • Singh A.D.
      • Platt S.M.
      • Lystad L.
      • et al.
      Optic Nerve Assessment Using 7-Tesla Magnetic Resonance Imaging.
      In order to address the hypothesis, it was necessary to assess the image quality through the region of interest (ROI) calculations and MRI parameter investigations along with reproducibility of the image quality. Our findings can be used to improve MRI investigations of an array of ocular conditions such as myopia, glaucoma, ocular tumors, cataract, and presbyopia.
      • Liebo G.B.
      • Fuller M.L.
      • Witte R.
      • et al.
      High-Resolution Volumetric MR Ocular Imaging at 3T: A Pictorial Review With Ophthalmic and Sonographic Correlation.

      Methods

      Apparatus and Safety Assessment

      The study protocol was approved by the Human Research Ethics Committee at The University of Melbourne (ID 1646275 and 1340926). Participants provided written informed consent before imaging. In total, 11 healthy participants (aged 20–41 years) free of ocular pathology were recruited. They were emmetropic, with spherical refractive errors between −0.50D and +0.75D, with normal visual acuity (6/7.5 or better) and normal ocular health as determined by a screening eye examination (BNN).
      We performed ultrahigh-field MRI using a 7 T whole-body scanner (Magnetom, Siemens Healthcare, Erlangen, Germany) and a six-element transmit/receive RF eye coil (MRI.TOOLS GmbH, Berlin, Germany), which is based on a prototype published by Graessl and colleagues
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      2013 (Fig. 2). The RF coil has a symmetric design with six overlapping 36-mm diameter transceiver loop elements tuned to 297 MHz. A custom-built coil holder was fabricated from Perspex plastic to assist with coil placement (see Fig. 2A). The aim of the holder design was to enable coil stabilization and participant comfort and to minimize the distance from the skin to the eye coil without any skin contact. The design included the need to have the ability of height adjustment to accommodate for differing participant head shapes. Positioning the coil as close to the eye as possible allowed for maximum posterior eye anatomy to be visible without any hindrance from the signal loss related to the use of a surface coil.
      • Christoforidis J.B.
      • Wassenaar P.A.
      • Christoforidis G.A.
      • et al.
      Retrobulbar vasculature using 7-T magnetic resonance imaging with dedicated eye surface coil.
      ,
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      ,
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      Scanning was conducted at the Melbourne Brain Centre Imaging Unit (Parkville, VIC, Australia) for no more than 1 hour. All participants were screened by the radiographer to ensure they were safe to be scanned.
      Figure thumbnail gr2
      Fig. 2(A) Position of the participant within the dedicated eye coil (sagittal view). The eye coil was a 6-channel transmit/receive eye coil array (297 MHz, icoil, MRI.TOOLS GmbH, Berlin, Germany). The custom-built coil holder could be adjusted to the optimal height for each participant. (B) The position of the participant within the dedicated eye coil (transverse view).

      Participant Preparation

      There are inherent problems associated with imaging the eye in vivo, the most common being motion artifacts caused by the involuntary eye movements (microsaccades) that occur several times per second.
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      Although the gain in the SNR of ultrahigh-field MRI would be beneficial for imaging the small structures of the eye, based on previous 7T studies,
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      ,
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      it was expected that susceptibility artifacts would also increase significantly with increasing field strength. Hence, it is critical that patients are prepared well to optimize imaging, and we were careful to follow previous suggestions for participant preparation,
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      ,
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      ,
      • Zhang Y.
      • Peng Q.
      • Kiel J.W.
      • et al.
      Magnetic Resonance Imaging of Vascular Oxygenation Changes during Hyperoxia and Carbogen Challenges in the Human Retina.
      ,
      • Bert R.J.
      • Patz S.
      • Ossiani M.
      • et al.
      High-resolution MR imaging of the human eye 2005.
      such as taping the eyelid of the imaged eye shut to prevent blinking and reduce susceptibility artifacts from the air-tissue interface (Fig. 3A vs 3B).
      Figure thumbnail gr3
      Fig. 33D GRE axial images (TR/TE = 13/4 ms, FA = 12°, 0.2 × 0.2 × 0.9 mm resolution) of the eye demonstrating differing patient preparation: (A) the eye shut but not taped, (B) the eye open, producing susceptibility artifact, (C) the eye taped and closed, no fixation task with the opposing eye (motion artifact), and (D) with our fixation task showing very little artifact.
      In addition, we developed our own fixation technique to maximize participant stability (Fig. 3C vs 3D). With one eye taped shut (the eye to be imaged), the other was allowed to fixate. Under normal circumstances, the eyes are yoked and will move together. Thus, if the fixating eye stays still, so too will the imaged eye. The participant was asked to look into the mirror attached to the coil that showed the fixation target on an LCD screen positioned at the far end of the magnet bore. With proper placement of the target cross on the screen, so that the fixating eye could view the fixation target centrally (hence the fixation target was shifted laterally from the center of view according to each person’s individual interpupillary distance, ∼12° of visual angle), a comfortable neutral eye position could be maintained without strain and movement. To maintain attention, the participant was given a button box and asked to press a button whenever the fixation target changed color (Video 1). Furthermore, given the potential for dry eye within the scanner from the fans, as well as from a reduced blink rate due to concentration, we found that instillation of ocular lubricants (TheraTears, Akorn Consumer Health, Ann Arbor, MI, USA) before scanning assisted in participant comfort.

      Sequence Choice and Considerations for Optimization to Visualize Relevant Anatomy

      Our aim of acquiring data with a high SNR and contrast-to-noise ratio (CNR) required the optimization of essential imaging factors for clinical use. We addressed these in the following subexperiments:
      • Experiment one—flip angle assessment in 3D gradient echo (GRE) imaging;
      • Experiment two—increase in resolution of 3D GRE imaging;
      • Experiment three—echo time assessment in 3D GRE imaging;
      • Experiment four—reproducibility of the SNR and CNR in 3D GRE imaging;
      • Experiment five—Visualization of the optic nerve thickness in coronal T2 turbo spin echo (TSE);
      • Experiment six—echo time assessment and visualization of the optic nerve position in axial T2 TSE.

      3D Gradient-Echo Imaging—Fast Low-Angle Snapshot T1-Weighted MRI

      High-resolution true 3D MRI data can be acquired using an RF spoiled 3D FLASH sequence. The axial orbital length (the distance from the anterior surface of the cornea and the central fovea of the retina) can be measured, and all ocular tissues from the cornea through to the optic tract are demonstrated.
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      ,
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      Consideration was required to optimize the sequence CNR and SNR to visualize these structures.
      Experiment one was a GRE sequence with fat saturation, acquired for one eye in the axial plane aimed at optimizing flip-angle allocation. The participant had the chosen eye imaged 6 times. The acquisition was a 2′24″ second acquisition repeated 6 times with varied excitation flip angles (5, 10, 12, 15, 20, and 30°) to determine the optimal contrast for visualization of the main ocular tissues (Fig. 4). Imaging parameters for the GRE sequence were a repetition time/echo time (TR/TE) = 13/4 ms, in-plane 81-mm field of view (FOV), voxel dimensions of 0.2 × 0.2 × 0.9 mm, flip angle (FA) = 12, matrix size of 384 x 384 pixels, 44 slices and a GRAPPA acceleration factor of 2.
      Figure thumbnail gr4
      Fig. 4Multiflip angle experiment data. (A–F) Images of 2 representative axial slices acquired from a GRE with varied flip angles (FA or θ) of 5, 10, 12, 15, 20, and 30°. TR/TE = 13/4 ms, 0.2 × 0.2 × 0.9 mm. (G) The MRI signal (arbitrary units) as a function of the flip angle from the lens (red circle), vitreous (blue cross), optic nerve (000), lateral muscle (purple star), and fatty tissue (light blue square). Lines represent nonlinear least square fitting to a well-known nuclear magnetic resonance equation.
      Experiment two consisted of a 4-min, higher resolution version of the GRE at an optimized flip angle of 12° derived from the previous experiment. All parameters were the same as the first experiment, except the slice resolution was 0.4 mm and TR was 10 ms (Fig. 5A, B ). The aim was to increase the resolution as far as possible while maintaining an achievable scan time.
      Figure thumbnail gr5
      Fig. 5GRE = TR/TE 10/4, in-plane 81-mm FOV, FA = 12, matrix size of 384 x 384 pixels, 44 slices, and a GRAPPA acceleration factor of 2. (A) Voxel dimensions of 0.2 × 0.2 × 0.4 mm, acquisition time 4:26 minutes, (B) voxel dimensions of 0.2 × 0.2 × 0.9 mm, acquisition time 2:24 minutes, and (C) semi-automated segmentation and 3D rendering of those in panel A, performed with the ITK-Snap toolbox.
      Experiment three was a TE investigation with respect to optic nerve evaluation. The same GRE sequence was used as in experiment two: 0.2 x 0.2 x 0.4 resolution and repeated with a TE of 4 ms and 10 ms (Fig. 6).
      Figure thumbnail gr6
      Fig. 6GRE with varied TE of 4 and 10 ms, all other parameters set as TR = 13, in-plane 81-mm FOV, voxel dimensions of 0.2 × 0.2 × 0.9 mm, FA = 12, matrix size of 384 x 384 pixels, 44 slices, and a GRAPPA acceleration factor of 2. Note clearer delineation of the optic nerve borders on the TE = 4 ms (red arrows).
      Experiment four was a reproducibility study of the SNR, CNR, and artifacts of 11 healthy emmetropic participants using the parameters from experiment one with an optimized flip angle of 12°, derived from experiment one (Fig. 7).
      Figure thumbnail gr7
      Fig. 7(A) Four central slices from an example set of 3D GRE axial images (TR/TE = 13/4 ms, FA = 12°, 2 × 0.2 × 0.9 mm resolution) of the eye and orbit. For optimal image quality, the imaged eye is taped shut while the participant fixates on a flashing target with the contralateral eye. (B, C) Box plots displaying the distribution of SNR and CNR across various orbit tissues in 11 healthy, emmetropic participants, respectively. Tissue ROIs are shown as color patches: lens (blue), vitreous humor (vit, dark purple), choroid (chor, red), optic nerve (on, dark green), subarachnoid space (ss, cyan), optic nerve sheath (ons, yellow), retro-orbital fat (fat, light purple), artifact (orange), and noise (light green). The central mark is the median, the edges of the box are the 25th and 75th percentiles, the whiskers extend to the most extreme data points the algorithm considers to be not outliers, and the outliers are plotted individually.

      Coronal Oblique T2 Measurements of Retrobulbar Ocular Anatomy

      A significant advantage of MRI is the ability to visualize structures behind the eyeball in vivo. We exploited this fact and conducted T2-weighted measurements of retrobulbar ocular anatomy. After the optic nerve exits the eyeball, it becomes myelinated and as it travels posteriorly to the optic chiasm, the optic nerve is surrounded by a subarachnoid space (ss) (filled with fluid, contiguous with cerebrospinal fluid), which is surrounded by the optic nerve sheath (ons) (membrane). We aimed to use ultrahigh-field MRI with slices less than 1 mm to minimize partial volume artifact to enable more accurate demarcation of the retrobulbar optic nerve and its surrounding anatomy.
      • Harrigan R.L.
      • Plassard A.J.
      • Bryan F.W.
      • et al.
      Disambiguating the optic nerve from the surrounding cerebrospinal fluid: Application to MS-related atrophy.
      ,
      • Mashima Y.
      • Oshitari K.
      • Imamura Y.
      • et al.
      High-Resolution Magnetic Resonance Imaging of the Intraorbital Optic Nerve and Subarachnoid Space in Patients With Papilledema and Optic Atrophy.
      In experiment five, we conducted 2D coronal oblique cross-sectional scans (TR/TE = 2000/64 ms, in-plane 155 mm FOV, voxel dimensions of 0.4 × 0.4 × 0.7 mm, matrix = 384 x 384 pixels, 14 slices (7 each eye), scan time = 2′34” (Fig. 8A–D )) on all 11 participants. The slices were angled perpendicular to each individual’s optic nerve with these slice positions allocated according to an initial 2D T2-weighted fast spin echo axial planning scan without fat suppression (TR/TE = 2130/73 ms, FOV = 150 mm, voxel dimensions = 0.4 × 0.4 × 0.7 mm, matrix = 384 x 384 pixels, 12 slice, scan time = 2′06″) (Fig. 8E). The acquisitions were optimized to acquire both optic nerves in the same acquisition using 2 slice groups. It is important to consider the signal void from crosstalk if the two groups intercept close to the optic nerve.
      Figure thumbnail gr8
      Fig. 8(AD) Coronal oblique slices through the optic nerve starting from the insertion point (A), with anatomic regions of interest available for identification and quantification if required in panel F. (E) Slice locations of the coronal oblique images in panels A to D representing slice 4, 3, 2, and 1 of the right eye. Sequence parameters for the coronal oblique—TR/TE = 2000/64 ms, voxel dimensions of 0.4 × 0.4 × 0.7 mm, 14 slices, 7 on each optic nerve. Slice locations demonstrated manually using in-built tools in biomedical imaging software. (G, H) Box plot showing SNR and CNR distributions from 11 healthy emmetropic participants, respectively. The central mark is the median, the edges of the box are the 25th and 75th percentiles, the whiskers extend to the most extreme data points the algorithm considers to be not outliers, and the outliers are plotted individually.

      Axial Oblique T2 Measurements of the Different Positions of Gaze and Created Tension on the Optic Nerve

      High-resolution orbital MRI has recreated interest in using the technology in conjunction with gaze and the ability to contribute to the investigation of mechanical forces on the optic nerve.
      • Kuo A.N.
      • Verkicharla P.K.
      • McNabb R.P.
      • et al.
      Posterior Eye Shape Measurement With Retinal OCT Compared to MRI.
      ,
      • Demer J.L.
      Optic nerve sheath as a novel mechanical load on the globe in ocular duction.
      In experiment six, we demonstrate the use of ultrahigh-field MRI to image the eye and its optic nerve during horizontal duction (left or right gaze, Fig. 9), relative to a neutral position using 2D T2-weighted fast spin echo axial scans without fat suppression (TR/TE = 2130/73 ms, FOV = 150 mm, voxel dimensions = 0.4 × 0.4 × 0.7 mm, matrix = 384 x 384 pixels, 12 slice, scan time = 2′06″). Initially, a single subject was scanned at varying echo times of 27, 45, and 73 ms (Fig. 10). Then, from 11 healthy emmetropic participants, data were acquired using the optimized protocol (TE = 45 ms without fat sat at 7T).
      Figure thumbnail gr9
      Fig. 9Axial oblique T2-weighted images aligned with the optic nerve in 3 gaze positions: neutral gaze (A), left gaze (B), and right gaze (C). Imaging parameters (AC): TR/TE = 2130/73 ms, voxel dimensions = 0.4 × 0.4 × 0.7 mm. Axial T2 image with fat saturation in the neutral gaze position is also shown (D). Note the enhancement of CSF surrounding the optic nerve but the loss of the signal from other anatomic structures. (E, F) A box plot of SNR and CNR distributions of various orbit tissues within 11 healthy emmetropic participants. The central mark is the median, the edges of the box are the 25th and 75th percentiles, the whiskers extend to the most extreme data points the algorithm considers to be not outliers, and the outliers are plotted individually.
      Figure thumbnail gr10
      Fig. 10Axial T2 with varied TE of 27, 45 and 73 ms on a healthy 26-year-old man. Other imaging parameters: TR = 2130 ms, FOV = 150 mm, voxel dimensions = 0.4 × 0.4 × 0.7 mm, matrix = 384x384 pixels, 12 slices.

      Statistics

      The SNR and CNR were quantified for ROIs (see Fig. 7A) from 11 emmetropes. The ROIs, shown in Fig. 7A, were drawn manually, by an MRI scientist (BAM) with substantial experience applying MRI to vision science, using MATLAB (Natick, MA, USA). The mean and standard deviations of variations of the SNR and CNR were then calculated; the SNR being defined as the ratio of image intensity to standard deviation of voxel intensities from air and the CNR being defined as the difference in the SNR from adjacent tissues.

      Results and Discussion

      Experiments 1 to 4: T1-Weighted GRE FLASH MRI

      Ipsilateral eye closed and contralateral eye fixated is the optimal participant setup

      As with previous studies, GRE FLASH images are much improved when the imaged eye is closed (see Fig. 3D).
      • Christoforidis J.B.
      • Wassenaar P.A.
      • Christoforidis G.A.
      • et al.
      Retrobulbar vasculature using 7-T magnetic resonance imaging with dedicated eye surface coil.
      ,
      • Richdale K.
      • Wassenaar P.
      • Teal Bluestein K.
      • et al.
      7 Tesla MR imaging of the human eye in vivo.
      ,
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      ,
      • Mashima Y.
      • Oshitari K.
      • Imamura Y.
      • et al.
      High-Resolution Magnetic Resonance Imaging of the Intraorbital Optic Nerve and Subarachnoid Space in Patients With Papilledema and Optic Atrophy.
      When the eye is open (see Fig. 3B), the magnetic susceptibility difference between the air and cornea creates severe magnetic field inhomogeneities leading to signal heterogeneity caused by decreased T2∗ relaxation times. This inhomogeneity extends beyond the lens and into the vitreous when the eye is left open. Although advanced automatic and manual shimming strategies including shim volume placement were explored, these proved to make little difference possibly because of the involuntary eye movement. Although in theory we expected shorter scan durations would improve imaging, this was not as problematic as initially thought. We found that even with a scan duration of 2 to 2.5 minutes, severe motion artifacts (see Fig. 3C) occurred when the participant was not asked to fixate using an interactive task. However, with fixation on a customized and interactive task (see Fig. 3D), these artifacts were dramatically reduced.

      Turn off unnecessary coil elements during reception

      An important consideration was coil application. During acquisition, only the 3 receiver loops over the eye of interest were activated, reducing motion artifact from the other eye.
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      For illustrative purpose, Fig. 11 displays the signal from the 6 individual coil elements. By eliminating the coils with limited signal but noticeable motion artifact, this will benefit the resultant image. This selective coil method also limited the amount of signal from outside the field of view being available for reconstruction and causing a wrap artifact in the displayed image.
      • Graessl A.
      • Muhle M.
      • Schwerter M.
      • et al.
      Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
      Figure thumbnail gr11
      Fig. 11Uncombined images of the six individual receive coils obtained with a 3D GRE from a left eye of a healthy 26-year-old man. Supporting the action to only acquire data from the three coils positioned closely to the eye of interest. Six-channel TX/RX eye coil array (297 MHz, icoil, MRI.TOOLS GmbH, Berlin, Germany).

      Optimize contrast by matching excitation the flip angle to ocular tissues of interest

      In experiment one by varying the flip angle (see Fig. 4) of these 3D FLASH images, it can be seen that the signal and contrast of the images can be controlled to optimize the data depending on which particular ocular tissue is of interest. By inspecting the change in the various tissue signals with the flip angle (see Fig. 4G), we determined that a nominal flip angle of 12° produced a good compromise of signal and contrast for visualization of the lens, vitreous, retina, and optic nerve.

      3D GRE achieves the highest possible image resolution

      In experiments two and three, while assessing the resolution and TE of the 3D GRE, on a single participant, a 4-min higher resolution (0.2 × 0.2 × 0.4 mm voxels) was achievable with minimal artifact (see Fig. 5A). The high-resolution images in experiment two displayed the depth of structure identification possible to have positive results for segmentation of anatomy. The marked borders of the globe itself allows for volumetric measurements and 3D modeling (see Fig. 5C).

      Minimum TE provides optimal contrast and minimized artifacts

      The results of experiment three (see Fig. 6) indicate that extending the echo time substantially decreases the contrast and signal, while also increasing T2∗-related susceptibility with artifacts easily seen on structure borders such as extraocular muscles. There was a clear visual difference in the borders of the optic nerve (see Fig. 6) when the minimum TE was used. This definition is vital for further studies where we aim to assess the optic nerve in more detail.

      Resolution is limited by participant compliance

      Subjectively through experiments one, two, and three, it was concluded that not all participants or future clinical cohorts would be able to complete the high-resolution scanning without motion artifacts despite fixation tasks. These observations meant the reproducibility study (experiment four) would be more successful with utilization of the 2′, 0.4 x 0.4 x 0.9 resolution protocol.

      3D GRE imaging provides visualization of all the major tissues in the orbit with a reproducibly excellent SNR and CNR

      The results of experiment 4 are shown as boxplots in Fig. 7B, C. The SNR results of the lens, vitreous humor, choroid, optic nerve, subarachnoid space, optic nerve sheath, and retro-orbital fat across 11 participants are also presented in Table 1. Although the coefficient of variation across participants were substantial, at an individual level, the CNRs between adjacent tissues were sufficient to distinguish them in all participants (see Fig. 7C). The mean CNRs between vitreous and lens, vitreous and choroid, optic nerve and subarachnoid space, subarachnoid space and nerve sheath, and nerve sheath and orbital fat are presented in Table 2. In one participant, susceptibility artifact prevented the lens from being resolved.
      Table 1SNR Values for the Different Ocular Tissues
      Coronal T2-Weighted TSEAxial T2-Weighted TSEAxial T1 GRE with Fat Saturation
      Anatomical RegionSNRStdCOV (%)SNRStdCOV (%)SNRStdCOV (%)
      Artifact7224125417120
      Lens165321827239
      vit22942181104642
      chor3713351535435
      Optic nerve18845281035984647
      ss3710266718271194941
      ons18738311446874349
      Fat7923291313326562239
      Abbreviations: chor, choroid; COV, coefficient of variation; on, optic nerve; ons, optic nerve sheath; ss, subarachnoid space; vit, vitreous humour.
      Table 2CNR Values for the Different Ocular Tissues
      Coronal T2-Weighted TSEAxial T2-Weighted TSEAxial T1 GRE with Fat Saturation
      Anatomical RegionCNRStdCOV (%)CNRStdCOV (%)CNRStdCOV (%)
      vit to lens2134119727097
      vit to chor1923619442046
      on to ss201158391642201785
      ss to ons19946361338322064
      ons to fat6023381002727312785
      Abbreviations: chor, choroid; COV, coefficient of variation; on, optic nerve; ons, optic nerve sheath; ss, subarachnoid space; vit, vitreous humour.

      Experiment 5

      Coronal T2 imaging is optional for quantitative imaging of retrobulbar ocular anatomy

      A significant advantage of MRI is the ability to visualize structures behind the eyeball in vivo. We exploited this and conducted T2-weighted measurements of the retrobulbar ocular anatomy (Fig. 8). After the optic nerve exits the eyeball, it becomes myelinated and effectively doubles in diameter at about 2 to 3 mm behind the eye. As it travels posteriorly to the optic chiasm, the optic nerve is surrounded by an subarachnoid space (filled with fluid, contiguous with cerebrospinal fluid), which is surrounded by the optic nerve sheath (membrane). The clinical application here is the ability to simultaneously measure these three compartments, which can become differentially compromised in disease (eg, smaller optic nerve diameter in glaucoma,
      • Moffat B.A.
      • Landman K.A.
      • Truscott R.J.
      • et al.
      Age-related changes in the kinetics of water transport in normal human lenses.
      increased fluid-filled subarachnoid space and/or optic nerve sheath diameter with raised intracranial pressure
      • Moffat B.A.
      • Atchison D.A.
      • Pope J.M.
      Explanation of the lens paradox.
      ). However, these three structures have not always been able to be distinguished from one another and the retro-orbital fat using conventional magnetic field strengths and head or orbital RF coils.
      • Christoforidis J.B.
      • Wassenaar P.A.
      • Christoforidis G.A.
      • et al.
      Retrobulbar vasculature using 7-T magnetic resonance imaging with dedicated eye surface coil.
      Although some data from 3 T MRI have shown promise in distinguishing between optic nerve, subarachnoid space, and optic nerve sheath,
      • Jones C.E.
      • Atchison D.A.
      • Pope J.M.
      Changes in lens dimensions and refractive index with age and accommodation.
      ,
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      it is hampered by the need for either low-resolution images or thick slices (>2 mm8) to achieve enough contrast and signal to visualize the optic nerve. For experiment five seen in Fig. 8, the use of sub-millimeter (0.7 mm) partial volume artifact is minimized to enable more accurate demarcation of the retrobulbar optic nerve and its surrounding anatomy.
      • Nguyen B.N.
      • Cleary J.O.
      • Glarin R.
      • et al.
      Normative retrobulbar measurements of the optic nerve using ultra high field magnetic resonance imaging.
      The SNRs (see Table 1) of the optic nerve, subarachnoid space, and nerve sheath were 18 (45%), 37 (26%), and 18 (38%), respectively. However, the CNRs (see Table 2) between the subarachnoid space, and nerve and sheath were 20 (58%) and 19 (46%), respectively. The CNRs in all subjects were above the Rose criteria
      • Watts R.
      • Wang Y.
      k-space interpretation of the Rose Model: noise limitation on the detectable resolution in MRI.
      of 5 for adequate detection of boundaries.

      Experiment 6

      Axial T2 imaging is optimal for differentiating positions of gaze that create tension on the optic nerve

      The ability of MRI to view the trajectory of the optic nerve behind the eye has allowed technology in conjunction with measures of strain and tension to quantify mechanical forces on the optic nerve.
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      ,
      • Liebo G.B.
      • Fuller M.L.
      • Witte R.
      • et al.
      High-Resolution Volumetric MR Ocular Imaging at 3T: A Pictorial Review With Ophthalmic and Sonographic Correlation.
      Typically, the eye can achieve a wide range of duction (rotation of the eye on the vertical or horizontal axis) of up to 60°.
      • Khurana A.
      • Eisenhut C.A.
      • Wan W.
      • et al.
      Comparison of the diagnostic value of MR imaging and ophthalmoscopy for the staging of retinoblastoma.
      It is thought that the outer connective tissue coat of the optic nerve, the optic nerve sheath, can exert significant mechanical force on the optic nerve head and peripapillary vascular tissue when the eye undergoes duction. Repetitive eye movements and therefore repetitive strain on the optic nerve as it enters the eye have therefore been proposed as a mechanism of damage to susceptible optic nerves in diseases such as glaucoma.
      • Niendorf T.
      • Paul K.
      • Graessl A.
      • et al.
      [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
      In Fig. 9A–C, showing experiment 6 results, the high-resolution axial T2 images of the eye and its optic nerve during horizontal duction (right and left gaze) relative to a neutral position can be easily visualized. These high-resolution images have adequate signal and contrast (Rose
      • Watts R.
      • Wang Y.
      k-space interpretation of the Rose Model: noise limitation on the detectable resolution in MRI.
      criteria >5) to easily quantify the gaze angle, optic nerve length, anterior-posterior depth, and ocular axial length. In addition, with fat saturation (see Fig. 9D), the cerebrospinal fluid surrounding the optic nerve can be clearly identified and quantified (if required). The variation in echo time appears to have no obvious visual effect on tissue differentiation as seen in Fig. 10. The decision to adopt the echo time of 45 ms was seen as a compromise of an appropriate echo time for T2 image weighting, without the signal loss that is known to arise from increasing the echo time, because of T2* dephasing.
      • Pohmann R.
      • Speck O.
      • Scheffler K.
      Signal-to-noise ratio and MR tissue parameters in human brain imaging at 3, 7, and 9.4 tesla using current receive coil arrays.
      ,
      • Cao Z.
      • Park J.
      • Cho Z.-H.
      • et al.
      Numerical evaluation of image homogeneity, signal-to-noise ratio, and specific absorption rate for human brain imaging at 1.5, 3, 7, 10.5, and 14T in an 8-channel transmit/receive array.

      Highest resolution and contrast anatomic imaging of the eye and orbit can be achieved with clinically available pulse sequence technology

      As clinical translation at 7T becomes more prominent, a reoccurring question is, can the application of sequences developed by research sites be implemented successfully clinically? In this study, all sequences used are product sequences, so they can be adapted to a clinical setting; however, consideration of coil choice would be significant. In order to benefit from the improved SNR and CNR that are attainable at high field, a dedicated eye coil would potentially be required. As discussed, the susceptibility artifact when scanning this region of the head must to be treated with caution and care. The susceptibility artifact and sequence time are just two of the factors that may hinder the move to advanced imaging techniques such as diffusion, fMRI, arterial spin labeling, or susceptibility imaging without input and adoption of current research sequences.
      • Paul K.
      • Huelnhagen T.
      • Oberacker E.
      • et al.
      Multiband diffusion-weighted MRI of the eye and orbit free of geometric distortions using a RARE-EPI hybrid.

      Summary

      In conclusion, we have shown that the synergistic combination of a dedicated phased array eye RF coil, a field strength of 7T, parallel image reconstruction, and careful participant preparation allows for high-resolution and contrast MRI images of the entire eye and orbit to be acquired. We believe these to be some of the highest resolution in vivo human MRI eye images published to date with protocols readily available on clinical 7T MRI scanners. We were able to image the retrobulbar optic nerve in 2D and 3D while avoiding significant motion artifact by adopting appropriate visual fixation and eye preparation techniques, the feasibility of which was demonstrated by reproducing high-quality images across a cohort of healthy emmetropic participants.

      Acknowledgments

      The authors thank MRI.TOOLS GmbH, Berlin, Germany, for guidance in implementing coil characteristics during pilot testing and Mr Trevor McQuilland and Mr Edward Green at the Department of Biomedical Engineering, University of Melbourne, for the design and fabrication of the coil stabiliser. The authors acknowledge the facilities and the scientific and technical assistance of the Australian National Imaging Facility at the Melbourne Brain Centre Imaging Unit. This work was supported by a research collaboration agreement with Siemens Healthineers.

      Disclosure

      A.M. McKendrick receives research support from Heidelberg Engineering GmBH. Funding was provided by Melbourne Neuroscience Institute Fellowship (B.N. Nguyen); The University of Melbourne McKenzie Fellowship (J.O. Cleary); and Melbourne Neuroscience Institute Interdisciplinary Seed Funding Grant (A.M. McKendrick, B.V. Bui, and R.J. Ordidge). The National Imaging Facility supports the 7T system at the Melbourne Brain Center Imaging Unit and the MRI Facility Fellow (B.A. Moffat).

      Supplementary data

      References

        • Jonas J.
        • Xu L.
        Histological changes of high axial myopia.
        Eye (Lond). 2013; 28: 113-117
        • Kuo A.N.
        • Verkicharla P.K.
        • McNabb R.P.
        • et al.
        Posterior Eye Shape Measurement With Retinal OCT Compared to MRI.
        Invest Ophthalmol Vis Sci. 2016; 57: OCT196-OCT203
        • Smith G.
        • Atchison D.A.
        • Iskander D.R.
        • et al.
        Mathematical models for describing the shape of the in vitro unstretched human crystalline lens.
        Vision Res. 2009; 49: 2442-2452
        • Moffat B.A.
        • Landman K.A.
        • Truscott R.J.
        • et al.
        Age-related changes in the kinetics of water transport in normal human lenses.
        Exp Eye Res. 1999; 69: 663-669
        • Moffat B.A.
        • Atchison D.A.
        • Pope J.M.
        Explanation of the lens paradox.
        Optom Vis Sci. 2002; 79: 148-150
        • Christoforidis J.B.
        • Wassenaar P.A.
        • Christoforidis G.A.
        • et al.
        Retrobulbar vasculature using 7-T magnetic resonance imaging with dedicated eye surface coil.
        Graefes Arch Clin Exp Ophthalmol. 2013; 251: 271-277
        • Jones C.E.
        • Atchison D.A.
        • Pope J.M.
        Changes in lens dimensions and refractive index with age and accommodation.
        Optom Vis Sci. 2007; 84: 990-995
        • Niendorf T.
        • Paul K.
        • Graessl A.
        • et al.
        [Ophthalmological imaging with ultrahigh field magnetic resonance tomography: technical innovations and frontier applications].
        Klin Monbl Augenheilkd. 2014; 231: 1187-1195
        • Liebo G.B.
        • Fuller M.L.
        • Witte R.
        • et al.
        High-Resolution Volumetric MR Ocular Imaging at 3T: A Pictorial Review With Ophthalmic and Sonographic Correlation.
        Neurographics. 2016; 6: 281-291
        • Khurana A.
        • Eisenhut C.A.
        • Wan W.
        • et al.
        Comparison of the diagnostic value of MR imaging and ophthalmoscopy for the staging of retinoblastoma.
        Eur Radiol. 2013; 23: 1271-1280
        • Harrigan R.L.
        • Smith A.K.
        • Mawn L.A.
        • et al.
        Short Term Reproducibility of a High Contrast 3-D Isotropic Optic Nerve Imaging Sequence in Healthy Controls.
        Proc SPIE Int Soc Opt Eng. 2016; 9783: 97831L
        • Kasthurirangan S.
        • Markwell E.L.
        • Atchison D.A.
        • et al.
        In vivo study of changes in refractive index distribution in the human crystalline lens with age and accommodation.
        Invest Ophthalmol Vis Sci. 2008; 49: 2531-2540
        • Pohmann R.
        • Speck O.
        • Scheffler K.
        Signal-to-noise ratio and MR tissue parameters in human brain imaging at 3, 7, and 9.4 tesla using current receive coil arrays.
        Magn Reson Med. 2016; 75: 801-809
        • Richdale K.
        • Wassenaar P.
        • Teal Bluestein K.
        • et al.
        7 Tesla MR imaging of the human eye in vivo.
        J Magn Reson Imaging. 2009; 30: 924-932
        • Graessl A.
        • Muhle M.
        • Schwerter M.
        • et al.
        Ophthalmic magnetic resonance imaging at 7 T using a 6-channel transceiver radiofrequency coil array in healthy subjects and patients with intraocular masses.
        Invest Radiol. 2014; 49: 260-270
        • Wezel J.
        • Garpebring A.
        • Webb A.G.
        • et al.
        Automated eye blink detection and correction method for clinical MR eye imaging.
        Magn Reson Med. 2017; 78: 165-171
        • Zhang Y.
        • Peng Q.
        • Kiel J.W.
        • et al.
        Magnetic Resonance Imaging of Vascular Oxygenation Changes during Hyperoxia and Carbogen Challenges in the Human Retina.
        Invest Ophthalmol Vis Sci. 2011; 52: 286-291
        • Singh A.D.
        • Platt S.M.
        • Lystad L.
        • et al.
        Optic Nerve Assessment Using 7-Tesla Magnetic Resonance Imaging.
        Ocul Oncol Pathol. 2016; 2: 178-180
        • Bert R.J.
        • Patz S.
        • Ossiani M.
        • et al.
        High-resolution MR imaging of the human eye 2005.
        Acad Radiol. 2006; 13: 368-378
        • Harrigan R.L.
        • Plassard A.J.
        • Bryan F.W.
        • et al.
        Disambiguating the optic nerve from the surrounding cerebrospinal fluid: Application to MS-related atrophy.
        Magn Reson Med. 2016; 75: 414-422
        • Mashima Y.
        • Oshitari K.
        • Imamura Y.
        • et al.
        High-Resolution Magnetic Resonance Imaging of the Intraorbital Optic Nerve and Subarachnoid Space in Patients With Papilledema and Optic Atrophy.
        Arch Ophthalmol. 1996; 114: 1197-1203
        • Demer J.L.
        Optic nerve sheath as a novel mechanical load on the globe in ocular duction.
        Invest Ophthalmol Vis Sci. 2016; 57: 1826-1838
        • Nguyen B.N.
        • Cleary J.O.
        • Glarin R.
        • et al.
        Normative retrobulbar measurements of the optic nerve using ultra high field magnetic resonance imaging.
        Invest Ophthalmol Vis Sci. 2019; 60: 6109
        • Watts R.
        • Wang Y.
        k-space interpretation of the Rose Model: noise limitation on the detectable resolution in MRI.
        Magn Reson Med. 2002; 48: 550-554
        • Cao Z.
        • Park J.
        • Cho Z.-H.
        • et al.
        Numerical evaluation of image homogeneity, signal-to-noise ratio, and specific absorption rate for human brain imaging at 1.5, 3, 7, 10.5, and 14T in an 8-channel transmit/receive array.
        J Magn Reson Imaging. 2015; 41: 1432-1439
        • Paul K.
        • Huelnhagen T.
        • Oberacker E.
        • et al.
        Multiband diffusion-weighted MRI of the eye and orbit free of geometric distortions using a RARE-EPI hybrid.
        NMR Biomed. 2018; 31: e3872