Skip to main content

Characterization of the retinal vasculature in fundus photos using the PanOptic iExaminer system

Abstract

Background

The goal was to characterize retinal vasculature by quantitative analysis of arteriole-to-venule (A/V) ratio and vessel density in fundus photos taken with the PanOptic iExaminer System.

Methods

The PanOptic ophthalmoscope equipped with a smartphone was used to acquire fundus photos centered on the optic nerve head. Two fundus photos of a total of 19 eyes from 10 subjects were imaged. Retinal vessels were analyzed to obtain the A/V ratio. In addition, the vessel tree was extracted using deep learning U-NET, and vessel density was processed by the percentage of pixels within vessels over the entire image.

Results

All images were successfully processed for the A/V ratio and vessel density. There was no significant difference of averaged A/V ratio between the first (0.77 ± 0.09) and second (0.77 ± 0.10) measurements (P = 0.53). There was no significant difference of averaged vessel density (%) between the first (6.11 ± 1.39) and second (6.12 ± 1.40) measurements (P = 0.85).

Conclusions

Quantitative analysis of the retinal vasculature was feasible in fundus photos taken using the PanOptic ophthalmoscope. The device appears to provide sufficient image quality for analyzing A/V ratio and vessel density with the benefit of portability, easy data transferring, and low cost of the device, which could be used for pre-clinical screening of systemic, cerebral and ocular diseases.

Background

The retina provides a direct, non-invasive, and easily accessible window for observing the microvascular system. The vasculature of the retina and brain are anatomically and physiologically similar [1]. Fundus photography is a useful tool to observe and monitor changes in the retinal vasculature. Large-scale epidemiology studies based on fundus photography reported that the changes of retinal vessels are associated with the risk of retinal and systemic diseases, including diabetic retinopathy [2], stroke [3], cardiovascular mortality [4] and dementia [5].

While the traditional fundus camera offers good-quality images for analysis of large retinal vessels, these fundus camera systems are based in hospitals and research facilities. The office-based systems require skillful technicians and are limited to the subjectivity of the clinician’s interpretation. These systems are also bulky and costly, which limits their use in community screening efforts, especially in remote areas [6]. Two recent remarkable breakthroughs are improvements in teleophthalmology and smartphone adapted devices. With the advantage of easy image capturing, easy data transferring, and low cost, the PanOptic ophthalmoscope adapted with a smartphone and data acquisition application (iExaminer) may provide an ideal solution for pre-clinical disease screening [7,8,9]. The PanOptic ophthalmoscope system has been used in some previous studies. However, none of these previous studies quantified retinal vasculature, which renders whether the system can be used to characterize retinal vasculature. The goal of the present study was to characterize retinal vasculature by quantitative analysis of arteriole-to-venule (A/V) ratio and vessel density in smartphone-acquired fundus photographs using the PanOptic iExaminer System.

Methods

Subject selection

This study was approved by the Institutional Review Board (IRB) of the University of Miami (ID: 20070492), and all study subjects were treated according to the tenets of the Declaration of Helsinki. A signed written informed consent was obtained from each subject. Ten subjects were recruited at the Bascom Palmer Eye Institute. A total of 19 eyes were imaged. One study subject agreed to have one eye dilated for the present study. The following exclusion criteria were used: shallow anterior chamber, ocular surgery history, refractive errors more than ±6.0 diopters, or any systemic inflammatory or infectious diseases.

Image capture

PanOptic ophthalmoscope (WelchAllyn, Skaneateles Falls, NY) was mounted on a slit-lamp base to facilitate the alignment of the system with the eye (Fig. 1). An external fixation target, placed in front of the contralateral eye, was also provided. The PanOptic ophthalmoscope was adapted to the iExaminer adapter (WelchAllyn, Skaneateles Falls, NY). Fundus images were captured by the iExaminer Pro application with the iPhone 6 s. A short video clip was recorded during acquisition; then, the five best images were manually selected based on the focus of vasculature and optimal visualization (Fig. 1). For the field of view calibration, a Zeiss field of view calibration tool (S/N4007) was placed in front of the ophthalmoscope under room light (Fig. 1). The field of view was calibrated to 7.0 mm × 7.0 mm with the emitted light on. Each eye was dilated with topical tropicamide 1.0%. During imaging in a dark test room, the subject was seated, and the chin rested on the chinrest while looking at the fixation target.

Fig. 1
figure1

PanOptic iExaminer System and field of view calibration. a PanOptic ophthalmoscope adapted with a smartphone (iPhone 6 s) was mounted on a slit lamp base with the chinrest. b iExaminer software was used for image acquisition. c A representative fundus photo was captured using the system. d The field of view (without internal illumination) was tested using the Zeiss field of view calibration tool placed in front of the PanOptic ophthalmoscope under room light. e The field of view was calibrated as 7.0 mm × 7.0 mm with the emitted light on

A/V ratio analysis

The exported raw image was 1024 (vertical) × 720 (horizontal) pixels. To make the image conform to a square ratio, the black portions on the top and bottom of the image were removed, and the raw image was trimmed to 720 × 720 pixels with a field of view of 7.0 × 7.0 mm (Fig. 2). The measurement area for arteriole and venule was done within 0.5 to 1 disc diameter from the edge of the disc margin, as done in previous studies [10, 11]. The selected parallel arteriole and venule diameters were measured using the straight-line measurement tool of ImageJ. Five non-overlapping measurements for each vessel were made (Fig. 2). The diameters were then averaged, and the arteriole to venule ratio was calculated. Two photographs with the best focusing and centration of the optic nerve head of each eye were selected and calculated. The measurement for the A/V ratio was performed by two researchers.

Fig. 2
figure2

Example of arteriole and venule diameter measurements. a Raw fundus photograph acquired using the Panoptic iExaminer system. b Cropped square photograph. c The measurement area was within 0.5 to 1 disc diameter of the edge of the disc margin. The arrow showed the selected parallel arteriole and venule for demonstration. d Vessel centerline was drawn along with the long vessel axis. e Five non-overlapping measurement lines for each vessel were made. f The width of each measurement line was calculated using ImageJ

Vessel extraction using deep learning U-NET and density analysis of retinal vessels

We developed the image processing codes in the MATLAB (MathWorks, Natick, MA) for image conversion and filtering. We also developed U-Net image segmentation in Python (ver. 3.6.5). The photograph was converted into a square with a resolution of 565 × 565pixels (Fig. 3) and then transformed into the grayscale format using Gray = MAX Contrast (R, G, B). Although the green channel is commonly used to convert RGB images to grayscale, we used all channels to convert the RGB images so that they would contain more of the red color in these fundus photos to preserve more details of the vessels. The weight of each channel was obtained from the training set (DRIVE), and the setting was red × 0.299 + green × 0.587 + blue × 0.114. The image was then processed by contrast enhancement using contrast limited adaptive histogram equalization (CLAHE). We used the function of adapthisteq with NumTiles (25, 25), Cliplimit (0.01). Other parameters were set to default. The image was then processed using the function of imadjust (stretchlim, output = 0 to 1, gamma = 0.6). The image was further processed by the linear filter and edge enhancement to enhance the vessel border (Fig. 3) [12]. We used the MATLAB function of imfilter [parameters: w = fspecial (average = 11, boundary options = replicate, others = default] to remove the artifact of the image edge, which was enhanced during the image contrast enhancement. This process will not affect the extraction of the vessels. In recent years, semantic segmentation methods have been used to segment the vessels from fundus photos, using deep learning approaches [13, 14]. One of the most representative networks is U-Net (convolutional neural network), which was developed for the biomedical imaging segment. This approach often uses the hand-labeled picture as the ground truth for the learning model. The U-NET deep learning method has been well described in previous studies [13,14,15,16]. In the present study, the U-NET structure based on the fully convolutional neural network [15] was developed and then used to enhance the contrast of vessels. After that, blood vessels were extracted (Fig. 3). The U-shaped network used 23 convolutional layers. The public datasets of digital retinal images for vessel extraction (Drive: https://www.isi.uu.nl/Research/Databases/DRIVE/) [13, 14, 17] were used as the training data. There were a total of 40 source images (565 × 584 pixels), which were used as training images. The random image patches were generated from the training image data and patched into the training image dataset [14]. A total of 38 blood vessel images were processed [13, 14]. The images were then converted to binary for vessel density analysis using ImageJ (Fig. 3). Vessel density was defined as the percentage of pixels within the vessels over the entire image.

Fig. 3
figure3

Example of retinal vessel density analysis. a Cropped square photograph. b Grayscale conversion using Gray = MAX Contrast. c Contrast enhancement using contrast limited adaptive histogram equalization. d Vessel border enhancement using the linear filter and edge enhancement. e Vessel extraction using U-Net. f Vessel binary conversion using ImageJ

Statistical analysis

All values are expressed as mean ± standard deviation. Significance was assessed with the Student’s t-test for two variable comparisons using SPSS Statistics package (ver. 25, IBM Corp., Armonk, NY, USA). Pearson’s regression was used to determine the relationships among parameters. The Bland-Altman plot was constructed between the two measurements to determine the 95% limit of agreement, which was calculated as 1.96 × (the standard deviation of the difference between repeated measurements). P < 0.05 was considered statistically significant.

Results

Subject characteristics

Demographic information is shown in Table 1. Three participants were male, seven were female. Three study participants were healthy subjects, four had a history of diabetes mellitus (DM), and three had a history of multiple sclerosis (MS). In a total of 19 eyes, 10 eyes were right eyes, and 9 eyes were left eyes.

Table 1 Demographic information

A/V ratio

The A/V analysis was successfully processed in all subjects of each group (Fig. 4). There was no significant difference of averaged A/V ratio between the first (0.77 ± 0.09) and second (0.77 ± 0.10) measurements (P = 0.53) (Fig. 5). The A/V ratio of the first measurement ranged from 0.58 to 0.96, while it ranged from 0.59 to 0.97 in the second measurement. Two measurements were significantly correlated (r = 0.94, P < 0.001) (Fig. 5). Bland-Altman plot for the two measurements showed a bias of − 0.005 with upper and lower confidence intervals of 0.063 and − 0.073, respectively.

Fig. 4
figure4

A/V ratio analysis of representative cases of healthy control (HC), diabetes mellitus (DM), and multiple sclerosis (MS) subjects. a-c Retinal photographs with circle defining the measurement. The arrow indicates selected parallel arteriole and venule. d-f Five non-overlapping diameter measurements of arteriole and venule by ImageJ

Fig. 5
figure5

A/V ratio calculation of first and second measurements. Bland-Altman plot of differences in the first measurement vs. the second measurement

Vessel density

Vessels were extracted from the fundus photos of subjects for analysis of vessel density (Fig. 6). There was no significant difference of averaged vessel density (%) between the first (6.11 ± 1.39) and second (6.12 ± 1.40) measurements (P = 0.85) (Fig. 7). In the first measurement, vessel density ranged from 3.64 to 8.42, while it ranged from 3.63 to 8.30 in the second measurement. The two measurements were significantly correlated (r = 0.97, P < 0.001). Bland-Altman plot for the two measurements showed no bias with upper and lower confidence intervals of 0.60 and − 0.63, respectively.

Fig. 6
figure6

Vessel density analysis of representative cases of healthy control (HC), diabetes mellitus (DM), and multiple sclerosis (MS) subjects. a, d, g Adjusted square photographs. b, e, h Vessel extraction images. c, f, i Binary images

Fig. 7
figure7

Vessel density analysis of first and second measurements. Bland-Altman plot of differences in the first measurement vs. the second measurement

Discussion

This study was to quantitatively analyze fundus photos for characterizing retinal vasculature by calculating the A/V ratio and vessel density in fundus photos taken using the PanOptic iExaminer system. This study provides evidence that fundus photos with sufficient quality for analysis of A/V ratio and vessel density, can be acquired using the portable PanOptic system for image analysis. The acquisition appeared to be facilitated by the modifications used in the present study. First, the ophthalmoscope was stabilized by the use of the slit-lamp base, which also made it easy to align with the eye. Second, the patient’s head was stabilized by resting on the chinrest. Third, the fixation target facilitated the gaze direction for imaging. Fourth, pupil dilation also helped in image acquisition. It may be worth noting that while these modifications may not alter the portability and data transfer, the slit-lamp base with the chinrest can be easily transported with the portable ophthalmoscope. The inconvenience of the use of the slit-lamp base may not outweigh the gain of easy acquisition of high-quality images for quantitative analysis. It is also worth noting that some of the inexpensive handheld fundus cameras are also available, although some of these cameras do not use smartphones [18, 19]. Based on the outcomes in the present study and previous studies [20], the PanOptic system could potentially be used for pre-screening in remote communities. It would be recommended that fundus photos may be analyzed using analysis software installed in the smartphone [20] or in fundus photo reading centers or research labs.

Previous studies (Table 2) demonstrated the usefulness of the PanOptic ophthalmoscope, while the present study provided an alternative solution to use the device with the add-on translator and chinrest. The PanOptic ophthalmoscope was used qualitatively for patients’ fundus imaging in the emergency room [23] and optic disc assessment [24], for diabetic retinopathy screening [7, 9] and medical students ophthalmoscopy skills assessment [8]. There are some studies using the PanOptic device, quantitatively analyzing the vessel diameter in the retina [20] and anterior lens capsule vascularity only [22]. In addition, vessel segmentation and analysis based on fundus photos taken with the PanOptic ophthalmoscope were also done in previous studies [20, 21]. Xu et al. did not analyze the A/V ratio and vessel density [20, 21], which are critical to the clinical applications of the PanOptic ophthalmoscope equipped with a smartphone. In addition, Xu et al. did not analyze the repeatability of these important measurements (i.e., A/V ratio and vessel density) [20, 21]. Our work focused on the characterization of retinal vasculature by quantitatively analyzing A/V and vessel density, which provides insightful information on the repeatability before testing whether the system can be used for clinical diagnosis. Although these previous studies [20, 21] and the present study applied a similar methodology to extract the vessel information, this study provides additional information to the feasibility and repeatability of measuring A/V ratio and vessel density. In addition, the alternative solution by adding the slit-lamp base may facilitate translating the portable ophthalmoscope equipped with the smartphone, such as the PanOptic device. However, more work needs to be done to test whether the portable ophthalmoscope equipped with the smartphone can differentiate diseased populations from the normal population.

Table 2 Summary of PanOptic ophthalmoscopy studies

Compared to images captured by traditional fundus cameras and Ultra-wide field scanning laser ophthalmoscopy (SLO), the resolution (1280 × 720 pixels in video recording) of PanOptic fundus images is not high due to the settings in the iExaminer software for video recording. However, there are some advantages to making it a practical instrument for retina blood vessel analysis of disease screening (Table 3), which are its portability, low cost, smartphone adaptability, and easy data transferring. Traditional fundus camera offers good-quality images but is bulky, office-based, and technician dependent, which limits its use as a community screening tool, especially in remote areas. SLO, with a 180–200° field of view, offers faster and easier image acquisition without pupil dilation when compared to traditional fundus cameras [28]. It is ideal for a hospital-based study. However, SLO is costly, which may reduce its availability for disease screening. In our future study, we will compare the retinal vascular network analysis results of the PanOptic iExaminer system with those of traditional fundus cameras and SLO.

Table 3 Comparison of fundus imaging systems

The A/V ratio acquired using the PanOptic iExaminer system appeared within the range of those found at using the traditional fundus camera of previous large sample studies [10, 29, 30]. The good repeatability of this analysis indicated its feasibility using the analysis method of McClelland et al. [31] although in this study, the A/V ratio’s correlation to other disease conditions was not analyzed as the focus of this paper was a feasibility test. Other studies already reported that the A/V ratio of the retina is a widely used parameter in the assessment of different ocular and systemic vascular diseases, such as open-angle glaucoma [32], incident stroke [32], coronary heart disease [33] and dementia [34].

There were several limitations to this study. First, the aim of this study was to characterize retinal vasculature using the commercially available ophthalmoscope equipped with a smartphone by reporting the feasibility and repeatability of the measurements. Although this study provided an alternative solution for resource-limited regions and countries, there are low-cost portable handheld fundus cameras such as the portable handheld camera with a retinal model (Pictor, Volk Optical, Inc., Mentor, OH, USA) [18, 35]. The camera could perform a similar task for pre-screening [18, 35]. Therefore, more work will need to be done to test whether the portable ophthalmoscope equipped with the smartphone, such as the PanOptic ophthalmoscopy, can be used to screen for retinal diseases. Second, although we aimed for the feasibility of analysis, the sample size was still small. We could not compare the differences among groups. Larger sample size and case-control studies are needed to further verify further whether the described method could differentiate vasculature alterations in the pathologic retina. Third, adding the slit-lamp translator and chin-and-head rest as an alternative add-on of the PanOptic ophthalmoscope will increase the cost, which amounts to an entry-level retinal camera. Nevertheless, our study provided an alternative solution to use the PanOptic ophthalmoscope with the translator and chinrest. Fourth, we did not compare the benefit of the smartphone to other portable fundus cameras without smartphones. While different approaches using the smartphone on the handheld fundus camera or ophthalmoscopes are available and continuously being developed, adding the smartphone may not necessarily make a device better. The applications using the smartphone on the portable fundus camera may be dependent on the availability and the need for particular functionalities.

Conclusions

This study demonstrated the feasibility of characterizing retinal vasculature using the PanOptic iExaminer System, which yielded sufficient image quality of the fundus photos for quantitative analysis of A/V ratio and vessel density. The device appears to provide the benefit of portability, easy data transferring, and cost-effectiveness for the possible use in pre-clinical screening of systemic, cerebral, and ocular diseases.

Availability of data and materials

The datasets used and analyzed for the present study are available from the corresponding author upon reasonable request.

References

  1. 1.

    Patton N, Aslam T, Macgillivray T, Pattie A, Deary IJ, Dhillon B. Retinal vascular image analysis as a potential screening tool for cerebrovascular disease: a rationale based on homology between cerebral and retinal microvasculatures. J Anat. 2005;206(4):319–48.

    Article  Google Scholar 

  2. 2.

    Stevenson CH, Hong SC, Ogbuehi KC. Development of an artificial intelligence system to classify pathology and clinical features on retinal fundus images. Clin Exp Ophthalmol. 2019;47(4):484–9.

    Article  Google Scholar 

  3. 3.

    Doubal FN, MacGillivray TJ, Patton N, Dhillon B, Dennis MS, Wardlaw JM. Fractal analysis of retinal vessels suggests that a distinct vasculopathy causes lacunar stroke. Neurology. 2010;74(14):1102–7.

    CAS  Article  Google Scholar 

  4. 4.

    Arnould L, Binquet C, Guenancia C, Alassane S, Kawasaki R, Daien V, et al. Association between the retinal vascular network with Singapore "I" Vessel Assessment (SIVA) software, cardiovascular history and risk factors in the elderly: the Montrachet study, population-based study. PLoS One. 2018;13(4):e0194694.

  5. 5.

    Schrijvers EM, Buitendijk GH, Ikram MK, Koudstaal PJ, Hofman A, Vingerling JR, et al. Retinopathy and risk of dementia: the Rotterdam Study. Neurology. 2012;79(4):365–70.

  6. 6.

    Panwar N, Huang P, Lee J, Keane PA, Chuan TS, Richhariya A, et al. Fundus photography in the 21st century--a review of recent technological advances and their implications for worldwide healthcare. Telemed J E Health. 2016;22(3):198–208.

    Article  Google Scholar 

  7. 7.

    Gill JM, Cole DM, Lebowitz HM, Diamond JJ. Accuracy of screening for diabetic retinopathy by family physicians. Ann Fam Med. 2004;2(3):218–20.

    Article  Google Scholar 

  8. 8.

    McComiskie JE, Greer RM, Gole GA. Panoptic versus conventional ophthalmoscope. Clin Exp Ophthalmol. 2004;32(3):238–42.

    Article  Google Scholar 

  9. 9.

    Tan A, Mallika P, Aziz S, Asokumaran T, Intan G, Faridah H. Comparison between the panoptic ophthalmoscope and the conventional direct ophthalmoscope in the detection of sight threatening diabetic retinopathy: the Kuching diabetic eye study. Malays Fam Physician. 2010;5(2):83–90.

    PubMed  PubMed Central  Google Scholar 

  10. 10.

    Hubbard LD, Brothers RJ, King WN, Clegg LX, Klein R, Cooper LS, et al. Methods for evaluation of retinal microvascular abnormalities associated with hypertension/sclerosis in the atherosclerosis risk in communities study. Ophthalmology. 1999;106(12):2269–80.

    CAS  Article  Google Scholar 

  11. 11.

    Knudtson MD, Lee KE, Hubbard LD, Wong TY, Klein R, Klein BE. Revised formulas for summarizing retinal vessel diameters. Curr Eye Res. 2003;27(3):143–9.

    Article  Google Scholar 

  12. 12.

    Setiawan A, Mengko T, Santoso O, Suksmono A. Color Retinal Image Enhancement using CLAHE. In: International Conference on ICT for Smart Society: IEEE; 2013. p. 1–3.

  13. 13.

    Luo L, Chen D, Xue D. Retinal blood vessels semantic segmentation method based on modified U-Net. In: 2018 Chinese Control And Decision Conference (CCDC): IEEE; 2018.

  14. 14.

    Wang C, Zhao Z, Ren Q, Xu O, Yu Y. Dense U-net based on patch-based learning for retinal vessel segmentation. Entropy. 2019;21(2):168.

    Article  Google Scholar 

  15. 15.

    Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention: Springer; 2015. p. 234–41.

  16. 16.

    Wang X, Zhangwei J, Wen X, Zhenyan J, Gu H, Li W, Miao B, Jing H. Retina Blood vessel segmentation using a U-Net based convolutional neural network. In: International Conference on Data Science; 2018.

    Google Scholar 

  17. 17.

    Staal J, Abràmoff MD, Niemeijer M, Viergever MA, van Ginneken B. Ridge-based vessel segmentation in color images of the retina. IEEE Trans Med Imaging. 2004;23(4):501–9.

  18. 18.

    Wadhwani M, Vashist P, Singh SS, Gupta N, Malhotra S, Gupta A, et al. Diabetic retinopathy screening programme utilising non-mydriatic fundus imaging in slum populations of New Delhi, India. Tropical Med Int Health. 2018;23(4):405–14.

    Article  Google Scholar 

  19. 19.

    Gosheva M, Klameth C, Norrenberg L, Clin L, Dietter J, Haq W, et al. Quality and learning curve of handheld versus stand-alone non-mydriatic cameras. Clin Ophthalmol. 2017;11:1601–6.

    Article  Google Scholar 

  20. 20.

    Xu X, Ding W, Wang X, Cao R, Zhang M, Lv P, et al. Smartphone-based accurate analysis of retinal vasculature towards point-of-care diagnostics. Sci Rep. 2016;6:34603.

    CAS  Article  Google Scholar 

  21. 21.

    Xu X, Wang R, Lv P, Gao B, Li C, Tian Z, et al. Simultaneous arteriole and venule segmentation with domain-specific loss function on a new public database. Biomed Opt Express. 2018;9(7):3153–66.

    Article  Google Scholar 

  22. 22.

    Patel M, Mukherjee D, Farsiu S, Munoz B, Blood AB, Wilson CG, et al. Estimation of gestational age via image analysis of anterior lens capsule vascularity in preterm infants: a pilot study. Front Pediatr. 2019;7:43.

    Article  Google Scholar 

  23. 23.

    Day LM, Wang SX, Huang CJ. Nonmydriatic fundoscopic imaging using the Pan Optic iExaminer system in the pediatric emergency department. Acad Emerg Med. 2017;24(5):587–94.

  24. 24.

    Petrushkin H, Barsam A, Mavrakakis M, Parfitt A, Jaye P. Optic disc assessment in the emergency department: a comparative study between the PanOptic and direct ophthalmoscopes. Emerg Med J. 2012;29(12):1007–8.

    Article  Google Scholar 

  25. 25.

    Desai AD, Peng C, Fang L, Mukherjee D, Yeung A, Jaffe SJ, et al. Open-source, machine and deep learning-based automated algorithm for gestational age estimation through smartphone lens imaging. Biomed Opt Express. 2018;9(12):6038–52.

    Article  Google Scholar 

  26. 26.

    Lee JY, Gallo RA, Alabiad CR. Evaluating the effectiveness of small-group training in teaching medical students integral clinical eye examination skills. J Acad Ophthalmol. 2020;12(1):e79–86.

    Article  Google Scholar 

  27. 27.

    Besenczi R, Szitha K, Harangi B, Csutak A, Hajdu A. Automatic optic disc and optic cup detection in retinal images acquired by mobile phone. In: 9th International Symposium on Image and Signal Processing and Analysis (ISPA 2015); 2015. p. 139–8.

    Google Scholar 

  28. 28.

    Nghiem AZ, Nderitu P, Lukic M, Khatun M, Largan R, Kortuem K, et al. Comparing diabetic retinopathy lesions in scanning laser ophthalmoscopy and colour fundus photography. Acta Ophthalmol. 2019;97(8):e1035–40.

    Article  Google Scholar 

  29. 29.

    Ponto KA, Werner DJ, Wiedemer L, Laubert-Reh D, Schuster AK, Nickels S, et al. Retinal vessel metrics: normative data and their use in systemic hypertension: results from the Gutenberg health study. J Hypertens. 2017;35(8):1635–45.

    CAS  Article  Google Scholar 

  30. 30.

    Ikram MK, de Jong FJ, Vingerling JR, Witteman JC, Hofman A, Breteler MM, et al. Are retinal arteriolar or venular diameters associated with markers for cardiovascular disorders? The Rotterdam study. Invest Ophthalmol Vis Sci. 2004;45(7):2129–34.

    Article  Google Scholar 

  31. 31.

    McClelland JF, O'Donoghue L, McIntyre M, Saunders KJ. Cup-to-disc and arteriole-to-venule ratios in children aged 6-7 and 12-13 years. Ophthalmic Physiol Opt. 2012;32(1):31–8.

    Article  Google Scholar 

  32. 32.

    Mitchell P, Wang JJ, Wong TY, Smith W, Klein R, Leeder SR. Retinal microvascular signs and risk of stroke and stroke mortality. Neurology. 2005;65(7):1005–9.

    CAS  Article  Google Scholar 

  33. 33.

    Wong TY, Cheung N, Islam FM, Klein R, Criqui MH, Cotch MF, et al. Relation of retinopathy to coronary artery calcification: the multi-ethnic study of atherosclerosis. Am J Epidemiol. 2008;167(1):51–8.

    Article  Google Scholar 

  34. 34.

    Bulut M, Kurtuluş F, Gözkaya O, Erol MK, Cengiz A, Akidan M, et al. Evaluation of optical coherence tomography angiographic findings in Alzheimer's type dementia. Br J Ophthalmol. 2018;102(2):233–7.

  35. 35.

    Bursztyn L, Woodward MA, Cornblath WT, Grabe HM, Trobe JD, Niziol L, et al. Accuracy and reliability of a handheld, nonmydriatic fundus camera for the remote detection of optic disc edema. Telemed J E Health. 2018;24(5):344–50.

    Article  Google Scholar 

Download references

Acknowledgments

Not applicable.

Funding

The work has been supported by NIH Center Grant P30 EY014801, NINDS 1R01NS111115–01 (Wang), the Ed and Ethel Moor Alzheimer’s Disease Research Program (Florida Health, 20A05, to Jiang) and a grant from Research to Prevent Blindness (RPB). Visiting scholar activities (Haicheng Wei and Mingxia Xiao) were supported by the North Minzu University Scientific Research Projects (Major projects No. 2019KJ37 and 2018XYZDX11), National Natural Science Foundation of China (No. 61861001) and Natural Science Foundation of Ningxia (No. 2020AAC03220).

Author information

Affiliations

Authors

Contributions

HH, HW, MX, LJ, HW, and JW collected, analyzed, and interpreted the data. HH, HJ, TR, and JW were the major contributors for drafting the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Jianhua Wang.

Ethics declarations

Ethics approval and consent to participate

All research methods were conducted following the tenets of the declaration of Helsinki and approved by the ethics committee board of the University of Miami (ID 20070492). All subjects were recruited voluntarily and were informed about the purposes, methods, and the potential risks of the study. A signed consent form was obtained from each volunteer.

Consent for publication

All study subjects gave informed consent.

Competing interests

The authors declare that they have no competing interests.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Hu, H., Wei, H., Xiao, M. et al. Characterization of the retinal vasculature in fundus photos using the PanOptic iExaminer system. Eye and Vis 7, 46 (2020). https://doi.org/10.1186/s40662-020-00211-5

Download citation

Keywords

  • Smartphone ophthalmoscope
  • Image analysis
  • Deep learning
  • Arteriovenous ratio
  • Vessel density
  • Retina