Purpose To evaluate the thickness of the macular ganglion cell and inner plexiform layer (GCIPL) using spectral domain optical coherence tomography (SD OCT) in patients with brain lesions.
Methods This case-control study included 58 healthy subjects and 98 patients with brain lesions confirmed by MRI. GCIPL and peripapillary retinal nerve fiber layer (pRNFL) thicknesses were determined using the Cirrus SD OCT. Area under the receiver operating characteristic curve (AUC) values of pRNFL and GCIPL thickness were used to discriminate patients with brain lesions from normal controls.
Results Average GCIPL thickness showed a good correlation with visual field mean deviation (r2=0.342, p<0.001). All GCIPL parameters, including average thickness (71.9±8.6 vs 85.1±4.8 μm, p<0.001), differed between the patient and control groups. The AUC of the average GCIPL thickness was significantly greater than that of average pRNFL thickness (0.941 vs 0.823, p<0.001).
Conclusions Our results suggest that various kinds of brain lesions with different locations show considerable reduction in GCIPL thickness. Thickness of the GCIPL performed better than conventional pRNFL thickness for the diagnosis of retinal ganglion cell damage induced by brains lesions. The pattern of GCIPL loss may be of particular usefulness in recognising a potential intracranial lesion in cases suspected of having normal-tension glaucoma. GCIPL thickness determined by OCT can be an early and useful marker to estimate the status of the visual pathway in various brain lesions.
- Visual pathway
Statistics from Altmetric.com
Optical coherence tomography (OCT) is a technology that can image the posterior pole of the eye. Determination of the thickness of the retinal nerve fiber layer (RNFL) by OCT is widely used to assess glaucomatous structural changes and evaluate neurophthalmological pathologies.1–4 With improved image resolution in spectral domain (SD) OCT compared with previous versions, the evaluation of segmented inner retinal layer thickness, which targets retinal ganglion cell (RGC) thickness, has become commercially available.5–7 More than 50% of RGCs are located in the inner area of the macula. As a result, measurement of the thickness of the ganglion cell and inner plexiform layer (GCIPL) in the macular area is used both in clinical practice and for research of diseases related to RGC dysfunction.8–11
Retrograde trans-synaptic degeneration of neurons, also known as trans-neuronal retrograde degeneration (TRD), occurs following damage to the human central nervous system. Recent studies have used OCT to demonstrate thinning of the RNFL in brain lesions.12–17 These findings confirm the occurrence of TRD of neurons in the human visual pathway and suggest the possibility of using OCT-derived measurements of the RNFL to diagnose TRD. Measurements of the thicknesses of both RNFL and GCIPL using OCT are good indicators of quantitative RGC damage. Assessment of GCIPL thickness may provide a more direct measure of RGC damage than that of RNFL thickness. Axons of RGCs have a characteristic topographic distribution in the visual pathway. Nasal-sided RGC axons in the optic nerve cross at the optic chiasm and join with temporal-sided RGC axons of the contralateral eye at the postchiasmal area and become the optic track of the contralateral side, which radiates to the lateral geniculate body, the optic radiation and the visual cortex. Temporal-sided RGC axons in the optic nerve become an ipsilateral side optic track after joining with nasal-sided RGC axons of the contralateral eye. Therefore, chiasmal and postchiasmal lesions manifest as a visual field (VF) defect with respect to the vertical meridian. This characteristic topography of RGCs in the visual pathway might aid the localisation of brain lesions in instances of TRD.
The current study aimed to use SD OCT to evaluate the thickness of the macular GCIPL in brain lesions. The detectable loss of RNFL precedes functional decay in the standard automated perimetry in glaucoma.18 ,19 RNFL changes and significant loss of RGC population can occur in macular zone before detectable VF changes take place in glaucoma.20 Thus, a structural measure enables earlier diagnosis of glaucoma than functional test. Hence, we also explored this possibility that a structural measure may enable earlier diagnosis of damages in visual pathway in patients with brain lesions who did not have apparent VF abnormality.
This case-control study included healthy control subjects and patients with brain lesions confirmed by MRI. Individuals who met the inclusion criteria and were examined at the ophthalmology clinic of the Asan Medical Center were selected by retrospective review of their medical records. The patients with brain lesion were referred for an ocular examination either from the neurology or neurosurgery departments of the same hospital, or from their local eye clinic. Brain lesions were diagnosed by brain MRI and confirmed by radiology specialists. MRI was performed either at the ophthalmologic or neurology/neurosurgery department. Age-matched healthy eyes formed the control group, which comprised hospital staff, their family, spouses of patients or volunteers from the eye clinic and hospital. Control data were adapted from a previous study.7
All participants received a complete ophthalmological examination that included recording medical, ocular and family histories; testing of visual acuity (VA), testing of the Humphrey field analyser Swedish Interactive Threshold Algorithm 24-2 or 30-2 (Carl Zeiss Meditec, Dublin, California, USA), measurement of the intraocular pressure (IOP) using Goldmann applanation tonometry, stereoscopic optic disc photography and SD OCT imaging.
For inclusion in the study, all participants had to meet the following criteria: a best corrected VA (BCVA) of 20/40 or better with a spherical refractive error of between −6.0 and +4.0 diopters (D), a cylinder correction of within ±3 D, and the presence of a normal anterior chamber and an open angle upon slit-lamp and gonioscopic examinations. Both the control group and the group of patients with brain lesions had IOP values <21 mm Hg and no histories of retinal disease (including diabetic or hypertensive retinopathy), eye trauma, intraocular surgery, optic nerve disease (including glaucoma) or any systemic disease other than a brain lesion that might affect the VF. Patients with optic disc and macular pathologies, such as papilloedema or epiretinal membranes, were excluded because these diseases prevent accurate measurements of the thicknesses of the GCIPL and peripapillary RNFL (pRNFL). Control eyes were confirmed to appear normal after examination of the optic disc by stereoscopic optic disc photography and VF of each eye. The patients with brain lesions were divided into two subgroups, which comprised patients with VF abnormality (perimetric brain lesion group (PBG)) and without VF abnormality (preperimetric brain lesion group (PPBG)). Members of the PBG had to have VF abnormalities in both eyes. Abnormal VF was defined as a cluster of three points with a probability <5% on the pattern deviation map, including at least one point with a probability of <1% or a cluster of two points with a probability of <1% with respect to the vertical meridian. Only reliable VF test results (false-positive errors <15%, false-negative errors <15% and fixation loss <20%) were included in analysis. The average value of both eyes in terms of BCVA, IOP, spherical equivalent, and VF mean deviation (MD) were calculated and used for analysis.
The study was approved by the Institutional Review Board (IRB) of the Asan Medical Center and abided by the tenets of the Declaration of Helsinki. Since the patients with brain lesions were selected by retrospective review of their medical recording, informed consent was exempted by IRB.
SD OCT imaging
All SD OCT images were obtained using a Cirrus HD OCT system (Software V.18.104.22.168, Carl Zeiss Meditec, Dublin, California, USA). Pupil dilatation was performed if necessary. The image acquisition procedure employed was previously described in detail.4 ,7 We obtained the pRNFL thicknesses using the optic disc cube mode and measured GCIPL thickness using the macular cube mode. Optic disc cube data were obtained from a three-dimensional dataset composed of 200 A-scans derived from 200 B-scans that covered a 6×6 mm area centred on the optic disc. After the creation of an RNFL thickness map from the cube dataset, the software automatically determined the centre of the disc and extracted a circumpapillary circle (1.73 mm radius) from the dataset to measure the pRNFL thickness. Macular cube data were obtained from a three-dimensional dataset composed of 512 A-scans derived from 128 B-scans that covered a 6×6 mm area centred on the fovea. The area where GCIPL thickness was determined included a layer from the outer boundary of the RNFL to the outer boundary of the inner plexiform layer. Hence, GCIPL thickness was determined from a combination of the RGC layer and the inner plexiform layer. Eight thicknesses were measured within an annular area centred on the fovea; these were the average, minimum (lowest thickness over a single meridian crossing the scanning annulus) and six sectoral (superotemporal, superior, superonasal, inferonasal, inferior and inferotemporal) values. Mean values for both the right eye and left eye were used to analyse thickness of both the GCIPL and pRNFL.
All accepted images exhibited either a centred optic disc or a fovea, were well-focused with even and adequate illumination, exhibited no eye motion within the measurement circle, displayed no segmentation failure and had a signal strength of ≥7.
The Wilk–Shapiro test was used to test the distribution of numerical data. Normally distributed data were compared among normal control, PBG and PPBG patients using the analysis of variance test with Bonferroni posthoc comparisons. The χ2 test was used to compare categorical data. Correlation between the VF MD value and the average GCIPL thickness was assessed by Pearson correlation analysis. Area under the receiver operating characteristic curves (AUC) values were calculated to assess the ability of each average pRNFL thickness and GCIPL thickness to differentiate patients with brain lesions from normal control subjects. Subgroup analysis was performed with PBG and PPBG patients. The DeLong method was employed to evaluate the statistical significance of differences in the AUC values.21 Statistical analysis was performed using SPSS V.21.0 (IBM SPSS, Chicago, Illinois, USA) and MedCalc V.9.6 (Mariakerke, Belgium). A p value <0.05 was considered statistically significant.
The final analysis included 98 patients with brain lesions and 58 normal control subjects. Among the 98 subjects in the patient group, 49 were men and 49 were women, whereas the control group comprised 24 men and 34 women. All participants were East Asians (Koreans). Among the 98 subjects in the patient group, 10 patients were tested by brain MRI later than OCT examination. Those 10 patients had been treated with IOP-lowering medication for suspected normal tension glaucoma when referred from a local eye clinic. Mean time from MRI to OCT examination of the remained 88 patients was 1045.6±1313.6 (range; 2–5810) days. The mean ages of members of the two groups were comparable. Mean corrected VA, IOP and VF MD values were lower in the brain lesion group. Demographics and baseline characteristics of the participants are described in table 1.
Among the 98 patients with brain lesions, 49 had a VF defect (and were thus classified as PBG), whereas 49 patients had a normal VF (and were thus classified as PPBG). Pituitary adenoma was the most frequent pathology in members of the PBG (16 patients), with cerebral infarction as the next most frequent pathology (7 patients). Cerebral infarction was the most frequent pathology in members of the PPBG (20 patients), with pituitary adenoma as the next most frequent pathology (14 patients). Specific diseases of the brain lesions in each PBG and PPBG patient are described in table 2.
The average pRNFL thickness was significantly thinner in the patient group than in the control group (84.8±18.2 vs 98.7±6.7 μm, p<0.001). All GCIPL parameters, including average thickness (71.9±8.6 vs 85.1±4.8 μm, p<0.001), also differed between the patient and the control groups. Average GCIPL thickness showed a good correlation with VF MD (r2=0.342, p<0.001).
Comparison of PBG and PPBG indicated no significant differences in the average thicknesses of GCIPL (p=0.133). Comparison of four GCIPL-related parameters between PBG and PPBG patients indicated significant differences in the minimum (p=0.007), superior (p=0.031), superonasal (p=0.002) and inferonasal (p=0.003) sectors (table 3).
The AUC of the average GCIPL thickness used to discriminate between the brain lesion group and normal control group was significantly greater than that of average pRNFL thickness (0.941 vs 0.823, p<0.001, figure 1). Of the GCIPL-related parameters, the minimum thicknesses showed the greatest AUC value (0.961). Comparisons of data from PBG and PPBG patients failed to reveal differences between any GCIPL parameters or the average pRNFL thickness in terms of AUC (table 4).
Clinical examples are shown in figures 2 and 3. The patient shown in figure 2 was a 49-year-old man who had been treated for left occipital lobe haemorrhagic infarction (figure 2A). His VF showed a right homonymous hemianopsia (figure 2B) and his GCIPL thickness map showed a vertical pattern of right-sided RGC loss, in keeping with the left occipital lobe infarction (figure 2C). The patient shown in figure 3 was a 63-year-old man referred from a neurosurgery department for ophthalmic evaluation before surgical removal of pituitary macroadenoma (figure 3A). The patient had no visual symptoms, and his VF showed no abnormality by the criteria used in this study, except suggestive but uncertain temporal depression in the right eye (figure 3B). However, his GCIPL thickness was reduced in nasal sectors compared with temporal sectors in both eyes, showing a vertical pattern of nasal-sided RGC loss in the GCIPL thickness map (figure 3C). This might have been caused by compression of the optic chiasm by the pituitary macroadenoma.
Our current results confirmed that various brain lesions with different locations caused injury in RGCs. GCIPL thickness showed good relationship with VF severity. Given that those RGC injuries reduced the thickness of the GCIPL, the use of SD OCT to determine GCIPL thickness showed excellent performance in discriminating those eyes with brain lesions from normal control eyes. GCIPL thickness revealed even better performance than conventional pRNFL thickness. pRNFL thickness was developed earlier than GCIPL thickness as a parameter to assess RGC injury in OCT. Hence, the usefulness of pRNFL thickness in diagnosing TRD has been described previously.12–17 Given that the retinal nerve fibre is an axon of the RGC, GCIPL thickness might be a more direct marker for estimating RGC injury than pRNFL thickness. A case series indicated a reduction in the thickness of the macular ganglion cell layer following posterior cerebral artery territory infarction has been issued.22 However, as far as we are aware, the current study is the first that has investigated the value of using macular ganglion cell thickness to diagnose various types of brain lesions.
Some sectors that included the average GCIPL thickness did not show a significant difference between PBG patients than in PPBG patients (p=0.138). Many brain lesions were characterised by a considerable reduction in GCIPL thickness despite an apparently normal VF. AUC values also revealed no significant difference between PBG and PPBG patients in discriminating between normal subjects and patients with brain lesions. In other words, GCIPL thickness showed good diagnostic performance in discriminating normal eyes from brain lesions with no VF abnormality. These observations suggest that some brain lesions show considerable loss of RGC before manifestation of localised field abnormality recognised by our criteria. Therefore, macular ganglion cell thickness might be an earlier indicator of damage in the visual pathway than VF assessment. Additionally, patients who are unable to undergo a VF test, or for whom the results of a VF test are considered to be unreliable, might instead be evaluated by the assessment of GCIPL thickness. VF is a psychophysical test, the outcome of which depends entirely on the capacity of the examinee to concentrate adequately and perform the necessary tasks. The results of this test are thus frequently subject to intertest variability.23 ,24 Some patients with poor cognitive function might not perform the VF test sufficiently well to provide accurate results. The potential for malingering is another limitation of the VF test. In those cases where a VF test is unavailable or inaccurate, measurement of the GCIPL thickness might provide a surrogate measure for RGC status in brain lesions.
Whereas a suprasellar lesion was the most frequent pathology associated with PBG patients, cerebral infarction was more common in PPBG patients than in PBG patients. This might be explained by damage to the RGC by pituitary lesions that cause direct compression of the optic chiasm. The observation that GCIPL thickness had a significant reduction in superior and nasal quadrants than the other regions in PBG patients may be explained that suprasellar tumour was the most frequent pathology in PBG and, thus, compression of the optic chiasm may damage the superior and nasal areas more aggressively.
By contrast, cerebral infarction can cause different levels of RGC damage, depending on the extent and location of the infarction and the time between the infarction and VF assessment. Brain lesions that are minimally connected with the visual pathway might or might not show injury in VF. Given that time is needed before the onset of TRD, this might also affect the rate at which the reduction in the thickness of the GCIPL becomes apparent.17 ,25
Interestingly, baseline IOP was significantly lower in the brain lesion group than the normal control group. One possible explanation would be that some of the patients (10 patients) had been treated with IOP-lowering medication when referred from a local eye clinic with suspicion of normal tension glaucoma, which might contribute to lower baseline IOP measurements.
In a previous study, minimum GCIPL thickness was the most sensitive among GCIPL parameters in the diagnosis of glaucoma.7 Our result also revealed the greatest AUC of minimum GCIPL thickness. Thus, it would be suggested that minimum GCIPL thickness has the best capability among all GCIPL parameters in the diagnosis of RGC-related diseases.
In addition to quantitative assessment of GCIPL thickness, the pattern of RGC loss might provide diagnostic value through facilitating the localisation of brain lesions that aid the differential diagnosis of brain lesions and other pathologies, such as glaucoma. As shown in figure 3, a pituitary lesion that showed a binasal pattern of RGC loss with respect to the vertical meridian might be caused by compression of the optic chiasm. Figure 2 shows a vertical right homonymous pattern of RGC loss, which was induced by TRD following infarction of a left occipital lobe.
Our study has several limitations. Our inclusion of various kinds of brain lesions with different locations increased the likelihood of considerable variability in the degree of RGC damage. The combination of heterogeneous data might thus bias our results. The time period between MRI and SD OCT examination also showed wide range. Elapsed time from MRI to SD OCT examination is very long and very variable, with a mean of almost 3.6 years and as long as 15 years in some patients. In this period of time, other changes could have occurred. Additionally, different strategies in VF evaluation: 24–2 and 30–2 were performed. Some very peripheral lesions could have been undetected in patients with 24–2 only. Furthermore, the data were collected from the same clinic, which was based at a tertiary university hospital, and all participants belonged to the same ethnic group. Hence, our results might not be applicable to other populations.
In conclusion, our current results suggest that various kinds of brain lesions with different locations show considerable reduction of GCIPL thickness even before a VF abnormality becomes apparent. Thickness of the GCIPL performed better than conventional pRNFL thickness when attempting to discriminate patients with brain lesions from normal controls. Furthermore, a map of the GCIPL thickness showed a typical vertical pattern of location-dependent RGC loss. GCIPL thickness might, therefore, be an early and useful marker to estimate visual pathway status in patients with various brain lesions.
HM and JYY contributed equally.
Contributors HM: Study design, data collection, analysis, manuscript writing. JYY: data collection, analysis, manuscript writing. KRS: study design, data collection, analysis, manuscript writing, critical revision. HTL: critical revision.
Funding This study was supported by a grant (2013-0513) from the Asan Medical Center, Seoul, Korea.
Competing interests Kyung Rim Sung, none; Joo Young Yoon, none; Haein Moon, none; Hyun Taek Lim, none.
Ethics approval Institutional Review Board of the Asan Medical Center.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.