Competing/conflict of interest: Professor Yogesan Kanagasingam invented the EyeScan (Ophthalmic Imaging System, CA, USA) but he is not/has not received any direct/indirect funding from the manufacturer.
Light and portable novel device for diabetic retinopathy screening
Article first published online: 23 DEC 2011
© 2011 The Authors. Clinical and Experimental Ophthalmology © 2011 Royal Australian and New Zealand College of Ophthalmologists
Clinical & Experimental Ophthalmology
Volume 40, Issue 1, pages e40–e46, January/February 2012
How to Cite
Ting, D. S., Tay-Kearney, M. L. and Kanagasingam, Y. (2012), Light and portable novel device for diabetic retinopathy screening. Clinical & Experimental Ophthalmology, 40: e40–e46. doi: 10.1111/j.1442-9071.2011.02732.x
Funding sources: Diabetes Australia Research Trust and Royal Perth Hospital have provided research funding to this project. The sponsor or funding organization had no role in the design or conduct of this research.
- Issue published online: 5 FEB 2012
- Article first published online: 23 DEC 2011
- Accepted manuscript online: 2 NOV 2011 03:32AM EST
- Received 2 June 2011; accepted 12 September 2011.
- diabetic retinopathy;
- diagnostic technique
Background: To validate the use of an economical portable multipurpose ophthalmic imaging device, EyeScan (Ophthalmic Imaging System, Sacramento, CA, USA), for diabetic retinopathy screening.
Design: Evaluation of a diagnostic device.
Participants: One hundred thirty-six (272 eyes) were recruited from diabetic retinopathy screening clinic of Royal Perth Hospital, Western Australia, Australia.
Methods: All patients underwent three-field (optic disc, macular and temporal view) mydriatic retinal digital still photography captured by EyeScan and FF450 plus (Carl Zeiss Meditec, North America) and were subsequently examined by a senior consultant ophthalmologist using the slit-lamp biomicroscopy (reference standard). All retinal images were interpreted by a consultant ophthalmologist and a medical officer.
Main Outcome Measures: The sensitivity, specificity and kappa statistics of EyeScan and FF450 plus with reference to the slit-lamp examination findings by a senior consultant ophthalmologist.
Results: For detection of any grade of diabetic retinopathy, EyeScan had a sensitivity and specificity of 93 and 98%, respectively (ophthalmologist), and 92 and 95%, respectively (medical officer). In contrast, FF450 plus images had a sensitivity and specificity of 95 and 99%, respectively (ophthalmologist), and 92 and 96%, respectively (medical officer). The overall kappa statistics for diabetic retinopathy grading for EyeScan and FF450 plus were 0.93 and 0.95 for ophthalmologist and 0.88 and 0.90 for medical officer, respectively.
Conclusions: Given that the EyeScan requires minimal training to use and has excellent diagnostic accuracy in screening for diabetic retinopathy, it could be potentially utilized by the primary eye care providers to widely screen for diabetic retinopathy in the community.
The worldwide prevalence of diabetes is estimated to double to 366 million (4.8%) by 2030.1 People with diabetes should receive regular eye screening, as early detection could prevent diabetic retinopathy-related visual impairment.2–4 Retinal still photography remains the mainstay screening tool in various diabetic retinopathy screening programs worldwide, and it has been shown to increase the primary eye care providers' confidence and desire to detect diabetic retinopathy in the community.5 To date, mydriatic retinal photography has been shown to be the most effective means to detect diabetic retinopathy in a screening setting.6,7
The rising prevalence of diabetes will require implementation of more community screening programs. This will lead to a rise in health-care costs8 and maintenance of an effective quality assurance system.9 The cost of a retinal camera is and will continue to be an important factor in the health-care cost equation. Numerous retinal cameras (mydriatic and non-mydriatic) with various functions, including anterior and posterior segment imaging and fluorescein angiography, are currently available on the market. However, many are not inexpensive.
The purpose of our study was to validate the efficacy of the EyeScan (Ophthalmic Imaging System, CA, USA) machine to screen for diabetic retinopathy in the community. The EyeScan is a flash camera that weighs around 0.9 kg. Apart from retinal imaging, this device also captures the anterior segments of the eye, such as the cornea and the lens, and this is especially important in determining the causes of some ‘ungradable’ retinal still images secondary to cataracts or pterygium. It has a maximum image capture rate of three images per second and a 5.3-megapixel Sensor. It is able to capture the photos with near infrared or visible light and has a field of view of 35 degrees. This machine was approved by the US Food and Drug Administration in 2010 and is currently available in United States, Australia and other European countries. Given that this is an extremely portable device, which can be carried around in a suitcase, it will be a suitable imaging device in routine, mobile and teleretinal diabetic retinopathy screening clinics in both metropolitan and rural areas.
This single-centre study aimed to validate and compare a new diagnostic device, EyeScan, with the currently accepted diagnostic device FF450 plus (Carl Zeiss Meditec, Inc., North America), with reference to slit-lamp examination by a consultant ophthalmologist.
We enrolled 136 consecutive patients (272 eyes) from the diabetic retinopathy screening clinic of Royal Perth Hospital, Western Australia, into our study. All patients signed an informed consent for participation. This study was approved by the Royal Perth Hospital Ethics Committee.
Patients' characteristics and diabetes history
We collected information on patients' demographics (e.g. age and ethnicity), current and past ocular history, diabetes history [e.g., type, duration, glycated hemoglobin level, macro- and microvascular complications (cerebrovascular accidents, ischaemic heart disease, peripheral vascular disease, nephropathy and neuropathy)] and other associated cardiovascular risk factors (e.g. smoking status, blood pressure and lipid profile).
Upon arrival at the screening clinic, all patients received pupil-dilating drops (2.5% phenylephrine and 0.5% tropicamide) in both eyes. They underwent three sets of retinal examinations in the following order: (i) non-stereo colour retinal still photography (FF450 plus); (ii) non-stereo colour retinal still photography (EyeScan) and (iii) slit-lamp biomicroscopy examination with a 78-dioptre lens by a senior consultant ophthalmologist.
Retinal still photography using the EyeScan and FF450 plus was performed by a medical officer and a retinal photographer, respectively. In order to assess the usability of the EyeScan device, a medical officer with no previous experience in ocular imaging was recruited and compared to the retinal photographer who has had 10 years experience in performing retinal still photography for diabetic retinopathy screening. Three retinal fields (optic disc, macula and temporal views) were captured using both devices, and the images were subsequently de-identified, randomized and interpreted by a consultant ophthalmologist and a medical officer (who has graded more than 1000 colour fundus photos of patients with diabetes) on a 27-in. iMac (Apple, Cupertino, CA, USA) with a display resolution of 2560 × 1440 pixels in a dimly lit room.
The retinal digital still images of the EyeScan and FF450 plus were all downloaded in Joint Photographic Experts Group format. The colour resolution of the still images of EyeScan and FF450 plus were 640 × 480 and 2392 × 1944 pixels, respectively. The field angles of EyeScan and FF450 plus were 35 and 30 degrees, respectively, centring on optic disc, macular and temporal views.
The retinal images were graded on the basis of the presence or absence of diabetic retinopathy signs (microaneurysms, retinal haemorrhages, hard exudates, cotton wool spots, venous beading, intraretinal microvascular abnormalities, new vessel formation and preretinal/vitreous haemorrhage) using the International Clinical Diabetic Retinopathy Severity Scale (Table 1). The retinal photographs were classified as ‘unacceptable’, ‘average’ or ‘excellent’ depending on their quality; the retinal photograph was graded as ‘unacceptable’ if more than one-third of it was ‘blurred’ or ‘uninterpretable’.
|Mild NPDR||Microaneurysms only|
|Moderate NPDR||More than just microaneurysms but less than severe NPDR|
|Severe NPDR||Any of the following:|
|i. Extensive (>20) intraretinal haemorrhages in each of four quadrants|
|ii. Definite venous beading in 2+ quadrants|
|iii. Prominent IRMA in 1+ quadrant|
|AND no signs of PDR|
|PDR||One or more of the following:|
|ii. Vitreous/preretinal haemorrhage|
We calculated the sensitivity and specificity of the two imaging devices (EyeScan and FF450 plus) in detecting and grading diabetic retinopathy with reference to slit-lamp examination. In addition, Cohen's kappa coefficient was utilized as a measure of agreement for diabetic retinopathy signs and grading using the two types of imaging devices. The technical failure rate was defined as the fraction of the ‘unacceptable’ retinal images captured by both devices; such images were excluded from the calculation for sensitivity, specificity and kappa coefficient. All data were analysed using Statistical Package for Social Sciences version 17 (SPSS, Chicago, IL, USA).
Patients' demographics and clinical characteristics
A total of 136 patients (272 eyes) participated in our study. The mean ± standard deviation age of participants was 53.9 ± 15.3 years, duration of diabetes was 13.9 ± 9.9 years, and glycated hemoglobin was 8.0 ± 1.7%. Among the recruited patients, 74% (n = 101) were Whites, 17% (n = 23) were Asians, and 9% (n = 12) were from other ethnic groups. Of these, 96 patients (71%) had type 2 diabetes. Figure 1 shows the colour fundus images captured by EyeScan and FF450 plus.
The best-corrected visual acuity of 240 eyes (88%) was 6/6 or 6/9, and that of 23 eyes (9%) was between 6/12 and 6/36, and that of nine eyes (3%) was 6/60 or less. Of the consecutively recruited eyes, nearly 35% had diabetic retinopathy ranging from mild non-proliferative diabetic retinopathy to proliferative diabetic retinopathy (Table 2). Nearly 15% (n = 37) of eyes had previously received panretinal photocoagulation, and cataracts were diagnosed in 28 eyes (10.3%) on the basis of slit-lamp biomicroscopy examination using Lens Opacities Classification III.11 Almost 45% (n = 118) of the patients had never undergone any diabetic retinopathy screening. Of the self-reported diabetes-related complications, diabetic neuropathy (23%, n = 62) and nephropathy (22%, n = 60) were the leading complications ( Table 3).
|Diabetic retinopathy severity||Eyes (n)||%|
|Mild non-proliferative diabetic retinopathy||51||18.8|
|Moderate non-proliferative diabetic retinopathy||26||9.6|
|Severe non-proliferative diabetic retinopathy||8||2.9|
|Proliferative diabetic retinopathy||6||2.2|
|Diabetes complications||Patients n (%)|
|Ischaemic heart disease||23 (17)|
|Peripheral vascular disease||13 (10)|
|Cerebrovascular disease||9 (7)|
Main outcome measure
Compared with the slit-lamp biomicroscopy examination, EyeScan, in detecting any grade of diabetic retinopathy, had a sensitivity and specificity of 93% (95% confidence interval [CI]: 84.9–97.1) and 98.2% (95% CI: 94.3–99.5), respectively, when graded by the ophthalmologist; however, they were 91.7% (95% CI: 83.2–96.3) and 94.7% (95% CI: 89.9–97.4), respectively, for the medical officer (Table 4). In contrast, FF450 plus images graded by the ophthalmologist had a sensitivity and specificity of 95.1% (95% CI: 87–98.4) and 98.8% (95% CI: 95.4–99.8), respectively, whereas for the medical officer, he had a sensitivity and specificity of 91.9% (95% CI: 83.4–96.4) and 95.9% (95% CI: 91.5–98.2), respectively. For the detection of sight-threatening diabetic retinopathy (severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy), the sensitivity and specificity of images from both devices (EyeScan and FF450 plus) graded by both readers increased to 100%.
|(95% CI)||(95% CI)|
The technical failure rate for EyeScan and FF450 plus were 8.5 and 7%, respectively, and they were not statistically significant (χ2 = 0.23, d.f. = 1, P = 0.63). Of the failed retinal photographs captured by the EyeScan, 39% (n = 9) were photographs of eyes with cataracts and 9% (n = 2) with dark fundi; in 52% (n = 12) of the photographs, failure was due to intolerance to bright flash. On the other hand, the ‘uninterpretable’ Zeiss retinal images were secondary to cataracts (42.1%, n = 8), dark fundi (10.5%, n = 2) and intolerance to bright flash (47.4%, n = 9).
The overall kappa statistics for diabetic retinopathy grading for EyeScan and FF450 plus were 0.93 and 0.95 for ophthalmologist and 0.88 and 0.90 for medical officer, respectively (Table 4). The kappa coefficients for all diabetic retinopathy signs, except macular oedema, based on the analysis of EyeScan and FF450 plus images by both readers, with reference to the slit-lamp biomicroscopy examination, were more than 0.8 (Table 5). The kappa coefficients for the ophthalmologist in detecting diabetic maculopathy using EyeScan and FF450 plus were 0.70 and 0.74, respectively, whereas for the medical officer, they were 0.71 and 0.76, respectively.
|Retinal findings||Kappa statistics|
|Retinal photography||Retinal photography|
|Cotton wool spots||1||1|
|Intraretinal microvascular abnormalities||0.93||0.93|
|New vessels formation||1||1|
|Cupped optic disc||1||1|
|Cotton wool spots||1||1|
|Intraretinal microvascular abnormalities||0.76||0.76|
|New vessels formation||1||1|
|Cupped optic disc||0.87||0.89|
In our study, we utilized the International Clinical Diabetic Retinopathy Severity Scale (Table 1)10 as the grading system as it is much more simplified, with less severity levels and diagnostic criteria, as compared with the Early Treatment Diabetic Retinopathy Study classification system (Table 6).12 In addition, the slit-lamp examination was chosen as the reference standard, given that it has been shown to be compared favourably with the seven-field stereoscopic 30 degrees Early Treatment Diabetic Retinopathy Study.13 In addition, it is easy to perform, less time consuming and more tolerable especially when all patients had to undergo two sets of retinal still imaging on EyeScan and FF450 plus. Should the quality of the retinal still images were compromised due to the presence of cataracts or pterygiums (encroaching the visual axis), the slit-lamp examination by an ophthalmologist would be more accurate than the retinal still images captured using the seven-field Early Treatment Diabetic Retinopathy Study.
|Diabetic retinopathy stage||Retinal findings|
|Minimal NPDR||MA and one or more of the following: retinal haem, Hex, CWS, but not meeting the criteria for moderate NPDR|
|Moderate NPDR||H/Ma > std photo 2A in at least one quadrant and one of more of: CWS, VB, IRMA, but not meeting severe NPDR|
|Severe NPDR||Any of:|
|H/Ma > std photo 2A in all four quadrants,|
|IRMA > std photo 8A in one or more quadrants,|
|VB in two or more quadrants|
|NVE or NVD < std photo 10A, vitreous/preretinal haem|
|NVE < 1/2 DA without NVD|
|High-risk PDR||Any of:|
|NVD > 1/4 to 1/3 disc area,|
|or with vitreous/preretinal haem,|
|or NVE > 1/2 DA with vitreous/preretinal haem|
|Advanced PDR||High-risk PDR with tractional detachment involving macula or vitreous haem obscuring ability to grade NVD and NVE|
In this study, we have shown that with reference to the slit-lamp examination, both readers had comparable sensitivity and specificity in grading diabetic retinopathy from retinal photographs captured using the EyeScan machine or FF450 plus (Table 4). Both devices had a sensitivity and specificity of 100% in detecting sight-threatening diabetic retinopathy changes. Additionally, the kappa statistics of EyeScan and FF450 plus for the detection of all diabetic retinopathy signs, except for macular oedema, were more than 0.8 (Table 5). Because the images obtained from EyeScan showed excellent sensitivity, specificity and kappa coefficients in diagnosing diabetic retinopathy, it could be used as reliably as currently used cameras for the screening of diabetic retinopathy.
The colour fundus images from both devices graded by the ophthalmologist and medical officer had kappa coefficients of less than 0.8 for detecting diabetic maculopathy (Table 5). Because retinal photographs provide only two-dimensional views of the retina, colour fundus images do not afford easy identification of any retinal thickening or macular oedema. In this study, we diagnosed diabetic maculopathy on the basis of the presence of hard exudates, microaneurysms and retinal haemorrhages close to the macular area. Presence of these lesions in the macula region should always prompt an urgent referral.
In order to evaluate the usability of EyeScan, a medical officer with no previous experience in ocular imaging was recruited as the operator of the device. He received a single day's training on the device prior to the screening. The technical failure rate of the FF450 plus operated by an experienced retinal photographer (10 years of experience on retinal still photography for diabetic retinopathy screening) and the EyeScan were similar (8.5% vs. 7%, P > 0.05). These figures show that EyeScan may be used by non-experienced personnel with minimal training. Further studies will be of great value in evaluating the user-friendliness of the EyeScan for both medical and non-medical personnel in the primary health-care and teleophthalmology setting.
In our study, assessments by both readers had excellent sensitivity, specificity and kappa coefficients in detecting and grading diabetic retinopathy using the colour fundus images captured by both devices (Tables 4,5). These results indicate that non-specialist personnels, such as primary care physicians and allied health personnels, can be trained to screen for diabetic retinopathy in the primary health-care setting. The EyeScan will be a suitable device to be utilized in a teleophthalmology setting as it requires small storage capacity for its lower resolution files, and thus, the available bandwidth would better handle the transfer and archives of the images form community to centralized screening centre. Further research is required to evaluate the cost effectiveness of using Eyescan in a community and teleophthalmology setting using primary eye care providers, such as optometrists, general practitioners, orthoptists and diabetes nurses, to screen for diabetic retinopathy. The ophthalmologist will not be able to service the projected increase in numbers of patients with diabetes. Optimizing the use of people resources, embracing new and affordable technology will be necessary to effectively screen these patients and therefore reduce the impact of diabetic retinopathy-related visual impairment.
The strength of our study was that all recruited patients underwent colour fundus photo-imaging by using both devices (EyeScan and FF450 plus), and this allowed a head-to-head comparison between the two devices. In addition, we have utilized two different statistical methods (sensitivity/specificity and Cohen's kappa statistics) and two readers (an ophthalmologist and a medical officer) to increase the reliability and validity of our study findings. In contrast, our study carries a few limitations, one of which was that the slit-lamp examination (reference standard) was performed by a single senior consultant ophthalmologist. In addition, all patients successively underwent two sets of retinal still photography within a short period of time, and this could have contributed to the technical failure that had occurred as a consequence of patients' intolerance to bright light from both the devices.
- 12Early Treatment Diabetic Retinopathy Study Research Group. Grading diabetic retinopathy from stereoscopic color fundus photographs – an extension of the modified Airlie House classification. ETDRS report number 10. Early Treatment Diabetic Retinopathy Study Research Group. Ophthalmology 1991; 98: 786–806.