Teleophthalmology of Retinal Diseases

Updates on telemedicine for diabetic retinopathy screening and smartphone funduscopy

Teleophthalmology of Retinal Diseases

Updates on telemedicine for diabetic retinopathy screening and smartphone funduscopy


The way we practice medicine is changing every day, and telemedicine is part of this change. The question for us is whether telemedicine has a role to play in ophthalmology. Teleophthalmology, as it has been called, is definitely part of our future in ophthalmology.

Of key importance in ophthalmology is early detection. Teleophthalmology provides an easy, cost-effective way to diagnose many retinal diseases and ultimately to preserve a patient’s eyesight. In retina, there has been a trend toward teleophthalmology, particularly for diabetic retinopathy and retinopathy of prematurity (ROP) screenings, especially in areas where retinal specialists are not readily available.


Important points to consider when assessing the effectiveness of telemedicine are the quality of the images obtained and the accuracy in the interpretation of these images for the screening of retinal diseases, such as DR. The number of patients with diabetes in the United States will continue to grow, and the only way to potentially screen all of these patients for sight-threatening DR would be to utilize non-eyecare professionals through access to telemedicine.

Several studies have provided proof of the concept that telemedicine for referral-warranted DR could serve as a screening tool to identify those patients who would not otherwise be seen by an ophthalmologist or eyecare professional.

Chin et al1 conducted a retrospective review of all patients screened for DR using nonmydriatic fundus photography (nFP) between July 2006 and May 2011 and whose electronic images were reviewed by a retinal specialist remotely.

Luis J. Haddock, MD, and Allie Nadelson, MD, practice at the Bascom Palmer Eye Institute of the University of Miami Medical School in Florida. Neither author reports any financial interests in any products mentioned here. Dr. Haddock can be reached via e-mail at

The patients were seen either at rural medical clinics, where they had images obtain using a Topcon (Livermore, CA) TRC-NW6S nonmydriatic fundus camera, or at an urban family practice, where they had images obtained using a Nidek (Fremont, CA) AFC-210 nonmydriatic fundus camera.1 A complete summary of the results from this study is provided in Tables 1 and 2 (page 37).

Table 1. Retrospective Review
Total number of subjects 872 517 -
Fundus photographs with adequate image qualitya 1,437 (82.4%) 886 (85.7%) <0.05
Normal fundus 639 (73.3%) 290 (56.1%) <0.001
Signs of diabetic retinopathy 110 (12.6%) 153 (29.6%) <0.001
Fundus abnormalities other than diabetic retinopathy 123 (14.1%) 74 (14.3%) 0.94
a Image quaility was deemed adequate if either good or excellent quality, as per the retinal specialist’s interpretation

Table 2. Abnormalities Found
Total number of fundus photographs 1, 744 1, 034 -
Hypertensive retinopathya 1 (0.1%) 6 (0.6%) <0.05
Optic nerve cupping (cup/disc >0.5) 183 (10.5%) 79 (7.6%) <0.05
Optic nerve disease unspecifiedb 11 (0.6%) 7 (0.7%) 0.81
Retinal scarringc 18 (1.0%) 13 (1.3%) 0.46
Pigmented lesionsd 2 (0.1%) 3 (0.3%) 0.17
Vascular pathologye 7 (0.4%) 5 (0.5%) 0.77
Vitreous pathologyf 3 (0.2%) 3 (0.3%) 0.68
Otherg 6 (0.3%) 5 (0.5%) 0.55

a Includes copper wiring, silver wiring, flame hemorrhages

b Optic nerve drusen, optic nerve pallor, optic nerve gliosis, myelinated nerve fiber layer

c Panretinal photocoagulation: macular/retinal scarring, presumed ocular histoplasmosis syndrome

d Congenital hypertrophy of the retinal pigment epithelium, choroidal nevi, fundus heterochromia

e Retinal vein occlusion: sclerosed vessels, lipemic vessels, arteriovenous nicking

f Posterior vitreous detachment, vitreous hemorrhage

g Epiretinal membrane, amelanotic neovascular membrane, fleck crystals

The study showed that the images were more than 80% adequate in quality deemed by a retina specialist’s interpretation, which allowed for the diagnosis of DR, as well as incidental findings of hypertensive retinopathy, pigmented lesions, and even optic nerve cupping.

Another study, conducted by Raman et al in India,2 compared the use of both mydriatic and nonmydriatic fundus photography for screening. The nonmydriatic technique of fundus photography was found to have sensitivity of 62.5% (95% CI 24%-91%) and specificity of 98.7% (95% CI 93%-99%), compared with indirect ophthalmoscopy. In comparison, the mydriatic technique of photography had sensitivity of 70% (95% CI 35%-93%) and a specificity of 98% (95% CI 93%-99%), compared with indirect ophthalmoscopy. The authors concluded that, using three nonsimultaneous 45º field stereoscopic fundus images, an improvement in the sensitivity of the nonmydriatic technique could be expected, but their images were of sufficient quality to screen for DR.

Greater Automation

Telemedicine has the potential to screen large numbers of patients with DR; thus, a recent study by Walton et al3 evaluated the use of automated computer algorithm-based interpretation of images via the Intelligent Retinal Imaging System (IRIS; Pensacola, FL) to screen patients with diabetes as an effective and cost-efficient way of evaluating large populations and identifying those with DR warranting referral.

In their study, the patients received teleretinal screening in primary care clinics using nonmydriatic fundus photography. The images were uploaded to a HIPAA-compliant, cloud-based server, where they were hosted and processed by IRIS. The IRIS-integrated grading module used a neural network that had a specific classifier defined and trained by a standard data set of images from the Early Treatment Diabetic Retinopathy Study. The uploaded images were compared with images from the standard data set, and the program searched for the highest probabilistic match.

Once identified, the patient’s image was classified into a “referral” or “observation” category, whereby any patient with sight-threatening diabetic eye disease, which included severe nonproliferative diabetic retinopathy or more advanced disease, was classified as a “referral.”

Additionally, an ophthalmologist in a reading center manually graded the level of retinopathy based on the ETDRS classification criteria for each image, and those patients who were graded as having severe NPDR or worse had a retinal examination within two weeks.

Based on the reading center diagnosis, the sensitivity of the IRIS algorithm in detecting severe threatening diabetic eye disease was 66.4% (95% CI, 62.8%-69.9%) with a false-negative rate of 2%. The positive predictive value of the IRIS algorithm compared with the read by the specialist in the reading center was 10.8% (95% CI <9.6%-11.9%), and the negative predictive value of the IRIS algorithm compared with the reading center was 97.8% (95% CI, 96.8%-98.6%).

For the subset of patients who were referred to a retinal specialist for a complete examination, when the algorithm was compared with the clinical ophthalmoscopic examination, the algorithm demonstrated 74.8% sensitivity (95% CI, 66.8%-81.8%), with a false negative rate of 1%. When the reading center interpretation was compared with clinical ophthalmoscopic examination, the center’s positive predictive value was 62% (95% CI, 55.4%-68.4%), and when the IRIS algorithm interpretation was compared with clinical ophthalmoscopic examination, the algorithm’s positive predictive value was 70.3% (95% CI 65.5%-75.2%).

From these findings, the IRIS computer-automated algorithm was found to be an effective screening program for identifying patients with sight-threatening diabetic eye disease. This study provided proof of the concept that an automated system could screen large patient populations and identify those patients at risk of vision loss who would otherwise not be evaluated by any eye professional.


In addition to nonmydriatric and mydiatric fundus photography, today’s smartphones have cameras with resolution of 5 megapixels and higher, enabling their users to capture high-quality images. Recent studies have shown their capabilities in acquiring high-quality fundus images with and without adapters.

Smartphones have now become useful ophthalmic devices to capture both anterior- and posterior-segment images4 (Figure 1 provides an example of an anterior and posterior adapter used to capture these images).

Figure 1. Magnifi photoadapter for iPhone ( and iExaminer PanOptic Ophthalmoscope (

Furthermore, the option of zooming on the smartphone touch screen to visualize and enlarge images instantly is an added advantage for viewing images from the patient, as well as for reviewing the images at a remote location by a trained retina specialists.5

Toy et al6 conducted a recent prospective comparative series of 100 eyes in participants undergoing ophthalmic screening for diabetic eye disease using standard in-clinic methods, compared with smartphone-assisted telemedicine methods using the Paxos Scope (DigiSight, San Francisco, CA) and the Paxos Scope Telemedicine App.

Figure 2 shows the Paxos posterior-segment adapter with the Volk (Mentor, OH) Digital Clear Field lens that was used to capture the macula spanning 45º and optic nerve images, and Figure 3 (page 38) shows the quality of the images captured.

Figure 2. Photograph of Paxos Scope posterior-segment adapter. The iPhone 5S (Apple Inc., Cupertino, CA) with external light-emitting diode (LED) illumination was attached to the Paxos Scope posterior-segment adapter.

Figure 3. Images acquired with “Paxos scope. Left image shows “Paxos” acquired fundus image on device. Right top images show diabetic retinopathy. Right lower images show no retinopathy.

Using clinical examination grading as the reference to detect referral-warranted retinopathy, the smartphone photography grade was found to be 91% sensitive and 99% specific, with a 95% positive predictive value and a 98% negative predictive value, thus showing that the images captured and reviewed with a smartphone could provide an efficient and cost-effective way of identifying patients at risk of vision loss from DR.

Another study, performed in Italy, compared smartphone ophthalmoscopy with slit-lamp biomicroscopy for grading DR using the D-Eye 5 (Pasadena, CA) adapter attached to an iPhone 5 (Apple, Cupertino, CA) (Figure 4, page 38).

Figure 4. Left image depicts the D-Eye prototype magnetically attached to the smartphone. Right images shows representative retinal images of diabetic retinopathy taken with D-Eye. Top left) Optic disc in a retina with no apparent diabetic retinopathy. Top right) Mild nonproliferative diabetic retinopathy. Bottom left) Moderate nonproliferative diabetic retinopathy. Bottom right) Panretinal photocoagulation scars in proliferative diabetic retinopathy.

The assessment of DR by a masked retina specialist using color digital images and videos obtained using the D-Eye device was in an exact agreement in 204 of the 240 eyes studied (85%), and agreement within one step was observed in 232 eyes (96.7%). Compared to biomicroscopy, the sensitivity and specificity of smartphone ophthalmoscopy for the detection of clinically significant macular edema were 81% and 98%, respectively.7

Not only has the D-Eye adapter been useful with DR screening, but it has also been used to screen and diagnose optic nerve diseases, such as glaucoma and optic nerve edema.


Teleophthalmology has already played an important role in the evaluation and screening of other retinal diseases, such as retinopathy of prematurity (ROP). The Stanford University Network for Diagnosis of Retinopathy of Prematurity (SUNDROP) is an ongoing telemedicine-based community initiative for in-hospital screening for high-risk infants for treatment-warranted ROP (TW-ROP) at six satellite neonatal intensive care units (NICUs) situated throughout Northern California.8

Based on the American Academy of Pediatrics (AAP), American Academy of Ophthalmology (AAO), and Association of Pediatric Ophthalmology and Strabismus (AAPOS) guidelines, infants were screened using RetCam II (Clarity Medical Systems, Pleasanton, CA) images, which were then sent to the Stanford University Byers Eye Institute reading center for remote interpretation by an ROP specialist.

To evaluate the diagnostic accuracy of telemedicine screening, the clinical diagnosis determined using bedside binocular ophthalmoscopy (BIO) was considered the gold standard reference, and all of the patients received at least one mandatory bedside BIO examination from a pediatric retina specialist before discharge.

Compared with bedside BIO, remote interpretation of RetCam II images had a sensitivity of 100%, specificity of 99.8%, positive predictive value of 95.5%, and negative predictive value of 100% for the detection of TW-ROP.8

The six-year results of this study confirmed that telemedicine appears to be a safe, reliable, and cost-effective way of screening and increasing access to infants with potentially sight-threatening retinopathy.


Is there a down side to teleophthalmology? We think the simple answer is no. More than 29 million people in the United States have diabetes, and there just simply are not enough eyecare professionals or retina specialists to screen every individual affected with this disease. Teleophthalmology provides a way for patients who would otherwise not have been seen by an eyecare professional to be screened and therefore diagnosed earlier and treated sooner before more permanent visual complications occur.

Medicare studies have demonstrated a cost savings of $36 to $48 per teleretinal-screened patient over unscreened patients, who are more likely to present with advanced disease and therefore require more invasive and costly treatment.7 With early detection come more referrals, better patient care, and more disease-specific referrals for the retina specialists.

In addition, the advent of smartphone funduscopy in combination with teleophthalmology may provide a platform for nonophthalmic physicians to capture images that could be remotely sent to retina specialists for review to aid in diagnosis, management, and appropriate referrals. The future is bright, but more studies will need to be performed in large-scale populations to further assess the efficacy of this technology in providing adequate screening for retinal diseases. RP


1. Chin EK, Ventura BV, See K-Y, et al. Nonmydriatic fundus photography for teleophthalmology diabetic retinopathy screening in rural and urban clinics. Telemed J e-Health. 2014;20:102-108.

2. Raman R, Rani PK, Sharma T. The sensitivity and specificity of nonmydriatic digital stereoscopic retinal imaging in detecting diabetic retinopathy. Diabetes Care. 2007;30:e47.

3. Walton OB 4th, Garoon RB, Weng CY, et al. Evaluation of automated teleretinal screening program for diabetic retinopathy. JAMA Ophthalmol. 2016;134:204-209.

4. Zvornicanin E, Zvornicanin J, Hadziefendic B. The use of smart phones in ophthalmology. Acta Informatica Medica. 2014;22:206-209.

5. Rajalakshmi R, Arulmalar S, Usha M, et al. Validation of smartphone-based retinal photography for diabetic retinopathy screening. PLoS One. 2015;10:e0138285.

6. Toy BC, Myung DJ, He L, et al. Smartphone-based dilated fundus photography and near visual acuity testing as inexpensive screening tools to detect referral warranted diabetic eye disease. Retina. 2016 Jan 19. [Epub ahead of print]

7. Russo A, Morescalchi F, Costagliola C, et al. Comparison of smartphone ophthalmoscopy with slit-lamp biomicroscopy for grading diabetic retinopathy. Am J Ophthalmol. 2015;159:360-364.

8. Wang SK, Callaway NF, Wallenstein MB, et al. SUNDROP: six years of screening for retinopathy of prematurity with telemedicine. Can J Ophthalmol. 2015;50:101-106.