Comparison of Portable Cameras for Diabetic Retinopathy Community Screening

Diabetic retinopathy (DR) is a leading cause of blindness in functioning ages.^1,2

Regular retinal evaluation among patients with diabetes and effective screening strategies could improve the ability to detect vision-threatening DR. Risk of severe visual loss by early diagnosis of DR sign and timely treatment could be decreased by more than 90%.³ A variety of handheld and portable retino graphs have been introduced to improve community screening and early diagnosis of retinal diseases in remote settings, but none of them have been experienced in Iran. We aimed to compare the accuracy of two portable cameras and one smartphone adaptor for selection of the appropriate tele-screening instrument for DR screening in the community. This study was approved by the Institutional Review Board of Tehran University of Medical Sciences (TUMS35052).

Based on the camera protocols, both undilated and dilated modes were used to take images with two handheld cameras (Horus DEC 200; MiiS, Hsinchu, Taiwan and Welch Allyn RetinaVue (TM) 100 Imager, Skaneateles Falls, NY, USA) and the dilated mode was used for iPhone-based adaptor (Volk iNview, Volk Optical Inc., USA). The images were taken by one health care professional (nonphysician) and independently reviewed by a retina specialist and one trained general practitioner and compared blinded with slit lamp biomicroscopy using +90 D lens in the dilated status as gold standard.

After obtaining informed consent, a total of 250 patients with diabetes (500 eyes) who referred to a diabetes center were included. Exclusion criteria were previous history of DR treatment including laser or any other surgical treatment, advanced corneal opacity or cataract in one or both eyes, and any media opacity that might affect the quality of the images such as active uveitis. The mean age of the participants was 58.2 ± 12.43 years. Moreover, 96.4% had type 2 diabetes, and 59% of patients didn’t report any history of eye examination.

Total images rated as gradable in 93%; 87% of nondilated version of RetinaVue and Horus, respectively; and 93%, 97%, and 22% of postdilation for RetinaVue, Horus, and iNview, respectively. Of ungradable images, the majority (88%) were taken with the iNview, so this device was excluded from grading. A sensitivity of 98%-100% and specificity of 80%-99% were reported for referable patients in pre- and postdilation images for Horus and RetinaVue, respectively. The Kappa coefficient between the general practitioner and the specialist was 65%, 62% for Horus and RetinaVue, respectively.

Based on questionnaire, Horus was more user friendly (ease of use, low failure rate) and comfortable for both operator and patients. According to the original formula (cost/accuracy), Horus was more cost effectiveness than the RetinaVue. Detailed of clinical evaluation has been summarized in Table 1.

Table 1.

Clinical and Quality Evaluation by Cameras and Pupil Status.

Cameras	Sensitivity %	Specificity %	PPVa%	NPVb%	Image quality score %	Kappa coefficient %c	Accuracy %	Satisfaction score (0-10)
Horus Predilation	100	80	98	100	Poor: 13.3 Fair: 26.7 Excellent: 60	65	96	9 (operator)-7 (patients)
Horus Postdilation	98	99	94	96	Poor: 3.3 Fair: 30 Excellent: 66.7		97
RetinaVue Predilation	100	80	98	100	Poor: 7.4 Fair: 52.6 Excellent: 40	62	98	8 (operator)-6 (patients)
RetinaVue postdilation	99	96	96	96	Poor: 7.1 Fair: 39.3 Excellent: 53.6		98

^aPositive predictive value.

^bNegative predictive value.

^cFor agreement on referral cases between the general practitioner and the vitreoretinal specialist.

Sensitivity and specificity of Horus and RetinaVue were almost the same nearly 100% (post dilation), and there was no significant difference in the sensitivity and specificity of cameras before and after pupil dilation. Considering the adequate of images quality, ease of use by nonophthalmologists, and the appropriate accuracy, they could be an appropriate alternative to overcome access barriers in underserved communities.