Retinal and Optic Nerve Magnetic Resonance Diffusion-Weighted Imaging in Acute Non-Arteritic Central Retinal Artery Occlusion

Abstract

Objectives:

Diffusion weighted imaging hyperintensity (DWI-H) has been described in the retina and optic nerve during acute central retinal artery occlusion (CRAO). We aimed to determine whether DWI-H can be accurately identified on standard brain magnetic resonance imaging (MRI) in non-arteritic CRAO patients at two tertiary academic centers.

Materials and Methods:

Retrospective cohort study that included all consecutive adult patients with confirmed acute non-arteritic CRAO and brain MRI performed within 14 days of CRAO. At each center, two neuroradiologists masked to patient clinical data reviewed each MRI for DWI-H in the retina and optic nerve, first independently then together. Statistical analysis for inter-rater reliability and correlation with clinical data was performed.

Results:

We included 204 patients [mean age 67.9±14.6 years; 47.5% females; median time from CRAO to MRI 1 day (IQR 1-4.3); 1.5 T in 127/204 (62.3%) and 3.0 T in 77/204 (37.7%)]. Inter-rater reliability varied between centers (κ=0.27 vs. κ=0.65) and was better for retinal DWI-H. Miss and error rates significantly differed between neuroradiologists at each center. After consensus review, DWI-H was identified in 87/204 (42.6%) patients [miss rate 117/204 (57.4%) and error rate 11/87 (12.6%)]. Significantly more patients without DWI-H had good visual acuity at follow-up (p=0.038).

Conclusions:

In this real-world case series, differences in agreement and interpretation accuracy amongst neuroradiologists limited the role of DWI-H in diagnosing acute CRAO on standard MRI. DWI-H was identified in 42.6% of patients and was more accurately detected in the retina than in the optic nerve. Further studies are needed with standardized novel MRI protocols.

INTRODUCTION

Non-arteritic central retinal artery occlusion (CRAO) occurs when blood flow is interrupted to the retina from thromboembolism or local thrombus formation. Patients present with acute painless monocular severe vision loss.¹ Occlusion duration determines visual outcomes because the retina requires a continuous oxygen supply.² CRAO must be promptly recognized in the acute phase to consider potential timely reperfusion therapies.^3,4

However, delay in diagnosis of CRAO is common.⁵ Patients can present to medical settings without appropriate diagnostic expertise. Furthermore, the typical fundoscopic examination findings seen in CRAO may not be present acutely.^6,7 Clinicians may therefore rely on ancillary testing such as optical coherence tomography to confirm the diagnosis, often delaying appropriate intervention.⁸ Brain magnetic resonance imaging (MRI) with diffusion-weighted imaging (DWI) is commonly used to assess for acute cerebral ischemia. DWI-hyperintensity (DWI-H) represents cytotoxic edema secondary to infarction and can be present within minutes and for up to 14 days after ischemic stroke.⁹ In hyperacute MRI-based stroke centers, MRI is an effective routine screening modality that reduces time to thrombolytic therapy and increases the number of patients treated within the first 60 minutes after stroke.^10,11 Additionally, MRI can screen patients with an unknown time of stroke onset to determine eligibility outside traditional time windows for IV thrombolysis^12-15 or endovascular reperfusion therapies.^15,16 DWI-H in the retina and/or optic nerve may also suggest acute ischemia, and case reports have demonstrated retinal and optic nerve DWI-H during acute CRAO.^17-19 A recent retrospective case-controlled study of 20 patients identified DWI-H in most non-arteritic CRAO patients.²⁰

Further studies are required to determine whether DWI-H can truly diagnose acute CRAO, particularly in a real world setting with larger sample sizes. We aimed to determine whether DWI-H of the retina and/or optic nerve can be accurately identified on standard brain MRI obtained within 14 days of non-arteritic CRAO as part of the standard stroke workup in patients from two large tertiary academic centers.

METHODS

Patients

This retrospective observational cross-sectional study evaluated the presence of retinal and optic nerve DWI-H on standard MRI brain in patients with acute non-arteritic CRAO at Emory University (Atlanta, GA) (Center 1) and the Mayo Clinic (AZ, FL, and MN) (Center 2). The study was approved by each site’s Institutional Review Board. The need for informed consent was waived.

All CRAO patients who presented to Center 1 from January 1, 2010 to July 31, 2020⁵ and to Center 2 from January 1, 2013 to April 31, 2021 were retrospectively identified and their medical records reviewed. Patients at Center 1 were identified in a local database using key words. Patients at Center 2 were identified using ICD-9 and ICD-10 diagnostic codes. All patients were evaluated in an affiliated ophthalmology clinic at each center. The diagnosis of CRAO was confirmed with history, ophthalmological examination, and ancillary testing, including ocular fundus photographs, optical coherence tomography, and fluorescein angiography in some patients.

We included all adult patients with non-arteritic CRAO and standard MRI of the brain and/or orbits performed within 14 days of visual loss onset as part of their stroke workup. Patients with an unclear diagnosis, incidental prior history of CRAO, age less than 18 years, those evaluated at an outside clinic, those who had an MRI performed at an outside facility or not available for direct review or without DWI sequences were excluded.

Demographic data collected included age, sex, past medical history, time to presentation, and time to MRI. Clinical characteristics included side of CRAO and visual acuity at baseline and follow-up when available. MRI details included protocol ordered (either brain or orbits), type of scanner used, magnet strength, and slice thickness.

DWI Analysis

At each center, two experienced neuroradiologists independently reviewed each MRI to determine whether DWI-H was present in the retina and/or optic nerve (A.S and R.H. for Center 1, and J.H. and T.R. for Center 2). Neuroradiologists were masked to the CRAO side. Each DWI was correlated with apparent diffusion coefficient (ADC) to ensure diffusion restriction. Patients were deidentified and presented in random order. Orbital MRI was preferentially used when available. After the initial independent review, the two neuroradiologists reassessed each MRI together to determine whether DWI-H was present. The consensus DWI review was compared to the confirmed side of the CRAO identified on physical examination to determine detection accuracy. A “miss” occurred when DWI-H was not identified in a patient. “Error” was defined as DWI-H identified in the retina and/or optic nerve contralateral to the affected eye.

Statistical Analysis

Descriptive statistics were used to summarize demographic data and time between CRAO (vision loss) and MRI. Cohen’s kappa coefficient (κ) was used to determine the overall inter-rater reliability for DWI-H between neuroradiologists, as well as the inter-rater reliability for retinal and optic nerve DWI-H separately. Agreement was categorized as poor (κ<0.00), slight (0.00≤κ≤0.20), fair (0.21≤κ≤0.40), moderate (0.41≤κ≤0.60), substantial (0.61≤κ≤0.80), and almost perfect (κ>0.80). The error and miss rates between readers were compared using Fisher’s exact test. Two-sample test of proportions was used to determine whether there was a difference in proportion of good visual acuity between patients with and without DWI-H at presentation and follow-up. A p-value of <0.05 was considered significant. Statistical analysis was performed using Stata/MP 15.0 (StataCorp. TX).

RESULTS

A total of 204 patients from two tertiary academic centers were included (Figure 1). The mean age was 67.9 ± 14.6 years and 97 (47.5%) were female. The affected eye was right in 110 patients (53.9%), left in 93 patients (45.6%), and bilateral in 1 patient (0.5%). The median time between CRAO and MRI was 1 day (IQR 1-4.3 days). MRI magnetic field strength was 1.5 T in 127/204 (62.3%) and 3.0 T in 77/204 (37.7%) patients.

Figure 1.

Flow diagram for patient selection at two tertiary centers.

Abbreviation: CRAO, central retinal artery occlusion; MRI, magnetic resonance imaging; DWI, diffusion-weighted imaging.

Imaging acquisition protocols differed between centers. At Center 1, patients were imaged on 6 different MRI scanners, and 61/75 had brain and 14/75 had orbits protocol performed. All DWI sequences were axial Echo Planar Imaging (EPI) with 5 mm slice thickness. At Center 2, patients were imaged on 14 different MRI scanners, and 126/129 had brain and 3/129 had orbits protocol performed. DWI sequences were EPI in 106/129 and Readout Segmentation Of Long Variable Echo-trains (RESOLVE) in 23/129. Slice thickness was variable with median thickness 4 mm (IQR 4-5 mm). All patients had axial DWI imaging, and 31 patients also had coronal DWI imaging available.

Imaging interpretation results for each center are summarized in Table 1. Representative examples of MRI findings are shown in Figure 2. Overall, DWI-H was present in 87/204 patients (42.6%). The miss rate was 117/204 (57.4%) and the error rate was 11/87 (12.6%). DWI-H was misidentified in 10 optic nerves and 1 retina.

Table 1.

DWI-H on standard MRI performed within 14 days of acute non-arteritic CRAO identified independently by experienced neuroradiologists at two different tertiary centers.

	Center 1 (n = 75)			Center 2 (n = 129)
	Reader 1	Reader 2		Reader 1	Reader 2
Total	27/75	22/75	κ=0.27	54/129	58/129	κ=0.65
Retina	25	6	κ=0.38	23	18	κ=0.77
Optic nerve	2	14	κ=0.19	16	25	κ=0.65
Retina and optic nerve	0	1		14	15
Bilateral	0	1		1	0
Miss rate	48/75 (64.0%)	53/75 (70.7%)	p=0.038	75/129 (58.1%)	71/129 (55.0%)	p<0.001
Error rate	1/27 (3.7%)	6/22 (27.3%)	p=0.024	1/54 (1.9%)	5/58 (8.6%)	p<0.001

Abbreviation: DWI-H, diffusion-weighted imaging hyperintensity; MRI, magnetic resonance imaging; CRAO, central retinal artery occlusion.

Figure 2.

Axial DWI for 4 separate patients showing hyperintensity (white arrows) correctly identified in the retina (A) and optic nerve (B). Hyperintensity correctly identified in both the retina and optic nerve (C) with inset image showing corresponding ADC hypointensity in the optic nerve (dashed circle). Hyperintensity correctly identified in the retina but incorrectly identified in the optic nerve (arrowhead) (D). Days between CRAO onset and MR-DWI are listed for each scan.

Abbreviation: DWI, diffusion-weighted imaging; ADC, apparent diffusion coefficient; CRAO, central retinal artery occlusion; MR, magnetic resonance.

At Center 1, DWI-H was identified in 27 patients by reader 1 and 22 patients by reader 2. The miss rate was 48/75 (64.0%) for reader 1 and 53/75 (70.7%) for reader 2 (p=0.038). The error rate between readers was 1/27 (3.7%) versus 6/22 (27.3%) (p=0.024). The overall agreement between neuroradiologists was fair (κ=0.27; standard error 0.060), for retina only was fair (κ=0.38; standard error 0.067), and for optic nerve only was slight (κ=0.19; standard error 0.055). After consensus, DWI-H was deemed present in 31/75 patients (41.3%). The miss rate was 44/75 (58.7%) and the error rate was 5/31 (16.1%).

At Center 2, DWI-H was identified in 54 patients by reader 1 and 58 patients by reader 2. The miss rate was 75/129 (58.1%) for reader 1 and 71/129 (55.0%) for reader 2 (p<0.001). The error rate between readers was 1/54 (1.9%) versus 5/58 (8.6%) (p<0.001). The overall agreement between neuroradiologists was substantial (κ=0.65; standard error 0.045), for retina only was substantial (κ=0.77; standard error 0.066), and for optic nerve only was substantial (κ=0.65; standard error 0.065). After consensus, DWI-H was deemed present in 56/129 patients (43.4%). The miss rate was 73/129 (56.6%) and the error rate was 6/56 (10.7%). For the 23 patients with RESOLVE sequencing, the inter-rater reliability was substantial (κ=0.62; standard error 0.117) and median time between CRAO and MRI was 4 days (IQR 1-8 days).

Good visual acuity at presentation (defined as 20/200 or better) was reported in 13/158 (8.2%) patients with DWI-H and 24/158 (15.2%) patients without DWI-H (p=0.40). At follow-up, good visual acuity was identified in 13/135 (9.6%) patients with DWI-H and 31/135 (23.0%) patients without DWI-H (p=0.038). Median follow-up occurred at 123 days after CRAO onset (IQR 42-213 days).

There were 7/204 patients (3.4%) who received IV thrombolytics. The median time from CRAO to receiving thrombolytics was 180 minutes (IQR 165-192.5 minutes). The median time from CRAO to MRI among these 7 patients was 1 day (IQR 0-4 days). DWI-H was identified correctly after consensus review in 4 of 7 patients (3 retinas and 1 optic nerve) and was not seen in 3 of 7 patients who received thrombolysis prior to MRI.

DISCUSSION

Our study showed that DWI-H was present in only 42.6% of 204 patients with acute CRAO, and was incorrectly identified in the unaffected eye in 12.6% of these patients. Miss and error rates between neuroradiologists significantly differed. Inter-rater reliability between academic and experienced neuroradiologists was fair at one center and substantial at the other, demonstrating that DWI-H in the retina and optic nerve is overall difficult to detect and routine brain MRI likely has a limited role in diagnosing acute non-arteritic CRAO.

Case reports have identified DWI-H in the retina^17,18 and optic nerve^19,21 during acute spontaneous or post-operative non-arteritic CRAO. One case series showed retinal DWI-H after infectious thromboembolism in a patient with endocarditis and a cardiac vegetation, but also in another patient with acute retinal necrosis secondary to herpes simplex virus.¹⁸ Another patient developed acute blindness with bilateral optic nerve restricted diffusion in the context of cavernous sinus thrombophlebitis.²² Acute ischemia was postulated as the mechanism for DWI-H in these patients. The largest retrospective case-controlled study of DWI in acute non-arteritic CRAO patients included only 20 CRAO patients and 20 age- and sex-controlled patients with ischemic cerebral stroke.²⁰ As in our study, two blinded readers evaluated all brain MRIs in a random order. DWI-H was identified in the retina in 75% of patients and in the optic nerve in 25-55% of patients. Inter-rater reliability was substantial for retinal and fair for optic nerve DWI-H. Despite careful review by expert neuroradiologists in two large academic centers, our retrospective identification rate was lower, and we noted variability in the inter-rater reliability between the two centers.

MRI can be used to diagnose acute ischemic stroke even despite limitations such as patient screening and prolonged acquisition times.²³ DWI identifies diffusion restriction that occurs with cytotoxic edema, and is an excellent modality to assess for acute cerebral ischemia.²⁴ However, imaging the retina and optic nerves poses unique challenges. The retina is a thin sheet of tightly packed cells measuring less than 1 mm in thickness.²⁵ The optic nerve measures approximately 2.2-5.2 mm in diameter and this varies throughout its course.²⁶ Optic nerve tortuosity can position the nerve out of plane and result in volume averaging. Tissue inhomogeneities in the orbits, skull base, and optic nerve sheath can create artifactual distortion and signal dropout that alter the optic nerve appearance.²⁷ Head and eye motion can lead to further artifact, and instructing patients to focus their eyes does not always reduce eye movements.^28,29 Although EPI has been the commonly used DWI technique for decades,³⁰ there can be significant artifact secondary to EPI-related geometric distortion, adjacent metal or dental amalgams, and orbit-sinus interface.²⁷ Newer DWI techniques such as RESOLVE have been studied in cerebral ischemia and may have better diagnostic capabilities with less artifact.²⁷ Image post-processing methods may also help visualize the optic nerves but currently have limitations.³¹

The large variability in agreement between the two centers likely reflects the real-world nature of our study. Patients were imaged on multiple different MRI scanners with different acquisition protocols at different time points after CRAO onset. The center with lower inter-rater reliability imaged all patients with 5 mm axial EPI slices. Better inter-rater reliability was seen at the center with access to RESOLVE sequencing, thinner slice thickness, and both axial and coronal slice orientations. These differences likely allowed for superior viewing of orbital anatomy and improved DWI-H detection in the retina and optic nerve. We did find substantial agreement between neuroradiologists for the 23 patients who had RESOLVE sequencing. Our neuroradiologists were aware that all patients had CRAO. Real-world miss and error rates will likely be much higher if radiologists are not prompted to a potential CRAO diagnosis because retinal and optic nerve DWI-H is not routinely included as part of the search pattern when interpretating brain MRI.

Interestingly, 41 of our CRAO patients had DWI-H in the ipsilateral optic nerve, suggesting acute optic nerve ischemia in addition to CRAO. The ophthalmic artery arises from the internal carotid artery and, after feeding the posterior ciliary arteries, becomes the central retinal artery, traveling within the optic nerve to the supply the inner layers of the retina.³² Whether the central retinal artery itself supplies some of the optic nerve in certain patients has been proposed.³³ DWI-H in the optic nerve has been identified during acute CRAO and posterior ischemic optic neuropathy.²¹ For CRAO, DWI-H occurring in the optic nerve may represent a proximal occlusion that results in optic nerve ischemia. However, the optic nerve is difficult to image, and optic nerve DWI-H was misidentified in the unaffected eye of 10 patients whereas DWI-H was only erroneously identified in one retina. There was also better inter-rater reliability between neuroradiologists for the retina compared to for the optic nerve.

There is an early window for intervention during which the retina can potentially be saved.³⁴ For example, patients treated with thrombolysis may have better visual improvement, and thus CRAO management commonly follows well-established algorithms for cerebral ischemia.^35,36 Clot resolution would presumably re-establish central retinal artery perfusion, and the reversal of ischemia would prevent DWI-H from occurring. Patients without DWI-H tended to have better vision at presentation, but this was not statistically significant. However, there were significantly more patients with good visual outcomes who did not have DWI-H at follow-up. DWI-H may represent more severe ischemia and portend a poorer prognosis. In our patients, treatment conclusions cannot be made because only 7 patients received thrombolysis, and only 4 of these patients had DWI-H detected. Future studies are needed to correlate the presence of retinal and/or optic nerve DWI-H with treatment and clinical outcomes.

Our study has several limitations, mostly related to its retrospective design and lack of a specific MRI protocol. There was potential selection bias as we could only include patients with available MRI scans. There was no control group for comparison, and thus sensitivity and specificity analysis could not be calculated. MRI scanners, imaging acquisition protocol, timing after CRAO, and neuroradiology interpretation schedules were not standardized. DWI images were mostly acquired using 5 mm slices, which may be too thick to effectively view the retina and optic nerve. Significant DWI-related artifacts may have limited scan interpretation and contributed to erroneous hyperintensities identified in the optic nerves. However, the goal of our study was to evaluate DWI-H on standard brain MRI routinely obtained as part of the stroke workup in various locations, and not to test a specific MRI protocol on a curated patient cohort.

Future research is needed to study newer MRI techniques. This could include prospectively imaging consecutive non-arteritic CRAO patients using dedicated orbital imaging with a standardized MRI protocol and an advanced DWI technique such as RESOLVE to improve overall resolution of the optic nerves and minimize artifact.²⁷ Half-Fourier Acquisition Single-shot Turbo spin Echo imaging (HASTE) is another diffusion-weighted sequence that can detect small middle ear cholesteatomas and may merit further investigation for the diagnosis of CRAO.³⁷ Whether non-radiologists can identify these subtle MRI changes would also need to be determined in a case-controlled study. Our study included a single MRI scan within two weeks of CRAO onset, but DWI-H dynamics and appearance over time are not known.

Conclusion

DWI-H on routine brain MRI had a limited role in detecting non-arteritic CRAO in this retrospective real-world case series. There were significant differences in inter-rater reliability miss rates, and error rates between neuroradiologists at different centers. After consensus review, DWI-H was identified in less than half of patients. Although MRI visualization of acute retinal ischemia may be useful clinically in selected centers without access to eye care providers or when the diagnosis of CRAO is uncertain, standard MRI performed as part of the routine stroke workup has current limitations in detecting acute CRAO. Innovative and more accurate DWI techniques that optimize imaging resolution and minimize artifact may be able to reliably diagnose CRAO in the future.