Full Text Article Open Access
Citation: Khallouli A, Oueslati Y, Bouchoucha S, Gouider D. Structure-Function correlation using optical coherence tomography in
normal subjects, preperimetric and manifest primary open-angle glaucoma. Jr. med. res. 2022; 5(1):3-6.
Khallouli et al © All rights are reserved. https://doi.org/10.32512/jmr.5.1.2022/3.6
Submit your manuscript: www.jmedicalresearch.com
Background
Primary open-angle glaucoma is a chronic optic neuropathy. Diagnosis and
monitoring require several functional and structural investigations.
Structure-function correlation is a capital step of the management.
The aim of this study was to assess the correlation between tomographic
and functional parameters in normal, preperimetric and manifest glaucoma
cases.
Methods
This retrospective analytical study included 275 eyes (152 cases).
Participants were divided into 3 groups: 33 normal subjects, 32 patients
with preperimetric glaucoma and 87 patients with manifest glaucoma. All
subjects underwent a complete ophthalmologic examination, a visual field
and spectral-domain optical coherence tomography (SD-OCT).
Results
Correlation between functional and tomographic parameters was non-
significant in the group of normal or preperimetric glaucoma subjects.
Regarding manifest glaucoma group, mean deviation (MD) was
significantly correlated with all tomographic parameters (p<0.001). The
loss variance (LV) was significantly correlated with tomographic
assessment of Retinal nerve fiber layer (RNFL) and ganglion cell complex
(GCC). The regression studies of (MD - RNFL /GCC) and (LV - GCC) had
significant results with nonlinear models (p <0.001). Linear and polynomial
models were used to correlate LV and average RNFL (p=0,275).
Conclusions
No structure-function correlation was observed at the preperimetric stage.
However, MD correlated with tomographic parameters more than LV in
manifest glaucoma group. Curvilinear function might be the appropriate
model for the structure-function relationship assessment.
Key words
Primary open-angle glaucoma; Retinal ganglion cells; Retinal nerve fiber;
Optical coherence tomography; Correlation.
1: Department of ophthalmology, Principal Military
Hospital, Tunis, Tunisia
2: College of Medicine, Tunis, Tunisia
* Corresponding author
Correspondence to:
weslatiyasin10@gmail.com
Publication data:
Submitted: December 22, 2021
Accepted: February 18, 2022
Online: June 30, 2022
This article was subject to full peer-review.
This is an open access article distributed under the
terms of the Creative Commons Attribution Non-
Commercial License 4.0 (CCBY-NC) allowing to
share and adapt.
Share: copy and redistribute the material in any
medium or format.
Adapt: remix, transform, and build upon the
licensed material.
the work provided must be properly cited and
cannot be used for commercial purpose.
Khallouli Asma
1,2
, Oueslati Yassin
1,2*
, Bouchoucha Saker
1,2
, Gouider Dhouha
1,2
.
Abstract
Original Article
Structure-Function correlation using optical coherence tomography in normal subjects,
preperimetric and manifest primary open-angle glaucoma.
Structure-Function correlation using optical coherence tomography in normal subjects, preperimetric and manifest primary open-angle glaucoma
Citation: Khallouli A, Oueslati Y, Bouchoucha S, Gouider D. Structure-Function correlation using optical coherence tomography in
normal subjects, preperimetric and manifest primary open-angle glaucoma. Jr. med. res. 2022; 5(1):3-6.
Khallouli et al © All rights are reserved Submit your manuscript: www.jmedicalresearch.com
Global loss volume (GLV) was computed as the sum of negative
fractional deviation in the entire area. Only OCT results with a
signal strength index "SSI"> 50 were retained. All statistical
analyses were performed with IBM® SPSS® Statistics 23.0
software and Excel Microsoft Office 2016. Distribution’s normality
for numerical data were tested by the Kolmogorov-Smirnov test.
Age-adjusted ANOVA test was adopted for the comparison
between groups, and the Tukey-Kramer honest significant
difference (HSD) post hoc test was used in order to adjust for
multiple comparisons between groups
Correlations between quantitative variables were studied by
Pearson correlation coefficient (r). Analysis of the relationship
between structural (RNFL, GCC) and functional (MD, LV)
parameters was performed with linear and nonlinear regression
models (second and third-degree polynomial regression), in order
to better understand the trend and evolutionary profile of the
disease. The regression models were evaluated with Akaike
Information Criterion (AIC); the model with the weakest Akaike
information criterion is the most suitable, adjusted R2 and extra-
sum-of square F test. The F test was used to check whether the
alternative nonlinear model matched the presented data better
than the linear model.
Results
The mean age in normal subjects group was statistically lower than
manifest glaucoma and preperimetric glaucoma group (P < 0.001).
(Table 1). Statistically significant differences were found between
the three groups for the analysis of Average C/D area ratio,
Average RNFL and Average GCC.
Table 1 : Age distribution, tomographic and perimetric parameters in different groups
Normal subjects
Preperimetric glaucoma
manifest glaucoma
P
Age
49.24 +/- 14.58
53.3+/-9.4
62.15+/-14.12
<0.001*
Average C/D area
ratio
0.33+/-0.13
0.48+/-0.15
0.52 +/-0.2
<0.001*
Average RNFL
102.26+/-8.3
94.84+/-7.9
84.03 +/- 16.2
<0.001*
Average GCC
101.48+/-9.93
94.21+/-14.4
86.13 +/- 15.52
<0.001*
MD (M±SD)
-0,43 +/- 0,5
-0,66 +/- 0,7
8,85 +/-6,69
A 0,094
B <0,01*
LV (M±SD)
1,77 +/- 0,88
2,61 +/- 1,53
20,16 +/ 17,51
A 0,523
B <0,01*
A: Comparison between normal subjects versus preperimetric glaucoma, B: Comparison between (normal subject/ preperimetric
glaucoma) versus manifest glaucoma. MD, mean deviation; LV, loss variance; RNFL, retinal nerve fiber layer; GCC, ganglion cell
complex
No significant difference was found between normal subjects and
preperimetric glaucoma group. However, the comparison between
normal subjects and preperimetric glaucoma groups on one hand
and manifest glaucoma group on the other hand, showed a
significant difference (p <0.01) for studied perimetric parameters
(MD, LV).
Correlation of MD with tomographic parameters revealed that; in
normal subjects and preperimetric glaucoma groups, no
statistically significant correlation was found between tomographic
parameters and MD. In manifest glaucoma group, MD was
significantly correlated with most tomographic parameters.
Best correlations for MD were noted with the analysis of RNFL and
GCC (highly significant <0.001).
In normal subjects and preperimetric glaucoma groups, LV did not
show a significant correlation with most of tomographic
parameters. However, in manifest glaucoma group, LV was
significantly correlated with most of parameters. Best correlations
for LV were noted with the analysis of GCC (highly significant
<0.001) (table2,3).
Tableau 2 : Les donnes de l’examen cardiaque et pulmonaire.
Les caractéristiques cliniques
n(%)
Les râles sibilants
13 (20)
Les râles crépitants bilatéraux
40 (61,5)
Les signes de lutte
27 (41,53)
Patients and Methods
This 2 years study included 275 eyes of 152.
Participants were divided into 3 groups: 66 eyes from 33 normal
subjects, 51 eyes of 32 preperimetric glaucoma patients, and 158
eyes of 87 manifest primary open-angle glaucoma (POAG)
patients.
•Normal subjects were randomly selected from a cohort of non-
glaucomatous individuals who met the inclusion criteria: Age> 40
years old, intraocular pressure <21 mmHg, normal papillae,
normal visual field (VF) and normal OCT examination.
•Inclusion criteria for manifest glaucoma group were: Open
iridocorneal angle (ICA), glaucomatous changes of optic disc,
RNFL and GCC defect on SD-OCT, consistent glaucomatous
pattern on VF examinations.
•For preperimetric glaucoma group, the criteria were: age> 40
years old, glaucomatous structural findings as stated above in
criteria for manifest glaucoma group, and normal VF results.
Exclusion criteria were as follows: history of eye surgery (except
uncomplicated phacoemulsification cataract surgery), presence of
ophthalmological or other pathologies responsible for impaired
vision or VF, narrow ICA and all other forms of secondary open-
angle glaucoma (such as post-traumatic or post-uveitis).
Standard VF testing was performed using automated standard
white-on-white perimetry (Octopus 101 Haag-Streit USA, Inc;
normal test strategy, program 24-2). The test was considered
reliable when fixation losses were less than 20% and false positive
or false negative errors were less than 15%. Each VF defect was
confirmed in at least two VF tests.
The mean value of VF sensitivity was calculated by a software and
presented as mean deviation (MD) and Loss variance (LV); all
studied parameters were expressed in decibels (dB).
Optic nerve head parameters, GCC and RNFL thicknesses were
measured by SD-OCT RTVue-100 (Optovue, Inc., Fremont, CA,
USA). RNFL thickness was determined by ONH mode, in which
data along a 3.45-mm diameter circle around the optic disc was
recalculated with a map created from en-face imaging that used
6 circular and 12 linear data inputs.
Average, superior, and inferior RNFL thicknesses were calculated.
In addition, the SD-OCT RTVue 100 provide a sectoral analysis of
the RNFL according to the following sectors: Superior temporal
quadrant: (90 ° -180 °) RNFL st, inferior temporal quadrant: (180°
-270 °) RNFL it, inferior nasal quadrant:
(270 ° -360 °) RNFL in, superior nasal quadrant: (360 -90 °) RNFL
sn.
The GCC scan was centered 1-mm temporal to the fovea and
covered a square grid (7 × 7 mm) on the central macula. GCC
thickness was measured from the internal limiting membrane to
the outer inner plexiform layer boundary, and average, superior,
and inferior GCC thicknesses were calculated.
Two pattern-based diagnostic parameters were also obtained.
Focal loss volume (FLV) was computed as the integral of deviation
in areas of significant focal GCC loss divided by the map area.
4
Structure-Function correlation using optical coherence tomography in normal subjects, preperimetric and manifest primary open-angle glaucoma
Citation: Khallouli A, Oueslati Y, Bouchoucha S, Gouider D. Structure-Function correlation using optical coherence tomography in
normal subjects, preperimetric and manifest primary open-angle glaucoma. Jr. med. res. 2022; 5(1):3-6.
Khallouli et al © All rights are reserved Submit your manuscript: www.jmedicalresearch.com
t w
The relationships between visual field and SD-OCT parameters were
evaluated by regression analysis. (Figures)
The relation between MD (dB) and average RNFL / GCC was better
explained with a second-order polynomial model (R2 = 0.289, AIC
= 550.06 and p <0.001 for average RNFL) and (R2 = 0.203, AIC =
567.95 and p <0.001 for average GCC).
Regression study between LV and RNFL did not show a statically
significant difference between linear and non-linear model.
The regression between LV and average GCC was also better
expressed by a second-order polynomial model (R2 = 0.1, AIC =
891 and p= 0.034) (Table 4).
Table 2: Correlation OCT parameters /MD.
Normal
Preperimetric glaucoma
Manifest glaucoma
Cup/Disc Area Ratio
r=0.75 p=0.548
r=0.036 p=0.8
r=0.52 p<0.001
VCDR
r=0.41 p=0.74
r=0.007 p=0.96
r=0.46 p<0.001
HCDR
r=0.077 p=0.54
r=-0.17 p=0.23
r=0.2 p=0.015
Rim Area
r=0.006 p=0.96
r=-0.052 p=0.71
r=-0.17 p=0.035
Disc Area
r=0.91 p=0.468
r=0.07 p=0.63
r=0.11 p=0.153
Cup Volume
r=-0.18 p=0.884
r=0.18 p=0.21
r=0.38 p<0.001
Average RNFL
r=0.188 p=0.131
r=-0.136 p=0.34
r=-0.47 p<0.001
Superior hemisphere RNFL
r=0.2 p=0.107
r=-0.16 p=0.26
r=-0.43 p<0.001
Inferior hemisphere RNFL
r=0.15 p=0.23
r=-0.09 p=0.53
r=-0.48 p<0.001
Superior temporal RNFL
r=0.2 p=0.1
r=-0.136 p=0.34
r=-0.34 p<0.001
Inferior temporal RNFL
r=0.1 p=0.43
r=-0.14 p=0.3
r=-0.32 p<0.001
Superior nasal RNFL
r=0.15 p=0.21
r=-0.18 p=0.2
r=-0.44 p<0.001
Inferior nasal RNFL
r=0.12 p=0.33
r=0.123 p=0.4
r=-0.51 p<0.001
Average GCC
r<0.001 p=0.99
r=0.095 p=0.51
r=-0.42 p<0.001
Superior hemisphere GCC
r=0.08 p=0.5
r=0.084 p=0.56
r=-0.4 p<0.001
Inferior hemisphere GCC
r=-0.1 p=0.44
r=0.095 p=0.51
r=-0.42 p<0.001
FLV %
r=-0.06 p=0.6
r=-0.08 p=0.54
r=0.5 p<0.001
GLV %
r=-0.07 p=0.57
r=-0.07 p=0.62
r=0.51 p<0.001
Table 3: Correlation OCT parameters/ LV.
Normal subjects
Preperimetric glaucoma
Manifest glaucoma
Cup/Disc Area Ratio
r=0.18 p=0.14
r=-0.27 p=0.05
r=0.3 p<0.001
VCDR
r=0.14 p=0.24
r=-0.18 p=0.2
r=0.27 p=0.001
HCDR
r=0.23 p=0.06
r=-0.17 p=0.22
r=-0.85 p=0.3
Rim Area
r=-0.18 p=0.14
r=0.006 p=0.96
r=-0.03 p=0.7
Disc Area
r=-0.009 p=0.94
r=-0.32 p=0.02
r=-0.08 p=0.27
Cup Volume
r=0.05 p=0.686
r=-0.25 p=0.07
r=0.09 p=0.24
Average RNFL
r=-0.07 p=0.542
r=0.174 p=0.22
r=-0.25 p=0.001
Superior hemisphere RNFL
r=-0.05 p=0.67
r=0.08 p=0.54
r=-0.22 p=0.005
Inferior hemisphere RNFL
r=-0.06 p=0.61
r=0.22 p=0.12
r=-0.26 p=0.001
Superior temporal RNFL
r=-0.07 p=0.55
r=0.11 p=0.43
r=-0.16 p=0.04
Inferior temporal RNFL
r=0.09 p=0.44
r=0.33 p=0.017
r=-0.13 p=0.11
Superior nasal RNFL
r=-0.02 p=0.88
r=0.078 p=0.58
r=-0.24 p=0.003
Inferior nasal RNFL
r=-0.16 p=0.19
r=0.013 p=0.93
r=-0.26 p=0.001
Average GCC
r=0.12 p=0.33
r=-0.05 p=0.74
r=-0.3 p<0.001
Superior hemisphere GCC
r=0.06 p=0.64
r=-0.03 p=0.81
r=-0.27 p<0.001
Inferior hemisphere GCC
r=-0.15 p=0.21
r=-0.61 p=0.67
r=-0.31 p<0.001
FLV %
r=-0.25 p=0.04
r=-0.063 p=0.66
r=0.28 p<0.001
GLV %
r=-0.31 p=0.01
r=-0.07 p=0.61
r=0.33 p<0.001
MD, mean deviation; HCDR, horizontal cup-to-disc ratio; VCDR, vertical cup-to-disc ratio; RNFL, retinal nerve fiber layer; GCC, ganglion
cell complex; FLV, Focal loss volume; GLV, Global loss volume.
(A) Mean Deviation (MD) versus average ganglion cell complex (GCC), linear and nonlinear
regression (second-order and third-order polynomial).
(B) Mean Deviation (MD) versus average retinal nerve fiber layer (RNFL), linear and
nonlinear regression (second-order and third-order polynomial).
(C) Loss variance (LV) versus average retinal nerve fiber layer (RNFL), linear and nonlinear
regression (second-order and third-order polynomial).
(D) Loss variance (LV) versus average ganglion cell complex (GCC), linear and nonlinear
regression (second-order and third-order polynomial)
5
Structure-Function correlation using optical coherence tomography in normal subjects, preperimetric and manifest primary open-angle glaucoma
Citation: Khallouli A, Oueslati Y, Bouchoucha S, Gouider D. Structure-Function correlation using optical coherence tomography in
normal subjects, preperimetric and manifest primary open-angle glaucoma. Jr. med. res. 2022; 5(1):3-6.
Khallouli et al © All rights are reserved Submit your manuscript: www.jmedicalresearch.com
Table 4: Prediction of MD and LV from OCT parameters( regression analysis).
Linear
Second-Order Polynomial
Third-Order Polynomial
R
2
AIC
R
2
AIC
F
P*
R
2
AIC
F
P†
MD -RNFL
0,219
563,77
0,289
550,06
16,21
<0,001
0,284
552,05
8,05
<0,001
MD - GCC
0,176
838,86
0,203
567,95
716,92
<0,001
0,2
569,85
356,43
<0,001
LV - RNFL
0,056
897,38
0,058
898,17
1,2
0,275
0,058
899,06
1,143
0,321
LV -GCC
0,084
833,7
0,1
891,13
4,54
0,034
0,094
892,96
2,34
0,1
P*; comparison of linear and second-order polynomial model, P†; comparison of linear and third-order polynomial model, RNFL, retinal nerve
fiber layer; GCC, ganglion cell complex, AIC; Akaike Information Criterion, R2 ; regression coefficient, F; extra-sum-of square F test.
Discussion
No correlation between Mean Deviation (MD) and tomographic parameters
in normal and preperimetric glaucoma groups could be found due to the
normal visual field [2]. Analysis of glaucomatous eyes from eye bank
showed that at least 25 to 35% loss of retinal ganglion cells (RGC) is
required to detect the first functional alterations in visual field. In early
stages, the neuronal loss is undetectable in automated perimetry [3].
In manifest glaucoma group, MD was significantly correlated with all
parameters exploring RNFL, GCC and most of the Optic nerve head(ONH)
parameters. Similar results have been reported in the literature [4,5].
However, objective correlation might be still questionable due to the
heterogeneity of the published series and the difficulty of randomized trial
establishment [6].
Regarding loss of variance, no significant correlations were found in
normal subjects and preperimetric glaucoma groups. However, for
manifest glaucoma group, the correlation was significant between LV and
most tomographic parameters assessing RNFL and GCC. Best correlations
were noted with parameters analyzing GCC. In the literature as in our work,
correlations of structural parameters were more significant with MD than
LV which represent more the variance of the local default [7-9]. LV index
would be considerable in case of inhomogeneous localized deficit of retinal
sensitivity. The correlation of LV with the ONH parameters was absent in
manifest glaucoma group, except; Cup/Disc area ratio and Cup/Disc V
ratio. This could be explained by the fact that, the overall severity of the
disease in manifest glaucoma group was moderate with an average MD =
8,85 +/- 6,69. In fact, at early and moderate stage, neuronal loss
particularly affects the upper and lower part of the optic disc responsible
for the vertical elongation of the optic cup.
The study of linear and nonlinear regression models was mandatory to
assess structure-function relationship. Regression functions would then be
useful in order to understand the trend of disease progression and to select
an appropriate monitoring strategy [10].
The structure-function relationship was better explained with nonlinear
models (second and third-order model) evaluating MD (dB) with RNFL and
GCC. However, linear models describe more the relationship between LV
and average RNFL. The regression between LV and average GCC, was
better explained by second-order polynomial model, but the difference was
non-significant compared to linear model (p = 0.034). Similar results were
seen in other studies adopting similar statistical methodology [11].
Logarithmic scale could minimize changes in retinal sensitivity at large dB
values and optimize it at low dB values. Simple linear model could be an
alternative of the curvilinear model to describe the relationship between
perimetric and tomographic parameters in some glaucoma cases. However,
it may transform MD into a non-logarithmic function and generate a biased
curvilinear aspect for the scales expressed in decibel [12].
In the literature, curvilinear regression models, evaluating the relationship
between RNFL and VF sensitivity, were the most effective. Likewise, the
second and third order regression models correlating GCC and VF
sensitivity, showed a stronger structure-function correlation compared to
the first order regression models. The curvilinear regression models
suggest that structural functional changes are not simultaneous in cases
of Glaucoma. Non-axonal components thickness of RNFL
increase with age and disease progression. Glial cells activation
allows glial remodeling of the retinal nerve fiber layer during
the response to neuronal injury. This gliosis may be hiding the
decrease in RNFL and GCC thickness [13].
The functional impotence may be delayed in some advanced
structural alterations. That would explain the concept of
‘functional reserve’ and the difficulties of the assessment in
some cases. for curvilinear model, the correlation between VF
sensitivity (on a logarithmic scale) and structural parameters
(on a linear scale) seems to be more effective for early stage
functional impairment.
Monitoring the glaucomatous neuropathy progression consists
first in detecting the evolution of the retinal nerve fiber layer
damages, then that of the macular ganglion cell complex. In
order to assess damage progression to the targeted structures,
the analysis should always be integrated with the measurement
of the functional impairment [14].
Some type of regression curve minimizes visual field changes
in early-stage glaucoma and maximizes them in severely
affected glaucoma. It is recommended to rely more on
structural investigations such as OCT for the monitoring of
preperimetric and early glaucoma cases. Functional tests such
as conventional static perimetry are more effective in the
follow-up of advanced stage glaucoma.
Conclusions
Glaucoma is one of the leading causes of blindness worldwide
and Optical Coherence Tomography (OCT) is the cornerstone
imaging technique for its detection. Early detection of
structural alterations and the prediction of related visual
impairment was the challenging part of most of the published
analytic studies.
Conflit of interest: None
References
[1] Kim NR, Lee ES, Seong GJ, Kim JH, An HG, Kim CY. Structure-function relationship and diagnostic
value of macular ganglion cell complex measurement using Fourier-domain OCT in glaucoma. Invest
Ophthalmol Vis Sci. 2010; 51:4646-51.
[2] Ajtony C, Balla Z, Somoskeoy S, Kovacs B. Relationship between visual field sensitivity and retinal
nerve fiber layer thickness as measured by optical coherence tomography. Invest Ophthalmol Vis
Sci. 2007; 48:258‑63.
[3] Kerrigan-Baumrind LA, Quigley HA, Pease ME, Kerrigan DF, Mitchell RS. Number of ganglion cells
in glaucoma eyes compared with threshold visual field tests in the same persons. Invest Ophthalmol
Vis Sci. 2000; 41:741-8.
[4] Kostianeva SS, Konareva-Kostianeva MI, Atanassov MA. Relationship between visual field
changes and optical coherence tomography measurements in advanced open-angle glaucoma. Folia
Med (Plovdiv). 2016; 58:174-81.
[5] Takahashi M, Omodaka K, Maruyama K, Yamaguchi T, Himori N, Shiga Y, et al. Simulated visual
fields produced from macular RNFLT data in patients with glaucoma.Curr Eye Res.2013;38:1133-41.
[6] Lleó-Pérez A, Ortuño-Soto A, Rahhal MS, Sanchis-Gimeno JA. Relationship Between Visual Field
Sensitivity and Retinal Nerve Fiber Layer Thickness Measured by Scanning Laser Polarimetry and
Optical Coherence Tomography in Normal, Ocular Hypertensive and Glaucomatous Eyes. J Optom.
2009; 2:39-50.
[7] Tole DM, Edwards MP, Davey KG, Menage MJ. The correlation of the visual field with scanning
laser ophthalmoscope measurements in glaucoma. Eye (Lond). 1998;12 (Pt4):686-90.
[8] Lan YW, Henson DB, Kwartz AJ. The correlation between optic nerve head topographic
measurements, peripapillary nerve fibre layer thickness, and visual field indices in glaucoma. Br J
Ophthalmol. 2003; 87:1135-41.
[9] Iester M, Mikelberg FS, Courtright P, Drance SM. Correlation between the visual field indices and
Heidelberg retina tomograph parameters. J Glaucoma. 1997; 6:78-82.
[10] Jonas JB, Bergua A, Schmitz-Valckenberg P, Papastathopoulos KI, Budde WM. Ranking of optic
disc variables for detection of glaucomatous optic nerve damage. Invest Ophthalmol Vis Sci. 2000;
41:1764‑73.
[11] Leung CK, Chong KK-L, Chan W, Yiu CK-F, Tso M, Woo J, et al. Comparative Study of Retinal
Nerve Fiber Layer Measurement by StratusOCT and GDx VCC, II: Structure/Function Regression
Analysis in Glaucoma. Invest Ophthalmol Vis Sci. 2005;46:3702‑11.
[12] Garway-Heath DF, Caprioli J, Fitzke FW, Hitchings RA. Scaling the hill of vision: the physiological
relationship between light sensitivity and ganglion cell numbers. Invest Ophthalmol Vis Sci. 2000;
41:1774‑82.
[13] Wheat JL, Rangaswamy NV, Harwerth RS. Correlating RNFL Thickness by OCT with Perimetric
Sensitivity in Glaucoma Patients. J Glaucoma. 2012;21:95‑101.
[14] Wang L, Cioffi GA, Cull G, Dong J, Fortune B. Immunohistologic evidence for retinal glial cell
changes in human glaucoma. Invest Ophthalmol Vis Sci.2002;43:1088‑94.
6