Purpose To compare the vessel calibre measurements between optical coherence tomography angiography (OCTA) and colour fundus photography.
Methods In this retrospective comparative study, OCTA and colour fundus images of healthy eyes and eyes with optic atrophy were evaluated. The colour fundus image was registered manually using Image J software to the OCTA image of the optic disc. Two independent graders measured the vessel calibre of the widest vein and artery in each peripapillary quadrant on a 3.4 mm diameter circle centred on the optic disc in the same location on both images. The difference in vessel calibre between the two techniques was assessed.
Results A total of 312 vessels from 29 healthy eyes and 20 eyes with atrophic optic discs were included. There was a high level of agreement between graders for measurement of vessel calibre in both colour fundus (intraclass correlation coefficient=0.93, coefficient of variation=0.07) and OCTA images (intraclass correlation coefficient=0.94, coefficient of variation=0.05). The mean vessel calibre in colour fundus images (94.5±23.2 µm) and OCT images (112.2±26.1 µm) was correlated (r=0.8, p<0.001), but the difference was statistically significant (mean difference: 17.6±1.5 µm, p<0.001). This difference was evident for both arteries (mean difference: 18.2±16.3 µm, p<0.001) and veins (mean difference: 15.1±16.2 µm, p<0.001) individually, with a similar magnitude of difference for both vessel types (p=0.08). In addition, the magnitude of difference between imaging modalities was similar in atrophic and healthy discs (17.1±15.9 vs 18.4±15.2 µm, respectively, p=0.4). The difference, however, was significantly higher in vessels with a calibre of ≤94.5 compared with larger vessels (19.3±16.3 vs 15.6±14.4 µm, respectively, p=0.02).
Conclusions Vessel calibre measurements were significantly larger in OCTA images compared with colour fundus photographs, particularly for smaller vessels. These differences may need to be accounted for when using OCTA-derived metrics.
- Optic Nerve
Statistics from Altmetric.com
Optical coherence tomography angiography (OCTA) is a novel imaging technology that uses motion contrast imaging to generate three-dimensional high-resolution vascular maps of the retinal and choroidal microcirculation. OCTA has demonstrated its ability to depict vascular changes in various retinal and choroidal diseases including diabetic retinopathy, retinal vascular occlusive disease and choroidal neovasculartisation.1 ,2 In addition, several studies have reported OCTA-derived quantitative metrics such as vessel density measurements, and have correlated these measures with various structural parameters in both healthy and diseased eyes.1 ,2
Despite many clear advantages over traditional dye-based angiography, the interpretation of OCTA images can be cofounded by a variety of artefacts. These include, but are not limited to, projection, segmentation, motion, masking, unmasking and blink artefacts.3 ,4 Spaide et al3 speculated that in OCTA, the calibre of vessels may be depicted to be larger than in their real state, but vessels calibres were not specifically quantified in order to validate this hypothesis. Since vessel width can potentially affect many OCTA-derived metrics including vessel density measurements, the magnitude of this difference in representation of the vessel width requires careful assessment. Thus, the aim of this study was to compare the vessel calibres between the images obtained by colour fundus photography and OCTA techniques.
This analysis was a retrospective comparative study. The images of all healthy subjects and patients with optic atrophy who underwent optic disc OCTA imaging in the Imaging Unit of the Doheny Eye Center between January and May 2016 were reviewed. The study was approved by the Institutional Review Board of the University of California—Los Angeles. The research adhered to the tenets of the Declaration of Helsinki and the Health Insurance Portability and Accountability Act, and written informed consent was obtained from all participants.
The presence of optic atrophy was diagnosed by an expert neuro-ophthalmologist (AS) after a complete ophthalmic examination, and then confirmed by the presence of nerve fibre layer thinning on OCT. Healthy subjects were required to have a normal ocular examination. Subjects with vitreopapillary traction and significant media opacity precluding high-quality fundus imaging were excluded.
OCTA was performed with a Topcon OCT instrument (DRI OCT Triton plus, Topcon, Tokyo, Japan). The Triton swept-source OCT uses a wavelength of 1050 nm with a scan speed of 100 000 A-scans per second. The instrument employs an active eye tracker that follows the eye movement, detects blinking, and adjusts the scan position accordingly, thereby reducing motion artefact during OCTA acquisition. All eyes were scanned using a 6×6 mm protocol centred on the optic nerve head. The device automatically takes a colour fundus image at the completion of the OCTA scan. The resolution of the fundus image is ≥60 lines/mm with a field of view of 45°. The optic disc OCTA scans were automatically segmented by the OCTA software to generate a retinal slab from the internal limiting membrane to the ellipsoid zone.
The OCTA and colour fundus images were extracted from the OCT instrument and imported to publicly available ImageJ software (National Institutes of Health, Bethesda, Maryland, USA). Each colour fundus image was registered over the corresponding OCTA image using the transform function and anatomic landmark correspondences, and a two frame same-size image stack was created. A circle with a diameter of 3.4 mm was centred on the optic nerve head using the colour fundus image. For each of the peripapillary quadrants, the largest diameter vein and artery were marked. If there was a motion artefact on the corresponding location in the OCTA image, the vessel was excluded. Two independent, certified Doheny Image Reading Center graders (KGF and MA-S) measured the width of the marked vessels at the border of the previously fitted 3.4 mm circle (figure 1). For this purpose, an image magnification of 200× was selected. The graders selected the most appropriate image brightness and contrast to detect the borders of the vessels. The border of the vessels was manually determined using an edge-based approach.5 In this method, each pixel was compared against the pixel's surrounding window at the external edge of the vessel, and tracking of the vessel margin was used to better maintain the connectivity of the vessel structure.
Data were analysed using SPSS software (V.16, SPSS, Chicago, Illinois, USA). The agreement between the two graders was analysed using intraclass correlation coefficients (ICC) and coefficients of variation. Bland-Altman plots were generated to show the 95% limit of agreement between the graders. The mean of the measurements of the two graders were used for comparison between two techniques. A paired t-test was used to compare the measurements between the two techniques. Pearson correlation test was used to show the correlation between the measurements. The difference between vein and artery measurements, and the difference between the measurements in atrophic and normal optic discs were analysed using Student's t-test. A p value <0.05 was considered significant.
Overall, 49 eyes of 30 patients including 13 females and 17 males with a mean age of 31.2±1.2 years (range 13–61) were included. The optic disc was normal in 29 eyes and atrophic in 20 eyes. The diagnoses for patients with optic atrophy were Leber's hereditary optic neuropathy (16 eyes), autoimmune optic neuropathy (2 eyes) and anterior ischaemic optic neuropathy (2 eyes).
The agreement between graders was excellent for the measurement of vessel calibre in colour fundus (ICC=0.93) and OCTA images (ICC=0.94). The coefficient of variation was 0.07 for colour fundus and 0.05 for OCTA techniques. Bland-Altman analysis indicated that the 95% limits of agreement between the two graders ranged from −22.2 to 25.6 µm for colour fundus measurements and from −24.6 to 22.7 µm for OCTA measurements (figure 2). The mean difference between the two graders was 1.7±12.2 µm for colour fundus images and −0.92±12.1 µm for the OCTA measurements.
A total of 312 vessels including 152 veins and 160 arteries were measured. The mean vessel calibre was 94.5±23.2 µm and 112.2±26.1 µm in colour fundus and OCTA images, respectively (table 1). The difference was statistically significant (p<0.001). In 282 vessels (89.2%), the vessel calibres in OCTA were higher than colour fundus measurements. OCTA vessel calibre measurements were significantly correlated with colour fundus measurements (r=0.8, p<0.001, figure 3).
The mean difference in vessel calibre measurements between the two techniques was 17.6±1.5 µm for the cohort overall, with a mean difference of 18.4±15.2 µm in normal eyes and 17.1±15.9 µm in eyes with optic atrophy (p=0.4). When considering the type of vessel, the mean difference in vessel calibre measurements between the two techniques was 18.2±16.3 µm for arteries and 15.1±16.2 µm for veins (p=0.08). The size of the vessels appeared to be important as well, as the mean difference in vessel calibre measurements between the two techniques was 15.6±14.4 µm for large vessels (vessels with a calibre of >94.5 in colour fundus photography) and 19.3±16.3 µm for smaller vessels (vessels with a calibre of ≤94.5 µm in colour fundus photography, p=0.02).
In this study, the mean vessel calibre measured on OCTA images was significantly higher than the measurement derived from colour fundus images. The mean difference of 17.6 µm between the two techniques was significantly higher than the mean intergrader difference (1.7 µm), and this difference was greater for smaller vessels compared with larger vessels. Spaide et al3 suggested that the larger apparent width of the vessels in OCTA images is due to the ability of the instrument to detect the low flow signals by an eccentric laser beam. The flow saturates at a relatively low rate. With an eccentric laser beam that covers a small part of the vessel, the flow is detected, and therefore, the vessel is represented with a larger diameter than it is in actuality. In addition, the apparent widening proportionately affects the diameter for smaller vessels to a greater extent than larger ones. An alternative explanation, however, for the difference between the two techniques with regard to measurement of vessel calibre is that the presumed border of the vessels in the colour images is also not correct. Previous studies have suggested that the measurement of vessel width in colour fundus images may represent only the central blood column within the vascular lumen rather than the whole vessel width, and it is known that dye-filled vessels in fluorescein angiographic images measure wider than vessels in fundus images without dye.6 ,7 Other possible explanations for the differences between OCTA and colour fundus images could be differences in brightness and contrast characteristics, which may affect the determination of the vessel borders. As a result, one must concede the possibility (perhaps likely) that neither colour fundus photography nor OCTA may depict the ‘true’ vessel diameter due to inherent technical factors.
Several studies have used vessel density measurements to show structural changes in eyes with various retinal, choroidal and optic nerve head disorders.8–14 Vessel density is defined as the percentage of a region of interest that is occupied by vessels. While some studies used the original OCTA images to calculate vessel density, others employed skeletonisation of the vessels to reduce the vessels to a one-pixel thickness, ignoring variations in vessel width, calibre and shape.15 Considering the possibility of artifactitious widening of the vessels in OCTA images, skeletonisation of the vessels may be more reliable for comparing the vessel density between different patients, different follow-up visits, and different instruments. However, it should be noted that we did not compare vessel width measurements for very small vessels or capillaries between the two methods—the smallest vessel calibre measured in our study was 31 µm.
We used fundus photography as the reference standard for vessel calibre measurement. Careful registration of the colour fundus images over the OCTA images and creating the image stacks allowed the measurements of vessel width. Therefore, the measurements were performed using the same method for both image types, on the same location and using the same measurement unit. The vessels were often overlapped near the disc margin and the dense peripapillary microvascular network did not allow precise measurement of the vessel width in this region. Therefore, we selected an image size of 6×6 to fit a 3.4 mm diameter circle.
Our study has several limitations. The sample size is relatively small. Also, we did not measure all vessels exiting the optic disc. In addition, we did not use stereophotographs for the measurements. Although the graders tried to carefully register the images using anatomical landmarks, minor registration errors are likely unavoidable. Moreover, we used the OCTA images from a single OCT-A device (Topcon Triton swept-source OCT). The instrument and the associated OCTA software could use a variety of proprietary algorithms and processing techniques, which could potentially affect the displayed angiographic image. Hence, the observed findings in our study may not generalise to other OCTA devices.
In summary, our study demonstrates that vessel calibres may measure larger in OCTA images compared with conventional colour fundus images. Moreover, the smaller vessels are affected more than larger vessels. This potential confounder should be considered when correlating vessel density parameters with other structural or functional measurements. Future studies with larger sample sizes and other OCTA instruments are necessary to confirm our findings.
Contributors Concept and design; data collection; analysis and interpretation; writing the article; critical revision of the article; final approval of the article; provision of materials, patients or resources; literature search: KGF. Final approval of the article; data collection; provision of materials: MA-S. Data collection; critical revision of the article; final approval of the article; provision of materials, patients or resources: FD. Critical revision of the article; final approval of the article; provision of materials, patients or resources; administrative, technical or logistic support: AS. Concept and design; critical revision of the article; final approval of the article; provision of materials, patients or resources; administrative, technical or logistic support: SRS.
Competing interests SRS is a consultant for Optos, Genentech, and Allergan and receives research support from Optos, Genentech, Allergan and Carl Zeiss Meditec. AS receives unrestricted grant support from GenSight, Stealth Biotherapeutics and Edison.
Patient consent Obtained.
Ethics approval IRB of the UCLA.
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.