Ophthalmic Surgery, Lasers and Imaging Retina

Clinical Science 

Validation of Smartphone-Based Retinal Photography for Diabetic Retinopathy Screening

Yannick Bilong, MD; Jean-Claude Katte, MD; Godefroy Koki, MD; Giles Kagmeni, MD; Odile Pascale Nga Obama, MD; Hermann Rossi Ngoufo Fofe, MD; Caroline Mvilongo, MD; Oliver Nkengfack, MD; Andre Michel Bimbai, MPH; Eugene Sobngwi, MD, PhD; Wilfred Mbacham, DS, ScD; Jean Claude Mbanya, MD, PhD, FRCP; Lucienne Assumpta Bella, MD; Ashish Sharma, MD

Abstract

BACKROUND AND OBJECTIVE:

Screening for diabetic retinopathy (DR) is cost-effective when compared with disability loss for those who go blind in the absence of a screening program. We aimed to evaluate the sensitivity and specificity of a smartphone-based device for the screening and detection of DR.

PATIENTS AND METHODS:

A cross-sectional study of 220 patients with diabetes (440 eyes, all patients age 25 years or older) was completed. Tropicamide 0.5% was used for iris dilation followed by an indirect ophthalmoscopy using a 20-D lens. Retinal images were later obtained using a smartphone attached to an adaptable camera device. Retinal images permitted the visualization of the macular and papillary regions and were sent without compression via the internet to a retinal specialist for interpretation. Sensitivity and specificity were calculated for all cases and stages of DR.

RESULTS:

Using our standard examination method, the prevalence of DR and macular edema were 13.6% and 6.4%, respectively. With the smartphone-based retinal camera, the prevalence of DR and macular edema were 18.2% and 8.2%, respectively. Sensitivity and specificity for the detection of all stages of DR was 73.3% and 90.5%, respectively. For the detection of macular edema, sensitivity was 77.8%, and specificity was 95%. For severe nonproliferative DR (NPDR), sensitivity and specificity were 80% and 99%, respectively; for proliferative DR (PDR), they were both 100%. In the early stages of DR, specificity was 89.8% for mild NPDR and 97.1% for moderate NPDR. Sensitivity was 57.1% and 42.9%, respectively.

CONCLUSION:

Screening for DR using a smartphone-based retinal camera has a satisfactory specificity at all DR stages. Its sensitivity seems to be high only in the stages of DR necessitating a specific therapeutic decision (eg, macular edema, severe NPDR, and PDR). A smartphone-based retinal camera may be a useful device to screen for DR in resource-limited settings.

[Ophthalmic Surg Lasers Imaging Retina. 2019;50:S18–S22.]

Abstract

BACKROUND AND OBJECTIVE:

Screening for diabetic retinopathy (DR) is cost-effective when compared with disability loss for those who go blind in the absence of a screening program. We aimed to evaluate the sensitivity and specificity of a smartphone-based device for the screening and detection of DR.

PATIENTS AND METHODS:

A cross-sectional study of 220 patients with diabetes (440 eyes, all patients age 25 years or older) was completed. Tropicamide 0.5% was used for iris dilation followed by an indirect ophthalmoscopy using a 20-D lens. Retinal images were later obtained using a smartphone attached to an adaptable camera device. Retinal images permitted the visualization of the macular and papillary regions and were sent without compression via the internet to a retinal specialist for interpretation. Sensitivity and specificity were calculated for all cases and stages of DR.

RESULTS:

Using our standard examination method, the prevalence of DR and macular edema were 13.6% and 6.4%, respectively. With the smartphone-based retinal camera, the prevalence of DR and macular edema were 18.2% and 8.2%, respectively. Sensitivity and specificity for the detection of all stages of DR was 73.3% and 90.5%, respectively. For the detection of macular edema, sensitivity was 77.8%, and specificity was 95%. For severe nonproliferative DR (NPDR), sensitivity and specificity were 80% and 99%, respectively; for proliferative DR (PDR), they were both 100%. In the early stages of DR, specificity was 89.8% for mild NPDR and 97.1% for moderate NPDR. Sensitivity was 57.1% and 42.9%, respectively.

CONCLUSION:

Screening for DR using a smartphone-based retinal camera has a satisfactory specificity at all DR stages. Its sensitivity seems to be high only in the stages of DR necessitating a specific therapeutic decision (eg, macular edema, severe NPDR, and PDR). A smartphone-based retinal camera may be a useful device to screen for DR in resource-limited settings.

[Ophthalmic Surg Lasers Imaging Retina. 2019;50:S18–S22.]

Introduction

Diabetic retinopathy (DR) is a major microvascular complication of diabetes and one of the top causes of vision loss globally.1 Despite the growing burden of DR, it seems to have been neglected in health-care research and planning in many low-income countries, where access to trained eye care professionals and tertiary eye care services may be inadequate.2 The majority of patients with diabetes who develop DR are asymptomatic until the very late stages. Therefore, screening and early intervention are critical milestones in the management of DR.3 Screening for DR is cost-effective when compared with the disability loss for those who go blind in the absence of a screening program.4 Different modalities of screening exist depending on the number of available ophthalmologists and other trained health care professionals, equipment, and resources available for screening, which are usually technically cumbersome and expensive.5 Mydriatic seven-field photography and clinical fundus examination are considered to be the gold standards for DR screening, but more innovative and comprehensive approaches are needed to reduce the risk for vision loss.6 Alternative screening innovations such as handheld fundoscopy are available, and the relative low cost of smartphones with cameras make an attractive platform for both image acquisition, transmission, and interpretation.7–9 The Make In India Retinal Camera (MII Ret Cam; MII Ret Cam Inc, Coimbatore, TN, India), when coupled with a smartphone and 20-D lens, can capture high quality retinal photos and videos of the posterior pole up to the ora serrata within a short learning curve for the user.10 Posterior pole imaging can be performed within a week of practice as per our experience (unpublished data). Given that this device is an inexpensive tool that can be useful for mass screening of retinal pathologies, we therefore decided to screen for DR in adult patients with type 2 diabetes and to evaluate the sensitivity and specificity of this tool in the diagnosis of DR.

Patients and Methods

This was a cross-sectional study carried out at the National Obesity Centre of Yaoundé Central Hospital in Cameroon from August 2017 to January 2018.

Ethical clearance was obtained from the Centre Regional Ethics Committee for Human Research in Health (CE No 0696/CRERSHC/2017) at the Centre Regional Delegation for Public Health.

Participants were known patients with type 2 diabetes attending outpatient consultations at the National Obesity Centre of Yaoundé Central Hospital. Inclusion criteria were an age older than 25 years, no known allergy to tropicamide eye drops, readiness to provide a written informed consent, and a recent glycated hemoglobin value at fewer than 3 months before inclusion. Participants with overt cataract were not included in the study.

All participants underwent a complete physical and eye examination. Physical examination included anthropometric and biophysical measurements. Ocular examination consisted of a visual acuity test using an illuminated Snellen chart, followed by indirect ophthalmoscopy and then retinal imaging using the MII Ret Cam coupled with a smartphone and 20-diopter lens after pupil dilation.

Indirect ophthalmoscopy was performed using a headset ophthalmoscope with the participant in a supine position at the margin of the examination table and at an arm-distance away from the examiner. The anterior chamber of the eye was examined, followed by the posterior chamber. This examination permitted the complete visualization of the retina, and a diagnosis was made based on the presence or the absence of retinopathy. The diagnosis made after indirect ophthalmoscopy was considered to be the gold standard diagnosis.

The MII Ret Cam is designed with slots to fit a smartphone (a built-in camera and flash) and a lens. We used an iPhone 5s (iOS 10.3.3; Apple, Cupertino, CA) and a 20-D lens. The device is held close to the eye, and once the fundus glow is visualized through the dilated pupil, the device is slowly brought back until a clear image can be visualized in the smartphone camera. At least two photographs per eye focusing on the macula and papilla were taken (Figure 1). They were immediately stored into the memory of the smartphone against the personal data of the already-registered patient data in the MII Ret Cam application memory.

The MII Ret Cam smartphone device. (a) Body positioning to obtain a retinal image after pupil dilation. (b) Image focus on the macula. (c) Image focus on the papilla.

Figure 1.

The MII Ret Cam smartphone device. (a) Body positioning to obtain a retinal image after pupil dilation. (b) Image focus on the macula. (c) Image focus on the papilla.

The retinal images obtained were sent by email (unzipped) with the patient data code for interpretation to a separate ophthalmologist who was masked to the clinical diagnosis. Image quality was ascertained according to the clarity of the second-order retinal vessels (good quality: good view of vessels; moderate quality: blurred vessels; poor quality: no view of vessels).

Retinopathy grading was based on the International Classification of Diabetic Retinopathy scale.11 Each eye was graded separately (Figure 2). The overall retinopathy diagnosis per patient was determined by the diagnosis of the worst eye. Diabetic macular edema was defined as the presence of hard exudates within one disc diameter of the center of the macula.

Examples of diabetic retinopathy (DR) stages after interpretation of the retinal image. (a) Minimum to moderate nonproliferative DR (NPDR). (b) Moderate NPDR and macular edema. (c) Proliferative DR.

Figure 2.

Examples of diabetic retinopathy (DR) stages after interpretation of the retinal image. (a) Minimum to moderate nonproliferative DR (NPDR). (b) Moderate NPDR and macular edema. (c) Proliferative DR.

The sample size was calculated from a prevalence of DR of 17%, with a type 1 error of 5% and a confidence interval set at 95%.

Sensitivity and specificity were calculated from the numbers of DR cases screened using the MII Ret Cam smartphone device compared with indirect ophthalmoscopy. The degree of agreement between indirect ophthalmoscopy and MII Ret Cam smartphone images was calculated using k statistics. A P value of less than .05 was considered statistically significant for all analyses.

Results

A total of 220 patients (440 eyes) with type 2 diabetes were recruited to participate in this study, all of whom completed both retinal examinations. Women constituted the majority of the study population, (n = 139, 63.2%). The mean age was 57.7 years ± 10.2 years, and the mean duration of diabetes was 7.9 years ± 6.9 years (Table 1).

General and Biophysical Characteristics of the Study Population.

Table 1:

General and Biophysical Characteristics of the Study Population.

Results are presented as frequencies or mean and standard deviation.

Image quality was good in 3.4% (15 eyes), moderate in 85.2% (375 eyes), and poor in 11.4% (50 eyes). The prevalence of DR in the study population was 13.6% (n = 30) and 18.2% (n = 40) with indirect ophthalmoscopy and the MII Ret Cam, respectively. Among those with DR on indirect ophthalmoscopy, 33.3% (n = 10) had mild nonproliferative DR (NPDR), 43.3% (n = 13) had moderate NPDR, 13.3% (n = 4) had severe NPDR, and 10% (n = 3) had proliferative DR (PDR). The MII Ret Cam revealed that among the 40 participants who presented with DR, 57.5% (n = 23) had mild NPDR, 32.5% (n = 13) had moderate NPDR, 2.5% (n = 1) had severe NPDR, and 7.5% (n = 3) had PDR.

On indirect ophthalmoscopy, the prevalence of macular edema was 6.4% (n = 14), whereas it was 8.2% (n = 18) with use of the MII Ret Cam and the smartphone.

Sensitivity increased with DR severity, whereas specificity was good in all DR stages (Table 2). The degree of agreement was fairly good among all stages of DR.

Sensitivity and Specificity of the MII Ret Cam

Table 2:

Sensitivity and Specificity of the MII Ret Cam

Discussion

The MII Ret Cam tended to overdiagnose DR compared with the standard method. The possibility of artifacts on the retinal images may have influenced the reading and interpretation of the retinal images obtained using the MII Ret Cam, therefore leading to a higher prevalence rate compared with the result obtained from indirect ophthalmoscopy.

This study shows that the prevalence of DR using indirect ophthalmoscopy and the MII Ret Cam plus a smartphone device was 13.6% and 18.2%, respectively. These prevalence rates are lower than that reported by other studies in Cameroon.12–17 These studies report a prevalence of DR from 21.7% to 49.9% depending on the method of investigation. For the most part, these studies used mydriatic seven-field fundus photography, which is the gold standard of care in the detection of DR. The low rates of DR can be accounted for by the limited number of patients in our study compared with other studies in Cameroon and also can be alluded to the near-normal diabetes control as shown by the mean glycated hemoglobin of 7.2% in this study population. The mean glycated hemoglobin in all other studies was greater than 9%, and poor glycemic control has been shown to be a major risk factor for DR.16 Other well-established risk factors of DR are age and diabetes duration.18 Mean diabetes duration in our study was 7.9 years, with the majority of patients having a total duration of fewer than 5 years (48.6%). The low duration of diabetes in this study compared with other studies and the relatively young age of our study population are possibly additional factors contributing to the lowered rates of DR.

The effectiveness of any screening tool is assessed by its sensitivity and specificity. The sensitivity and specificity of retinal images using the MII Ret Cam smartphone device was 73.3% and 90.5%, respectively, for the diagnosis of DR and 77.8% and 95.0%, respectively, for macular edema. Similar studies using other smartphones devices have shown varied degrees of sensitivity and specificity depending on the techniques used and the definition of DR stages.19–21 The Fundus on Phone system (Remidio Innovative Solutions Pvt Ltd, Bangalore) versus the Zeiss retinograph (FF450 Plus Digital Fundus Camera; Carl Zeiss Meditec, Dublin, CA) showed a sensitivity and specificity of 92.7% and 98.4%, respectively.19 A smartphone with the macrolens adapter and external light source versus a slit lamp exam showed a sensitivity and specificity of 91% and 99%.20 The Cell-Scope Retina (San Francisco, CA) versus a slit lamp exam showed a sensitivity and specificity of 95.1% and 48.7%, respectively, in the diagnosis of DR.21 The sensitivity and specificity of the retinal images from the MII Ret Cam smartphone device in detecting PDR were 100% and 100%, respectively. This can be explained by the fact that the lesions seen during PDR are overt. However, the number of participants with PDR in our study population was low.

To our knowledge, this is the first study to evaluate and validate the MII Ret Cam smartphone device in the detection of DR. Assessment of the retinal images from the device was performed by an independent reader who did not take part in the clinical evaluation of the patients, hence limiting any eventual bias.

The main limitation of this study remains the lack of comparison to the mydriatic seven-field stereoscopic retinal imaging technique, which is the gold standard in the detection and diagnosis of DR. However, this study serves as a pioneer to portray the usefulness and clinical utility of this innovative and less expensive smartphone-aided imaging technique, which can be adapted to our setting to improve the early screening and referral of patients with DR for appropriate management. The higher number of moderate quality retinal images in this study can be attributed to poor pupil dilation in patients with diabetes, who often present a failure of the sympathetically-mediated pupil dilation mechanism.22 However, appropriate training similar to any other hand-held device is necessary to improve the quality of images obtained and augment the sensitivity and specificity in the detection and diagnosis of all stages of DR. The possibility to use this technique and images obtained in teleophthalmology also is a dimension to be explored both in clinical research and in health service delivery plans.

References

  1. Kempen JH, O'Colmain BJ, Leske MC, et al. The prevalence of diabetic retinopathy among adults in the United States. Arch Ophthalmol. 2004;122:552–563. doi:10.1001/archopht.122.4.552 [CrossRef]
  2. Zheng Y, He M, Congdon N. The worldwide epidemic of diabetic retinopathy. Indian J Ophthalmol. 2012;60:428–431. doi:10.4103/0301-4738.100542 [CrossRef]
  3. Vieira-Potter VJ, Karamichos D, Lee DJ. Ocular complications of diabetes and therapeutic approaches. Biomed Res Int. 2016;2016:3801570. doi:10.1155/2016/3801570 [CrossRef]
  4. Vashist P, Singh S, Gupta N, Saxena R. Role of early screening for diabetic retinopathy in patients with diabetes mellitus: an overview. Indian J Community Med. 2011;36:247–252. doi:10.4103/0970-0218.91324 [CrossRef]
  5. Lau HC, Voo YO, Yeo KT, Ling SL, Jap A. Mass screening for diabetic retinopathy--a report on diabetic retinal screening in primary care clinics in Singapore. Singapore Med J. 1995;36:510–513.
  6. Bragge P, Gruen RL, Chau M, Forbes A, Taylor HR. Screening for presence or absence of diabetic retinopathy: a meta-analysis. Arch Ophthalmol. 2011;129:435–444. doi:10.1001/archophthalmol.2010.319 [CrossRef]
  7. Kumar S, Wang E-H, Pokabla MJ, Noecker RJ. Teleophthalmology assessment of diabetic retinopathy fundus images: smartphone versus standard office computer workstation. Telemed J E Health. 2012;18:158–162. doi:10.1089/tmj.2011.0089 [CrossRef]
  8. Prescott G, Sharp P, Goatman K, et al. Improving the cost-effectiveness of photographic screening for diabetic macular oedema: a prospective, multi-centre, UK study. Br J Ophthalmol. 2014;98:1042–1049. doi:10.1136/bjophthalmol-2013-304338 [CrossRef]
  9. Suto S, Hiraoka T, Okamoto Y, Okamoto F, Oshika T. Photography of anterior eye segment and fundus with smartphone [in Japanese]. Nippon Ganka Gakkai Zasshi. 2014;118:7–14.
  10. Sharma A, Subramaniam SD, Ramachandran KI, Lakshmikanthan C, Krishna S, Sundaramoorthy SK. Smartphone-based fundus camera device (MII Ret Cam) and technique with ability to image peripheral retina. Eur J Ophthalmol. 2016;26:142–144. doi:10.5301/ejo.5000663 [CrossRef]
  11. ICO Guidelines for Diabetic Eye Care. International Council of Ophthalmology Website. http://www.icoph.org/downloads/ICOGuidelinesforDiabeticEyeCare.pdf. Updated January 2017. Accessed August 10, 2017.
  12. Nanfack C, Koki G, Mbuagbaw L, Bella AL. Diabetic retinopathy at the Yaoundé Central Hospital in Cameroon: epidemiology and angiographic findings. Pan Afr Med J. 2012;13:54.
  13. Jingi AM, Noubiap JJN, Ellong A, Bigna JJR, Mvogo CE. Epidemiology and treatment outcomes of diabetic retinopathy in a diabetic population from Cameroon. BMC Ophthalmol. 2014;14:19. doi:10.1186/1471-2415-14-19 [CrossRef]
  14. Sobngwi E, Mbanya JC, Moukouri EN, Ngu KB. Microalbuminuria and retinopathy in a diabetic population of Cameroon. Diabetes Res Clin Pract. 1999;44:191–196. doi:10.1016/S0168-8227(99)00052-2 [CrossRef]
  15. Moukouri EN, Moli TM, Nouedoui C. Epidemiological aspects of diabetic retinopathy in Yaounde [in French]. Médecine d'Afrique Noire. 1992;3905:327–334.
  16. Jingi AM, Nansseu JRN, Noubiap JJN, Bilong Y, Ellong A, Mvogo CE. Diabetes and visual impairment in sub-Saharan Africa: evidence from Cameroon. J Diabetes Metab Disord. 2015;14:21. doi:10.1186/s40200-015-0151-4 [CrossRef]
  17. Jivraj I, Ng M, Rudnisky CJ, et al. Prevalence and severity of diabetic retinopathy in Northwest Cameroon as identified by teleophthalmology. Telemed J E Health. 2011;17: 294–298. doi:10.1089/tmj.2010.0155 [CrossRef]
  18. Solomon SD, Chew E, Duh EJ, et al. Diabetic retinopathy: a position statement by the American Diabetes Association. Diabetes Care. 2017;40(3):412–418. http://care.diabetesjournals.org/content/diacare/40/3/412.full.pdf. Accessed January 10, 2018. doi:10.2337/dc16-2641 [CrossRef]
  19. Rajalakshmi R, Arulmalar S, Usha M, et al. Validation of smartphone based retinal photography for diabetic retinopathy screening. PLoS ONE. 2015;10:e0138285. doi:10.1371/journal.pone.0138285 [CrossRef]
  20. 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;36: 1000–1008. doi:10.1097/IAE.0000000000000955 [CrossRef]
  21. Kim TN, Myers F, Reber C, et al. A smartphone-based tool for rapid, portable, and automated wide-field retinal imaging. Transl Vis Sci Technol. 2018;7:21. doi:10.1167/tvst.7.5.21 [CrossRef]
  22. Kwon HJ, Kim HY. A pharmacologic pupillary test in the diagnosis of diabetic autonomic neuropathy. Korean J Ophthalmol. 2009;23:291–295. doi:10.3341/kjo.2009.23.4.291 [CrossRef]

General and Biophysical Characteristics of the Study Population.

CharacteristicTotalMenWomenP Value
No.22081139
Mean age (y)57.7 ± 10.257.6 ± 10.257.7 ± 10.2.95
Mean duration of diabetes (y)7.9 ± 6.98.5 ± 6.67.6 ± 7.1.36
Mean weight (kg)80.6 ± 15.484.0 ± 15.478.0 ± 15.0.01
Mean body mass index (kg/m2)30.1 ± 7.428.6 ± 5.031.0 ± 8.3.01
Abdominal circumference (cm)98.8 ± 10.599.3 ± 11.098.5 ± 10.2.57
Mean systolic blood pressure (mm Hg)130 ± 18133 ± 19127 ± 16.01
Mean diastolic blood pressure (mm Hg)77 ± 1080 ± 1176 ± 9.003
Mean glycemia (mg/dL)143 ± 61150 ± 61138 ± 60.16
HbA1c (%)7.2 ± 0.67.2 ± 0.57.2 ± 0.6.74

Sensitivity and Specificity of the MII Ret Cam

Status of RetinopathySensitivity (%)Specificity (%)K AgreementP Value
Any DR73.390.50.56< .001
Mild NPDR57.189.80.53< .001
Moderate NPDR42.997.10.54< .001
Severe NPDR80.099.00.64< .001
Proliferative DR1001001.0< .001
Diabetic macular edema77.895.00.59< .001
Authors

From the Department of Ophthalmology, (YB, GK, GK, CM, LAB), Department of Public Health (JCK), Department of Internal Medicine and Specialties (ES, JCM), Faculty of Medicine and Sciences, University of Yaounde 1, Cameroon; Mbalmayo District Hospital (OPNO), Ebolowa Regional Hospital (HRNF), Tibati District Hospital (ON), Ministry of Public Health, Cameroon; Department of Public Health, Faculty of Medicine, Paris IV University, Paris, France (AMB); Laboratory for Public Health Research and Biotechnologies, Biotechnology Center, University of Yaounde 1, Cameroon (WM); and Lotus Eye Hospital and Institute, Coimbatore, Tamilnadu, India (AS).

Presented October 25, 2018, at the Ophthalmology Times Retina Research Scholar Meeting in Chicago, IL.

Supported by an unrestricted grant from the French Society of Ophthalmology.

Dr. Sharma has financial and proprietary interest in the MII Ret Cam device. The remaining authors report no relevant financial disclosures.

Address correspondence to Yannick Bilong, MD, Department of Ophthalmology, Faculty of Medicine and Biomedical Sciences, P.O. Box 1364, University of Yaounde 1, Cameroon; email: bilongyan@yahoo.fr.

Received: January 09, 2019
Accepted: February 05, 2019

10.3928/23258160-20190108-05

Sign up to receive

Journal E-contents