Morphological and Topographical Appearance of Microaneurysms On Optical Coherence Tomography Angiography
Morphological and Topographical Appearance of Microaneurysms On Optical Coherence Tomography Angiography
Morphological and Topographical Appearance of Microaneurysms On Optical Coherence Tomography Angiography
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
Morphological and topographical appearance of
microaneurysms on optical coherence tomography
angiography
Vivian Schreur,1 Artin Domanian,1 Bart Liefers,1,2 Freerk G Venhuizen,1,2
B Jeroen Klevering,1 Carel B Hoyng,1 Eiko K de Jong,1 Thomas Theelen1
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
Materials and methods Germany). FFA was acquired after intravenous administration
Patients of 2.5 mL of 2.5% fluorescein solution, and images for analysis
We conducted an observational case series study on 52 eyes of 33 were selected from the early (30 s to 2 min) and late phase (5–15
patients with clinically significant DME that visited the outpa- min) of the dye transit.
tient clinic of the ophthalmology department of the Radboud
University Medical Center, Nijmegen, The Netherlands, Image analysis
between November 2015 and June 2016. This study adhered All OCTA and FFA images were exported to custom image anal-
to the tenets of Helsinki, and ethical approval was waived by ysis software created with Mevislab (MeVis Medical Solutions
the Research Ethics Committee of Radboud University Medical AG, Fraunhofer MEVIS, Germany). FFA images were resized and
Center, Nijmegen. Patients were included if FFA and OCTA matched to the 3×3 mm OCTA image slabs for pixel-wise anal-
were performed on the same day in the context of regular clin- ysis. Two independent experienced graders (AD, VS) evaluated all
ical practice. Exclusion criteria were presence of retinal vascular images. Microaneurysms were defined as hyperfluorescent dots
pathology other than DME or poor quality images due to severe on FFA. A synchronised cursor assisted in detecting the according
motion or blinking artefacts. Demographic and clinical char- locations on OCTA to analyse flow signals on the same location
acteristics were collected by reviewing patients’ medical files. in all retinal layers. All locations were checked throughout the
Diabetic retinopathy was staged according to the International total thickness of the retina to avoid false negative results by
Clinical Diabetic Retinopathy Severity Scale.18 segmentation errors. Microaneurysms non-visible on OCTA due
to blood flow below the detection threshold were classified as
Image acquisition and processing low-flow microaneurysms, in contrast to their high-flow coun-
3×3 mm OCTA images were obtained using a swept source terparts visible on OCTA. In case microaneurysms were detected
OCTA device (DRI Triton OCT, Topcon, Tokyo, Japan). The in more than one retinal capillary plexus, we used the according
images were manually segmented into three retinal capillary high-flow signal in the uppermost layer where the signal was
plexi using IMAGEnet (Topcon) processing software.12 For the detected for analysis to account for projection artefacts.
superficial capillary plexus (SCP), the boundaries were set at According to Dubow et al, we classified the shape of the
the inner border of the inner limiting membrane to the super- microaneurysms on OCTA into either saccular, fusiform, mixed
ficial portion of the inner plexiform layer (IPL). The borders saccular/fusiform, pedunculated, irregular or focal bulging.17
of the intermediate capillary plexus (ICP) were set between the We furthermore graded the microaneurysms for the presence of
deep portion of the IPL and the superficial portion of the inner focal leakage on the late phase FFA. Focal leakage was defined
nuclear layer (INL). The boundaries of the deep capillary plexus as the presence of one or several microaneurysms within a
(DCP) were set between the deep portion of the INL and the surrounding area of fluorescein leak with fluorescein signal
outer boundary of the outer plexiform layer (figure 1). intensity decreasing to the periphery of the microaneurysm in
Spectral-domain OCT and FFA were obtained using the Spec- the centre.19 Discrepancies between graders were solved by open
tralis HRA+OCT device (Heidelberg Engineering, Heidelberg, adjudication.
Figure 1 Illustration of the investigated imaging characteristics of microaneurysms in a patient with diabetic maculopathy. Optical coherence
tomography angiography of the (A) superficial capillary plexus (SCP); (B) intermediate capillary plexus (ICP); (C) deep capillary plexus (DCP); (D)
SCP with superimposed the retinal thickness map. Fluorescein angiography of the (E) early phase; (F) late phase. (G) infrared image with the green
arrow indicating the cross section of the optical coherence tomography B-scan. (H) Optical coherence tomography of the foveal centred B-scan,
showing the segmentation boundaries of the SCP, ICP and the DCP. White circles indicate leaking microaneurysms; blue circles indicate non-leaking
microaneurysms (A–C), as determined by late phase fluorescein angiography; solid circles indicate high-flow microaneurysms; dashed circles indicate
low-flow microaneurysms that were non-detectable on optical coherence tomography angiography (E, F).
2 Schreur V, et al. Br J Ophthalmol 2018;0:1–6. doi:10.1136/bjophthalmol-2018-312258
Clinical science
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
retinal thickening. The threshold for increased retinal thickening
Table 1 Demographic and clinical characteristics of study
was ≥325 µm in the central subfield of the ETDRS grid centred
participants
to the fovea, and ≥375 µm in the parafoveal and perifoveal
Variable (n=31 eyes of 24 subjects) areas.
Gender, n (%)
Male 14 (58%) Statistical analysis
Female 10 (42%) Intergrader agreement was assessed using the intraclass correla-
Age in years, mean (SD) 57 (13) tion coefficient (ICC) for the number of microaneurysms per
Diabetes type, n (%) eye. We used generalised estimating equations binary logistic
Type 1 8 (33%) regression analysis to evaluate the influence of leakage and thick-
Type 2 16 (67%) ening status on the visibility of microaneurysms on OCTA, using
Diabetic retinopathy stage, n (%) ‘no leakage and no thickening’ as reference category. The same
Mild NPDR 5 (16%) analysis was used to study the influence of different microaneu-
Moderate NPDR 16 (52%) rysm appearances on the presence of focal leakage or increased
Severe NPDR 5 (16%)
thickening, indicating the category with the lowest predictive
Proliferative DR 5 (16%)
probability as reference category. Results are presented as OR
with corresponding 95% CI. P values<0.05 were considered
Duration DM in years, mean (SD) 24 (12)
statistically significant. Statistical analysis was performed using
Central retinal thickness in µm, median (IQR) 335 (304–351)
SPSS V.22.
Oedema type, n (%)
Cystoid 25 (81%)
Results
Diffuse 6 (19%)
Patients
Prior treatment, n (%)*
In total, 52 eyes of 33 patients were diagnosed with clinically
Intravitreal anti-VEGF 19 (61%)
significant DME and underwent simultaneous FFA and OCTA.
Intravitreal corticosteroids 10 (32%)
Of those, we excluded 18 eyes of 12 patients due to poor OCTA
Focal photocoagulation 20 (65%)
quality, and 3 eyes of 2 patients due to poor FFA quality. Even-
Micropulse photocoagulation 14 (45%) tually, 31 eyes of 24 patients were eligible for evaluation in our
Panretinal photocoagulation 20 (65%) study. Demographic and clinical features of study patients are
*Because some patients received multiple treatments, the percentages do not add displayed in table 1.
up to 100%.
DR, diabetic retinopathy; DM, diabetes mellitus; NPDR, non-proliferative diabetic
retinopathy; VEGF, vascular endothelial growth factor.
Microaneurysms
The ICCs for the number of microaneurysms on FFA and the
SCP, ICP and DCP of OCTA were 0.911, 0.974, 0.774 and
A retinal thickness map derived from the Spectralis HRA+OCT 0.765, respectively. Of all 513 microaneurysms flagged on FFA,
was superimposed and aligned onto the FFA and OCTA images 295 (58%) also showed an increased flow signal on OCTA. Most
according to retinal landmarks using Adobe Photoshop (cs5 increased flow signals associated with microaneurysms were
Adobe Systems, San Jose, California, USA). The distribution of located in the ICP (117 out of 513, 23%), closely followed by
low-flow and high-flow microaneurysms was studied within the the DCP (111 out of 513, 22%). The SCP contained the least
different categories mentioned above, based on local leakage and amount of increased flow signals (66 out of 513, 13%).
#
##
"$
"$
!
!
!
!
Figure 2 Distribution of (A) high-flow and low-flow microaneurysms across different categories according to leakage and thickness information
in patients with diabetic maculopathy; (B) leaking and not-leaking microaneurysms according to their location within the plexi. * p<0.05. DCP, deep
capillary plexus; ICP, intermediate capillary plexus; SCP, superficial capillary plexus,
Schreur V, et al. Br J Ophthalmol 2018;0:1–6. doi:10.1136/bjophthalmol-2018-312258 3
Clinical science
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
showing focal leakage ranged from 33% in the SCP to 36% in
the ICP and 41% in the DCP (figure 2B).
Focal bulging, saccular, fusiform, mixed saccular/fusiform,
pedunculated and irregular microaneurysms were all observed on
OCTA (figure 3). Saccular microaneurysms were detected most
often (89 out of 291, 31%), while pedunculated microaneurysms
were detected least often (26 out of 291, 9%, figure 4A). The
morphological appearances of microaneurysms were distributed
across all three capillary plexi (figure 4B). All types of microan-
eurysms were present throughout all stages of DR (online
supplementary table 1). Irregular, fusiform and mixed fusiform/
saccular-shaped microaneurysms were more often showing focal
leakage than focal bulging ones (OR 7.6, 95% CI 2.4 to 23.8,
p<0.001; OR 5.3, 95% CI 1.6 to 17.6, p=0.006; and OR 4.1,
95% CI 1.9 to 9.1, p<0.001, respectively, figure 4C,D C,D).
There was no statistical significant difference in the presence
of retinal thickening across the six morphological groups. The
position of microaneurysms within the different capillary plexi
did not differ among the various morphological features (data
not shown).
Discussion
We investigated retinal microaneurysms in DME according
to their location and appearance on OCTA in correlation to
their clinical properties, leakage on FFA and retinal thickening
on structural OCT. OCTA detected less microaneurysms than
conventional FFA, but did provide information on the loca-
tion of the microaneurysms in the retinal layers as well as their
morphological appearances that formerly could only be appreci-
ated using adaptive optics FFA.17 In addition, we could correlate
the morphology of the microaneurysms observed on OCTA to
their clinical properties, like leakage and retinal thickening.
Of all microaneurysms detected on FFA, increased flow signals
associated with microaneurysms were found in 58%. This is in
the same range as the previously reported frequency of 41% by
Miwa et al, the 62% reported by Couturier et al and the 64%
reported by Salz et al.14–16 The variance in detected microaneu-
rysms by OCTA is, at least in part, based on divergent grading
definitions for microaneurysms in both FFA and OCTA. More-
over, these studies might not be directly comparable as both
swept-source and spectral-domain OCT systems were employed.
Figure 3 Morphology of microaneurysms detected by optical We hypothesised that microaneurysms showing no signal on
coherence tomography angiography in patients with diabetic OCTA did not meet the minimal blood flow velocity in order to
maculopathy. The detected appearances include (first row) focal bulging; gain a positive flow signal. Besides slow flow, some microaneu-
(second row) saccular; (third row) fusiform; (fourth row) mixed saccular/ rysms may not contain (moving) erythrocytes at all, as demon-
fusiform; (fifth row) pedunculated; (sixth row) irregular.Orientation of strated in a histopathologic study reporting occlusion of the
segmentation was determined by the location of the microaneurysm. vessel lumen.20 Furthermore, not all hyperfluorescent dots on
Optical cuts were performed (left column) superior, (middle column) FFA may represent true microaneurysms. Fluorescein may cause
central and (right column) inferior tangentially. extensive staining of capillary walls or dye leakage from an
impaired vessel, mimicking a microaneurysm, but not causing an
increased flow signal on OCTA.13 16
The majority of previous OCTA studies only assessed the SCP
All microaneurysms were characterised according to leakage
and DCP, derived from preset commercial image analysis soft-
and flow signal properties and projected on a topographical ware, dividing the ICP across those two plexi. We argue that
retinal thickness map (figure 1). Focal leakage was present in one should take the structural and the developmental differ-
154 out of 513 (30%) microaneurysms visible on FFA and 200 ences of three distinct retinal capillary plexi into account in
out of 513 (39%) microaneurysms were located within a thick- the analysis of macular diabetic vascular changes as this best
ened area. High-flow microaneurysms did not necessarily leak, represents the anatomical organisation of the retinal vascular
nor were they all located in a thickened area. Microaneurysms network. However, distinction of three capillary plexi is chal-
that showed focal leakage and were located in a thickened area, lenged by the presence of projection artefacts, more so than
were more often detected on OCTA than not leaking microan- when only identifying two plexi.12 To account for projection
eurysms in non-thickened retinal areas (OR 2.6, 95% CI 1.4 to artefacts, we classified microaneurysms into the most superficial
4.6, p=0.002, figure 2A). The percentage of microaneurysms layer in which they were detected, although this method might
4 Schreur V, et al. Br J Ophthalmol 2018;0:1–6. doi:10.1136/bjophthalmol-2018-312258
Clinical science
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
Figure 4 Distribution of microaneurysm types on optical coherence tomography angiography (OCTA) in patients with diabetic maculopathy. (A)
Total number of microaneurysms throughout the retina; (B) number of microaneurysms according to their intraretin allocation on OCTA. Properties of
microaneurysms in terms of leakage and retinal thickening in patients with diabetic maculopathy. (C) Percentage of microaneurysm types according
to leakage; (D) percentage of microaneurysm types according to retinal thickening.*p<0.05; **p<0.001.
result in an underestimation of microaneurysms in the deeper in the diagnostic workup of DME as these microaneurysms may
plexi. Therefore, to avoid overestimation or underestimation effectively be treated by focal laser treatment. Although the
of the presence of microaneurysms, one should carefully slide intravitreal administration of steroids or anti-vascular endo-
through the complete OCTA volume in advance of preset OCTA thelial growth factor (VEGF) agents has replaced primary laser
slabs. By this, confounders like local capillary tortuosity (pseu- treatment in patients with central macular oedema, this infor-
do-microaneurysm if observed tangentially in a single slab) and mation is still useful in cases with persisting oedema, if despite
microaneurysms displaced to an underlying layer (false negative anti-VEGF additional laser treatment is indicated.23 Currently,
if single slabs above are studied) can be detected, resulting in laser treatment is often guided by FFA, but may be replaced by
more reliable study outcomes. Future improvements in anatomic a non-invasive imaging technique such as OCTA if it enables
segmentation software may reduce the number of false positives identification of microaneurysms that are considered treatment
and negatives. targets. Our present study shows that for this purpose FFA
We found microaneurysms to be distributed throughout all cannot be replaced by OCTA as not all leaking microaneurysms
retinal capillary plexi, although most microaneurysms were were detected by OCTA. Therefore, future research should be
located in the ICP and DCP. This corroborates findings from directed towards improving OCTA techniques and processing
histopathological studies localising microaneurysms mainly in the software.
INL, which contains the lower portion of the ICP and the upper As with adaptive optics FFA, we observed different shapes
portion of the DCP.21 Therefore, we hypothesise that when using of microaneurysms by OCTA and our results corroborated
the preset segmentation method most microaneurysms will be the frequency of the single subtypes, despite using a different
detected in the DCP as it contains the entire INL. Recent retinal imaging technique.17 In addition to adaptive optics FFA, we
blood flow studies suggest a serial perfusion of retinal capillaries, were able to localise the microaneurysms three-dimensional in
with direct flow from the large retinal vessels to ICP and DCP, the retina and correlate these morphological subtypes with focal
and SCP.22 Hence, blood flow velocity in general may be higher leakage and retinal thickening. Irregular, fusiform and mixed-
in the ICP and DCP, and consequently, in the microaneurysms shaped microaneurysms tended to leak more often than other
located there, facilitating detection by the OCTA. types of microaneurysms. Irregular microaneurysms may have a
Microaneurysms showing focal leakage on FFA and located greater luminal surface area that may increase the risk of damage
in a thickened area on OCT were more often high flow than of the basement membrane and thus breakdown of the inner
not leaking microaneurysms that were not located in a thickened blood–retina barrier.Wang et al, on the other hand, demon-
area. However, high flow in a microaneurysm did not necessarily strated that size of the aneurysms on FFA does not necessarily
mean leakage activity, neither did low flow mean the opposite. correlate with FFA leakage status.17 24 Therefore, we argue that
Leakage of microaneurysms on FFA is an important observation the morphology of retinal microaneurysms is a clinically more
Schreur V, et al. Br J Ophthalmol 2018;0:1–6. doi:10.1136/bjophthalmol-2018-312258 5
Clinical science
Br J Ophthalmol: first published as 10.1136/bjophthalmol-2018-312258 on 20 June 2018. Downloaded from http://bjo.bmj.com/ on 21 June 2018 by guest. Protected by copyright.
important biomarker than size and should be considered in 5 Photocoagulation for diabetic macular edema. Early Treatment Diabetic Retinopathy
upcoming studies. Study report number 1. Early Treatment Diabetic Retinopathy Study research group.
Arch Ophthalmol 1985;103:1796–806.
The small study population is a limitation of our current 6 Snodderly DM, Weinhaus RS, Choi JC. Neural-vascular relationships in central retina of
study. Furthermore, we included patients with various stages macaque monkeys (Macaca fascicularis). J Neurosci 1992;12:1169–93.
of DME, regardless of whether or not treatment was adminis- 7 Weinhaus RS, Burke JM, Delori FC, et al. Comparison of fluorescein angiography with
tered, resulting in a heterogeneous study population. Further microvascular anatomy of macaque retinas. Exp Eye Res 1995;61:1–16.
8 Mendis KR, Balaratnasingam C, Yu P, et al. Correlation of histologic and clinical
research is warranted in a large, prospective and preferably treat- images to determine the diagnostic value of fluorescein angiography for studying
ment-naive cohort, to replicate our findings, and to evaluate the retinal capillary detail. Invest Ophthalmol Vis Sci 2010;51:5864–9.
clinical relevance. In addition, a consensus is needed for the defi- 9 Yannuzzi LA, Rohrer KT, Tindel LJ, et al. Fluorescein angiography complication survey.
nition of microaneurysms in FFA as well as OCTA to increase the Ophthalmology 1986;93:611–7.
10 Kwan AS, Barry C, McAllister IL, et al. Fluorescein angiography and adverse drug
reliability of the ground truth in future studies. reactions revisited: the Lions Eye experience. Clin Exp Ophthalmol 2006;34:33–8.
In conclusion, we have described the morphology and three-di- 11 Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with
mensional distribution of diabetic retinal microaneurysms optical coherence tomography. Opt Express 2012;20:4710–25.
in relation to their clinical properties in terms of leakage and 12 Campbell JP, Zhang M, Hwang TS, et al. Detailed vascular anatomy of the human
retina by projection-resolved optical coherence tomography angiography. Sci Rep
retinal thickening. Defining the morphology of microaneurysms 2017;7:42201.
on OCTA increases our understanding of pathophysiological 13 Ishibazawa A, Nagaoka T, Takahashi A, et al. Optical coherence tomography
mechanisms of DME and may aid in finding the optimal thera- angiography in diabetic retinopathy: a prospective pilot study. Am J Ophthalmol
peutic approach. Future research should be directed towards the 2015;160:35–44.
14 Salz DA, de Carlo TE, Adhi M, et al. Select features of diabetic retinopathy on
clinical relevance of the various morphological manifestations as swept-source optical coherence tomographic angiography compared with fluorescein
biomarkers for disease progression or treatment response. angiography and normal eyes. JAMA Ophthalmol 2016;134:644–50.
15 Couturier A, Mané V, Bonnin S, et al. Capillary plexus anomalies in diabetic
Contributors VS and TT designed the study, VS and AD carried out the retinopathy on optical coherence tomography angiography. Retina 2015;35:2384–91.
experiments, BL and FGV developed custom image analysis software, BJK, CBH, EKJ 16 Miwa Y, Murakami T, Suzuma K, et al. Relationship between functional and structural
and TT contributed to the interpretation of the results. VS took the lead in writing the changes in diabetic vessels in optical coherence tomography angiography. Sci Rep
manuscript with input from all authors. 2016;6:29064.
17 Dubow M, Pinhas A, Shah N, et al. Classification of human retinal microaneurysms
Funding The authors have not declared a specific grant for this research from any using adaptive optics scanning light ophthalmoscope fluorescein angiography. Invest
funding agency in the public, commercial or not-for-profit sectors. Ophthalmol Vis Sci 2014;55:1299–309.
Competing interests None declared. 18 Wilkinson CP, Ferris FL, Klein RE, et al. Proposed international clinical diabetic
retinopathy and diabetic macular edema disease severity scales. Ophthalmology
Patient consent Not required. 2003;110:1677–82.
Provenance and peer review Not commissioned; externally peer reviewed. 19 Classification of diabetic retinopathy from fluorescein angiograms. ETDRS report
number 11. Early treatment diabetic retinopathy study research group. Ophthalmology
1991;98(5 Suppl):807–22.
20 Stitt AW, Gardiner TA, Archer DB. Histological and ultrastructural investigation
REFERENCES of retinal microaneurysm development in diabetic patients. Br J Ophthalmol
1 Klein BE. Overview of epidemiologic studies of diabetic retinopathy. Ophthalmic 1995;79:362–7.
Epidemiol 2007;14:179–83. 21 Moore J, Bagley S, Ireland G, et al. Three dimensional analysis of microaneurysms in
2 Antonetti DA, Klein R, Gardner TW. Diabetic retinopathy. N Engl J Med the human diabetic retina. J Anat 1999;194(Pt 1):89–100.
2012;366:1227–39. 22 Spaide RF, Fujimoto JG, Waheed NK, et al. Optical coherence tomography
3 Ometto G, Assheton P, Calivá F, et al. Spatial distribution of early red lesions is a angiography. Prog Retin Eye Res 2018;64:1-55–.
risk factor for development of vision-threatening diabetic retinopathy. Diabetologia 23 Shah AM, Bressler NM, Jampol LM. Does laser still have a role in the management of
2017;60:2361–7. retinal vascular and neovascular diseases? Am J Ophthalmol 2011;152:332–9.
4 Ribeiro ML, Nunes SG, Cunha-Vaz JG. Microaneurysm turnover at the macula predicts 24 Wang H, Chhablani J, Freeman WR, et al. Characterization of diabetic microaneurysms
risk of development of clinically significant macular edema in persons with mild by simultaneous fluorescein angiography and spectral-domain optical coherence
nonproliferative diabetic retinopathy. Diabetes Care 2013;36:1254–9. tomography. Am J Ophthalmol 2012;153:861–7.