Microaneurysms, haemorrhages and exudates in a colour fundus image
2.3 DR Level Grading Standards
2.3.1 International Standard
Most of DR grading standards used in the DR screening programmes were based on the Early Treatment Diabetic Retinopathy Study (ETDRS) and Wisconsin Epidemiology Study of Diabetic Retinopathy publications .
In 2001, the American Academy of Ophthalmology (AAO) launched the Global Diabetic Retinopathy Project to promote the development of a common clinical severity scale for DR and DME . Later, the International Clinical Diabetic Retinopathy and Diabetic Macular Oedema Disease Severity Scales were recommended by AAO for clinical practice. The International Severity Scales proposed five levels for grading of DR and 4 levels for grading DME based on the risk of DR progression. The classification standard allowed clinically important grades of retinopathy and allowed non-ophthalmologists, as optometrists and primary care physicians, attend DR screening. The international DR and DME severity scales are summarized in Table 2.1.
International clinical DR disease and DME disease severity scales
Disease severity level
No apparent retinopathy
More than just MAs but less than severe NPDR
Any of the following:
Extensive (>20) intraretinal HMs in each of 4 quadrants
Definite VB in 2 more quadrants
Prominent IRMA in 1 more quadrant
No signs of PDR
One or more of the signs of:
DME apparently absent
No apparent retinal thickening or HEs in posterior pole
Retinal thickening and HEs in posterior pole but distant from the centre of the macula
Retinal thickening or HEs approaching the centre of the macula but not involving the centre (500 μm radius of macular centre)
Retinal thickening or HEs involving the centre of the macula (500 μm radius of macular centre)
The AAO recommended, for Type I diabetes, first eye examination should be within 3–5 years, and the follow-up should be yearly after the first examination. For Type II diabetes, annual eye examination was recommended .
2.3.2 UK NSC Standard
The UK is another country conducting large-scale and community-based DR screening programmes. From 1990 to 2002, the UK adopted the European guidelines as DR grading system . After 2003, the National Screening Committee (NSC), UK, introduced a new NSC grading system for digital photography-based DR screening . Table 2.2 gives a summary about the levels of DR and DME in the grading system. The NSC grading standard was defined based on the ETDRS grading standard. The major difference from the international standard is that for proliferative DR, two levels, R3a (active) PDR and R3s (stable) PDR, are defined. An unclassifiable (ungradable) category was also added in the standard.
UK NSC retinopathy grading standard
R0: none DR
No visible DR
R1: background DR
MAs, HMs, any exudates not with the definition of maculopathy
R2: pre-proliferative DR
VB, venous loop or reduplication, IRMA multiple deep, round or blot HMs
R3a: (active) NVD, NVE, pre-retinal or vitreous HM, pre-retinal fibrosis ± tractional retinal detachment
R3s: (stable posttreatment) evidence of peripheral retinal laser treatment and stable retina from photographs at discharge from the Hospital Eye Service
M0: no visible DME
Absence of any M1 features
Exudates within 1 DD of macular centre, circinate or group of exudates with the macula retinal thickening with 1DD of the macular centre (if stereo available), any MA or HM within 1DD of the centre of the fovea only if associated with a best VA of < = 6/12 (if stereo unavailable)
Evidence of focal/grid laser to macula, evidence of peripheral scatter laser
Another difference is the NSC standard that also defines grading pathway for DR screening. That is, firstly, all images are graded by a certificated primary grader. Then, the images will be graded by a certificated secondary grader who does not know the grading results from the primary grader. If the results from them are in disagreement, an arbitrator (retinal specialist) will grade the images and his decision will be used as the final ones.
2.3.3 Other Standards
The standards used by other countries as France, Australia, etc. are usually modified or simplified versions based on the ETDRS and Wisconsin grading systems. They are quite similar to the international standard but may have a minor difference.
Table 2.3 gives a standard recommended by the National Health and Medical Research Council, Australia.
A simplified version of the Wisconsin grading system for classifying DR (Australia)
MA and one or more of: retinal HM, HE, CWS, but not meeting moderate NPDR
HM/MA in at least one quadrant and one or more of: CWS, VB, IRMA
HM/Ma in all four quadrants, IRMA in one or more quadrants, VB in two or more quadrants
Any of: NVE, NVD, vitreous/pre-retinal haemorrhage
Any of: NVD > 1/4–1/3 disc area, or with vitreous/pre-retinal HM, or NVE > 1/2 disc area with vitreous/pre-retinal HM
High-risk PDR with traditional detachment involving the macula or vitreous haemorrhage obscuring ability to grade NVD and NVE
Retinal thickening within two disc diameters of macular centre
Clinically significant macular oedema (CSME)
Retinal thickening within 500 μm of macular centre or HEs within 500 μm of macular centre with adjacent retinal thickening
2.4 DR Screening by Morphological Means
As above-mentioned standards, morphological means based on the shape and colour characteristics of DR pathologies are common methods used for DR screening. In a DR screening programme, several key factors should be considered. They are the eye examination method, role of medical staff, pathology identification method, DR grading standard and structure of screening network.
Conventional eye fundus examination methods, such as using indirect ophthalmoscope or slit lamp biomicroscopy on dilated eyes by ophthalmologists, have been proved effective for detecting retinopathy , but direct ophthalmoscopy by GPs for DR detection are less effective . The limitation of the conventional methods is it is hard to offer a population-based DR screening. In the past decade, fundus photography, especially digital fundus imaging, has been proved be able to provide effective fundus examination as a role in DR screening [15–17]. Mydriatic fundus cameras on dilated eyes were used at the early stage. New non-mydriatic cameras without eye dilation have been commonly accepted now. Dilation is only provided for small pupil eyes if non-mydriatic imaging is not successful.
For DR screening, digital fundus camera has been proved its advantages , especially for providing excellent quality. The instant image acquisition with immediate image quality assessment can potentially reduce the technical failure rate for fundus imaging. And the digital images can make their storage and transmission much easy compared to the conventional eye fundus examinations. In the following, digital fundus imaging role in DR screening programme is emphasized.
For DR screening by using digital fundus camera, technically, a patient should undergo visual acuity measurement and fundus photography. 45° or 50° fundus cameras are commonly used for the fundus imaging in a darkened room. Several fundus images of each eye of the patient will be taken according to different DR screening standards. In the early stage of the DR screening programmes, most standards recommended five-field or three-field images (one is macula centred, one optic disc centred and one temporal to the macula) per eye [19, 20]. In the recent years, two-field imaging rule (macula centred and disc centred) has been accepted by most screening programmes .
Most screening standards recommend immediate image quality assessment after image acquisition. The French Association for the Study of Diabetes and Metabolic Diseases (ALFEDIAM) standard recommends five image quality levels: (1) excellent; (2) good definition of most retinal details; (3) definition limited, difficult to assess; (4) only gross detail visible; and (5) not gradable [19, 22]. UK NSC defines three image quality levels: (1) good, (2) adequate and (3) inadequate (ungradable). If the image quality is not satisfied, the imaging process can be repeated immediately.
Medical staff involving a DR screening programme can be simply categorized into two types. One is technician staff (or screeners) for taking fundus images by using digital fundus cameras. Through a training process, nurses, orthoptists, optometrists, etc. can fully undertake the role. Another kind of staff for DR screening is image grading staff, who undertake the task of reading fundus images, identifying pathologies and grading DR levels. Ophthalmologists are best persons for undertaking the DR grading task. However, because of lack of ophthalmologists, in a large-scale DR screening programme, other medical staff can be trained and undertake the image grading task. In UK NSC recommended screening system, the image grader’s role and image grading pathway are set up. Certificated primary and secondary graders and arbitrators (eye specialists) play the grading roles in the defined image grading workflow. There were other studies which aimed at evaluating other clinical staff, as orthoptists, optometrists and GPs, for taking the grading task. For example, the California state-wide DR screening programme has included optometrists in its image reading system, who could review the patient’s images as the same responsibility as ophthalmologists . A recent study in Spain showed that GPs after adequate training could give high-level accuracy by using non-mydriatic photography for screening diabetes patients and refer the DR patients identified to ophthalmologists . In Victoria, Australia, 45 orthoptists attended the DR image grading performance evaluation by using 36 digital images. A performance with the mean sensitivity of 86 % and specificity of 91 % for identifying abnormalities in the images was achieved . Although the studies indicated that GPs, orthoptists, etc. could potentially act as graders in DR screening programmes after good training process, there are arguments that the ophthalmologist’s role in DR image grading should not be reduced.
Image grading process usually includes identifying pathologies on colour fundus images and grading DR levels according to DR grading standards. Morphological features related to DR, as MAs, HMs, EDs, VB and NVs, are commonly identified by an image grader in an image reading phase. DR disease level can be graded according to the currently recommended DR grading standards, such as the international DR disease severity scales, UK NSC standard, etc. The screening results sent back to the GPs or screeners can include the severity of DR and advice for the referral/no referral of a patient for further examination or treatment or a follow-up examination in a defined interval. Basic criteria of a DR screening programme should target at providing sufficiently high sensitivity (true positive rate) and specificity (true negative rate). The high sensitivity can ensure that the patients with DR (as DR level above moderate NPDR or sight-threatening retinopathy) are not missed for referral and treatment. The high specificity, on the other hand, can ensure that the diabetes patients without reaching the level of referral are not referred. The UK NSC grading standard recommends 80 % sensitivity and specificity for DR screening programme .
Several solutions can be used to construct a DR screening network. One solution is to locate the fundus cameras in fixed community centres and let them be operated by trained health providers for image scanning. Another solution is to mount the fundus cameras on mobile vans for serving remote or rural areas. These two solutions also can be combined as a hybrid service model. A grading centre linked with the screening sites play an important role of image grading. Data transmission between the grading centre and screening sites with an efficient and secure mode becomes especially important for providing instant information feedback. With the universal application of the Internet in remote and rural areas, Internet-based patient information transmission can bridge the grading centre and screening sites efficiently. Especially, a specially designed tele-ophthalmology system will provide great convenience for a DR screening programme .
A tele-ophthalmology, which links a grading centre to its surrounding DR screening sites, can easily implement a DR screening workflow as the steps of: (1) the primary health providers scan patient’s fundus images and transmit the images with the patient’s other medical information and eye examination results to the grading centre ; (2) the eye grading experts (ophthalmologists or graders) read the images and grade the images according to a grading standard; (3) the grading results are transmitted back to the corresponding screening sites; and (4) the health providers inform the results to the patient and instruct for follow-up examination or treatment. In a tele-ophthalmology system, technically, the efficiency of the transmission of image is one of important factors. The original image size of the fundus image can be 1 Mbytes to several Mbytes depending on the resolution and dimensions of the image. Image compression method can be applied for the convenience of the image transmission through the Internet.
The telemedicine service model for DR screening programme can greatly save the ophthalmologists’ or graders’ time and let them provide efficient services to the patients living in a vast area. This service model has been implemented in some countries.
2.5 DR Screening Programmes in the World
In France, the French National Health Authority recommended an annual examination of the eyes on diabetes patients without examination before or no mild retinopathy . Two-field image scanning (macular centred and disc centred) was recommended. The French ALFEDIAM standard, which is quite similar to the international DR severity scales, was used for DR level classification .
In early 1990s, the DR screening programmes in France were practiced by using slit lamp biomicroscopy on dilated eyes. In France, the lack of qualified manpower (absence of optometrists) was a key problem for performing a wide DR screening, especially organizing an efficient and cost-effective DR screening programme . In February 2002, the first DR screening centre by using telemedicine technique was set up in Paris. The orthoptists in the centre used a non-mydriatic fundus camera to take images for diabetes patients referred by GPs. The images were sent to a grading centre for DR grading. In 18 months, a total of 912 DR examinations were performed . In June 2004, an ophthalmological diabetes telemedicine network “Ophdiat” was created in the Île-de-France area . Seven hospitals, 11 primary healthcare centres and 2 prisons joined the DR screening programme. An average of seven ophthalmologists in the grading centre supported the DR grading task. From June 2004 to December 2009, 38,596 patients (in 51,741 examinations) were screened. Of the screening examinations, 13,726 (26.55 %) were referred to ophthalmologists. Through the screening programme, the telemedicine system has shown its reliability for providing one grading centre and multiple screening sites service model in a vast area.