We provided a detailed protocol for a standardized method of optic nerve assessment and quantification using MRI, utilizing a widely available imaging sequence, and open access software for image analysis. Following this standardized protocol would provide meaningful data for comparison between different patients and different studies.
Optic nerve assessment is an important aspect of glaucoma diagnosis and follow-up. This project describes a protocol for a unified methodology of optic nerve cross-sectional assessment and quantification using 3 T MRI for image acquisition and ImageJ’s Fiji software for image processing quantification. Image acquisition was performed using 3 T MRI, with proper instructions for the patient to ensure straight fixation during imaging. A T2-weighted fat suppressed sequence was used. A coronal cut taken 3 mm behind the globe and perpendicular to the optic nerve axis should be uploaded to the software. Using the threshold function, the white matter area of the optic nerve is selected and quantified, thus, eliminating inter-individual measurement bias. We also described the normal limits for the optic nerve cross-sectional area according to age, based on previously published literature. We used the described protocol to assess optic nerve of a suspected glaucoma patient. The optic nerve cross-sectional area was found to be within the normal limits, a finding further confirmed via optical coherence tomography of the optic nerve.
Glaucoma is an optic neuropathy that is considered to be the most common cause of irreversible blindness1. Despite that, it is still poorly understood in terms of its pathophysiology and diagnosis, with no single standard reference for establishing the diagnosis2. According to the National Institute for Health and Care Excellence (NICE) the diagnosis of primary open-angle glaucoma (POAG) requires the assessment of multiple domains, including optic disc assessment on fundus exam or optical coherence tomography (OCT) imaging, visual field assessment, and intraocular pressure measurement 3. The idea behind diagnosing glaucoma is establishing the presence of progressing optic neuropathy, which can be done quantitively on OCT4. In this regard, MRI can also be used for optic nerve assessment and quantification of its white matter area5, but for this to be clinically meaningful, the protocol used in optic nerve white matter quantification needs to be standardized. Moreover, a protocol should also accommodate inter-individual variation, a factor that might affect accuracy in different diseases6.
Optic nerve assessment in glaucoma is optimally assessed via ophthalmic imaging, including OCT, where the most anterior part of optic nerve (e.g., optic disc) is assessed. On the other hand, the use of MRI for optic nerve assessment usually assesses the retrobulbar part of optic nerve at various distances from globe. Several studies found a strong correlation between optic disc assessment using OCT and MRI7,8. However, there is still no unified protocol for optic nerve assessment and quantification on MRI. Outlining the optic nerve border on MRI has been used to quantify its cross-sectional area5. However, this method has considerable inter-rater variability, as it needs to be done by an experienced rater and requires considerable time for outlining. The aim of the current project was to provide a protocol for a unified methodology for optic nerve cross sectional assessment and quantification using 3 T MRI for image acquisition and ImageJ's Fiji software for image processing and quantification.
The following study was approved by the research committee and institutional review board of the University of Jordan Hospital. The following protocol will describe the imaging technique used to acquire MRI images, followed by image processing and optic nerve quantification using Fiji software.
1. MRI image acquisition
NOTE: MR image acquisition was done using a 3 Tesla (3 T) MRI to perform multiplanar T2-weighted fat suppression sequence (Table of Materials).
2. Image analysis
The cup to disc ratio for a 30-year-old male patient presenting for a checkup ophthalmology exam was 0.8 (Figure 1A), which is suspicious and might be suggestive of glaucoma. Upon performing an optical coherence tomography for nerve fiber layer thickness, we found that the nerve thickness was within the normal limits for age (Figure 1B). The patient was scheduled for an orbit MRI, where a coronal cut for optic nerve assessment was ordered and performed as per the abovementioned protocol.
We obtained a coronal MRI cut, 3 mm behind the optic disc. The white matter mean intensity was 94.372 (SD 7.085), which resulted in white matter intensity range of:
Lower limit = 94.372 – 21.255 = 73.117
Upper limit = 94.372 + 21.255 = 115.627
Figure 2 shows the coronal image (Figure 2A), the coronal image after applying the white matter threshold using the calculated upper and lower limits (Figure 2B), and the optic nerve white matter for quantification (Figure 2C). The cross-sectional area for the white matter of the left optic nerve was 6.9 mm2 (0.069 cm2), which is within the normal limits for his age, as shown in Table 1.
Figure 1: Fundus image showing high cup to disc ratio, which might be suggestive of glaucoma (A). An optical coherence tomography for the nerve fiber layer (NFL) showing a within normal limits NFL (B). Please click here to view a larger version of this figure.
Figure 2: Coronal T2 -weighted fat suppressed MRI image obtained perpendicular to the optic nerve 3 mm behind the optic disc (A). The same coronal cut after applying a pre-calculated threshold range (B). Optic nerve white matter (C). Please click here to view a larger version of this figure.
Study | Sample size | Age (Years) | Mean cross sectional area (mm2) | Imaging sequence |
Bäuerle, 2013. 10 | 15 | Mean (SD) 24.5 ± 0.8 | 5.69 ± 0.77 | T2-weighted turbo spin echo (TSE) sequence |
Wang 2012. 11 | 21 | Mean (SD) 51.6±12.0 | 5.03 ± 0.35 | T2-weighted fast recovery fast spin echo (FRFSE) sequence |
Weigel, 2006. 12 | 32 | Mean (range) 25 (22–39) | 5.7 ± 0.6 | T2-weighted turbo spin echo (TSE) sequence |
Yiannakas, 2013. 13 | 8 | Mean (range) 31 (29–33) | 6.2 (1.3) | T2-fat-suppressed |
Al-Haddad, 2018. 14 | 211 | Median (interquartile) 8.6 (3.9–13.3) | 4.0 ± 0.20 * | T1-weighted inversion recovery sequence |
*Calculated using the provided optic nerve diameter. |
Table 1 shows the normal range of optic nerve cross sectional diameter using MRI 3mm from globe, as found by previous studies.
We described a protocol to assess and quantify optic nerve white matter that might be used for glaucoma patient assessment. The protocol uses widely available imaging sequences for image acquisition, and it uses the open-source Fiji software for image analysis. We standardized the image parameters that were previously found to be most accurate and highly reproducible in optic nerve image acquisition, including asking the patient to fixate straight ahead, using T2 with fat suppression sequence, and capturing the cross-sectional area 3 mm behind the globe. In addition, we described a detailed image analysis method that eliminates manual segmentation and corrects for signal variability between patients. The importance of this protocol is that it eliminates the variation in region of interest (ROI) segmentation by the radiologist, which is usually the main source of error in optic nerve assessment on MRI12. While we tried to provide normative data for optic nerve cross sectional area with Table 1, further data using the standard protocol described is needed for the use and comparison in clinical settings. Such data need to accommodate different age groups due to the age variation in optic nerve size, as shown in Table 1. Such variation is not evident between genders15, but was recently suggested to be present for refractive error16.
Previous studies applied different methodologies for optic nerve white matter quantification, and they mostly used the software present in their workstation for image analysis. Initial studies on optic nerve assessment adopted a quantification approach based on cross sectional area, using manual segmentation by technicians or radiologists12,17. Wang et al. also used manual segmentation of optic nerve cross-sectional area at different distances from the globe for correlation with OCT 11. Omodaka et al. used the average cross-sectional area on the coronal cut and the length of the optic nerve from the disc to the optic chiasm on the axial cut through manual annotation to extract indicators of the optic nerve for correlation with OCT8. Despite being correlated with OCT, reproducibility of this method might not yield the required accuracy for longitudinal assessment of optic nerve. Ramli et al. quantified the volume of the optic nerve through manual segmentation of the isointense signal at all axial sections 5, an approach that might miss the optic nerve matter not captured by the axial sections themselves, human error during image manual segmentation, or even in the determination of the optic nerve length to be included in the quantification assessment.
While different studies used the assessment of the cross-sectional area of the optic nerve, they differed in the distance of measurements from globe. Wang et al. assessed 3 mm, 9 mm, and 15 mm behind the globe, and found that the 3 mm cross-sectional assessment had the highest correlation with intraocular pressure11. Bäuerle et al. analyzed the reproducibility of assessment of the optic nerve on MRI at 3 mm and 5 mm behind the globe, and they found a good assessment for both cases10. Lagrèze et al. measured the cross-sectional area 5 mm, 10 mm, and 15 mm behind the globe and found that the cross-sectional assessment was most accurate in the 5 mm cross sectional area compared to measurements further from the globe17. In this protocol, we used a 3 T MRI for images acquisition, where its use in optic nerve assessment was previously found to be superior to 1.5 T MRI18,19. The increasingly utilized 7 T MRI might also provide superior results but will also require its normative values. Regarding MRI sequence used, we used T2 fat suppression sequence, mainly due to its wide availability and its inherent ability to delineate the optic nerve surrounding CSF after eliminating the surrounding intraconal fat. Previous studies used other sequences with reliable results, including the half-Fourier single-shot turbo spin-echo (HASTE) sequence and Diffusion Tensor Imaging (DTI) sequence7,12, which might not be widely available.
An important aspect to consider during image acquisition is making sure that the patient is fixating on a straight target, as fixating on a non-straight target during imaging will yield a non-accurate optic nerve quantification12. Fixation in OCT is monocular on a close target, requiring the patient to have good visual acuity in the eye to be assessed to see the near target with one eye, whereas for MRI the target is further away, the fixation is binocular, and requires fewer visual demands. However, fixation might still be an issue for patients who have high refractive error or poor vision. While the use of MRI to assess and follow patients with glaucoma might not be feasible in the presence of low cost, simpler imaging techniques, including the OCT, MRI can be useful in special situations where OCT do not provide conclusive data, or OCT itself cannot be obtained, such as in the presence of significant ocular medial opacity. Furthermore, the described protocol can be used when MRI imaging is warranted in cases of unexplained optic neuropathy to exclude secondary causes20,21.
One of the main limitations of this protocol is the inability to assess patients who cannot properly fixate, including patients with poor visual acuity in both eyes. In this regard, the use of sound stimulus will improve the quality of fixation during image acquisition22. Moreover, as a new methodology, future studies are needed to depict normal values for MRI-based cross-sectional areas for optic nerve white matter. The importance of establishing normal values is further stressed by the fact that optic nerve is also composed of significant amount of connective tissue23, a tissue that does not have similar functional abilities as nerve fibers. While quantification of optic nerve fiber layer thickness in OCT might provide a false impression of remanent nerve tissue due to the inclusion of connective tissue in the quantification process24, such a false impression is not present in this MRI-based quantification method. Motion artifacts might also lead to blur in images, especially upon eye movement during exam. While it should be avoided during imaging, establishing white matter range will reduce the impact of such artifacts on the accuracy of optic nerve white matter quantification, as the changes caused by the motion artifact on cerebral white matter is almost like the optic nerve's white matter.
The main strength of the current protocol is the elimination of inter-individual differences during optic nerve quantification, even when performed by non-specialized medical doctors or technicians. In addition, it used a widely available open-source software for image analysis. While it is not feasible to do a dedicated MRI imaging for optic nerve quantification, especially in the presence of the OCT, it is recommended to perform this protocol during MRI imaging done for other purposes, including the exclusion of secondary causes of optic neuropathy and glaucoma.
The authors have nothing to disclose.
We would like to thank Faris Haddad and Hasan El-Isa for their important contribution in video filming and development.
Magnetic resonance imaging (MRI) machine | Siemens Magnetom Verio | N/A | 3T MRI scanner |