16 Feb 2015
Ophthalmic problems increase in frequency with age and common, age-related, systemic diseases, including renal failure, hyperthyroidism and diabetes mellitus, may cause retinal abnormalities – notably vascular changes, associated with hypertension (Kobayashi et al, 1990; Crispin and Mould, 2001).
Thus, examination of the fundus, comprising the optic disc, the tapetal and non-tapetal fundus and the retinal vasculature, is important in geriatric screening (Brown et al, 2007; La Croix, 2005; Leblanc, 2011; Maggio, 2000; Maggs et al, 2008; Michell, 2010).
Direct ophthalmoscopy is the method of choice for retinal examination in first opinion veterinary practice (Mould, 1993) although indirect ophthalmoscopy gives a wider field of view of the fundus and improved optics in the presence of opacity. Whichever technique is used it can sometimes be difficult to fully visualise the fundus in elderly cats and dogs during routine screening and no permanent record exists for future evaluation.
In humans, the value of direct ophthalmoscopy in detecting early hypertensive changes has been questioned as there is high inter-observer agreement (70 per cent to 80 per cent) for late signs of haemorrhage and retinal detachment, but for early signs it is only moderate (30 per cent to 40 per cent; Dimmitt et al, 1989; Van den Born et al, 2005) so other diagnostic techniques, including retinal photography, are increasingly being used (Bennett and Barry, 2009).
Diabetes mellitus is common in dogs and cats (Bennett, 2002; Rand and Marshall, 2005), but diabetic retinopathy (a major cause of visual impairment in humans) is rarely recognised although it is reported to be similar in pathogenesis (Bloodworth and Molitor, 1965; Hatchell et al, 1995). Secondary cataracts occur in up to 66 per cent of diabetic dogs (Beam et al, 1999), obscuring the retina, and retinopathy may be present following cataract removal (Landry et al, 2004). Feline diabetic retinopathy has only been reported as experimentally induced in cats (Hatchell et al, 1995) and the extent of retinal pathology was directly related to the duration of diabetes.
Stored digital images can be sent to remote specialists (telemedicine) for evaluation (Boucher et al, 2008; Cavallerano and Conlin, 2008) and they allow for comparison of sequential images to monitor disease progression. Retinal photography is rarely done in first opinion veterinary practice, but this may change with the advent of portable, hand-held digital fundus cameras designed specifically for veterinary use (Hoang et al, 2001).
A study was performed by Lucy Dobree to assess the practicality of retinal imaging for geriatric screening. To be a useful tool the procedure needs to be easy to perform in a reasonable time and produce diagnostic images of sufficient quality, such that abnormalities can be identified locally or by remote ophthalmologists.
Dogs and cats more than nine years of age were recruited from three populations – a veterinary hospital and two animal shelters. A retinal camera designed for veterinary use (Optibrand) was used to acquire the images. The camera software was downloaded on to a laptop and the camera attached via a USB cable. Real-time images were displayed on the laptop screen while the photographer positioned the camera at the correct distance (about 25mm) from the patient’s cornea by focusing two camera-generated lights on the iris. Light intensity was set reasonably low to reduce pupillary constriction as a mydriatic was not used. The camera captures images at 19 frames per second and a continuous sequence of images was autosaved for as long as the image capture button was depressed. It was possible to capture photographs without assistance, but for a few non-compliant animals, the owner or an assistant provided restraint. The photographer held the camera with one hand while using the other hand to retract the eyelids to prevent blinking.
Images were reviewed to ensure all four components of the retina had been adequately captured in one or more images and they were evaluated for abnormalities locally by reference to a standard veterinary ophthalmology atlas (Millichamp and Dziezyc, 2004).
Images considered the best were emailed to six veterinary ophthalmologists, three RCVS diploma holders and three certificate holders, who were asked to score the images on a three-point scale as being of either no diagnostic quality (1), limited diagnostic quality (2) or good diagnostic quality (3). Further comments were also invited.
The camera was easily portable and could be held from above or below – allowing the photographer to keep handto-muzzle contact at a minimum. At no point did the photographer have face-to-muzzle contact with the animal. The software was easy to use.
Images from 31 animals met the study inclusion criteria. The fundus of three eyes could not be photographed, but 1,508 photographs in total were taken and 71 images from 59/62 eyes were sent for remote evaluation. In 71 per cent (22/31) of cases, one image was sent for assessment from each eye; for the remaining 29 per cent (9/31), two or more images were needed to see all the structures. There was no significant difference (p=0.605) in diagnostic image quality scores between the cases for which multiple images were sent and those with just one image. In 96.8 per cent (30/31) of cases, images were acquired within 15 minutes and only one dog took longer.
A wide range of diagnostic quality scores was attributed to the images by the veterinary ophthalmologists (Figure 1). In total, 44 per cent (31/71) of images were found to be of limited diagnostic quality, 25 per cent (18/71) of good diagnostic quality, but 31 per cent (22/71) of images were considered non-diagnostic. Images that contained all four retinal components were sometimes scored of no diagnostic quality, while images that did not contain all components were sometimes scored as good diagnostic quality.
Of the animals included in the study, 26 per cent (8/31) were dogs and 74 per cent (23/31) cats, which are small sample sizes so conclusions about breed differences could not be made. However, diagnostic image quality was significantly different between the species (p=0.02) with images of the feline fundus being scored of better diagnostic quality (canine average 1.78, feline average 1.99). Individually, ophthalmologists all scored feline fundus images higher and of better diagnostic quality (Figure 2), but overall the images of both species were of limited diagnostic quality.
Four of the six veterinary ophthalmologists (VO 1 to VO 6) chose to report abnormalities seen on the retinal images: VO 3 reported 8/31, VO 4 – 7/31, VO 5 – 13/31 and VO 6 – 4/31. Abnormalities included changes consistent with hypertension (4/31), retinal detachment (1/31), retinal degeneration (3/31), feline central retinal degeneration consistent with taurine deficiency in two non-related 18-year-old cats (Bellhorn et al, 1974; Knopf et al, 1978) and optic nerve pathology (1/31). Examples can be seen in Figure 3.
There was no significant difference in scoring for VO 3, VO 5 and VO 6 (p=0.39, p=0.32 and p=0.085 respectively) between the images in which they recognised pathology compared to images they considered normal. However, VO 4 gave images with abnormalities recognised in significantly higher scores (p<0.001) than images assessed as normal.
Diploma holders scored the images significantly lower than the certificate holders (p<0.001; Figure 4).
There was little agreement on interpretation between ophthalmologists; 98 per cent (70/71) of the images were not scored the same by all six veterinary ophthalmologists and there was a highly significant relationship between the individual veterinary ophthalmologist and the scores they gave (Figure 5).
Potential advantages of fundus imaging over direct ophthalmoscopy are reduced risk of injury because face-to-muzzle contact does not occur and the ability to capture images of the full fundus for subsequent assessment in the practice or by remote ophthalmologists through telemedicine.
A large number of images (31 per cent) considered adequate by the lead author were not considered diagnostic by the ophthalmologists. Nevertheless, most images (69 per cent) provided some diagnostic information, so this study supports medical literature suggesting mydriatics may not be needed in screening programmes (Tarabishy et al, 2011), although image quality may have been improved with their use (Baeza et al, 2009).
There was high variability in the interpretation of the images, but the three-point scoring system (although based on a system used in human medicine) was not independently validated – potentially resulting in ambiguity. Emphasis on factors such as image sharpness, blur or lighting, compared to ability to visualise the main retinal components, could have affected scoring between observers.
Diagnostic-quality retinal images can be taken by people with any level of training (Maberley et al, 2004) and training in retinal photography software is considered more important than clinical knowledge (Farley et al, 2008), so the poor quality of many images acquired in this study may reflect a need for further training (Ruamviboonsuk et al, 2006). However, the problems caused by conscious animals moving or flicking their eyes is a main reason for poor quality due to movement blur.
Variation in opinion between the lead author and ophthalmologists about the diagnostic quality of the images is of concern as this could frustrate the deployment of telemedicine across the veterinary profession. Poor images would require further photography with additional cost and time. This situation could be improved by using a real-time teleophthalmology system, whereby the remote ophthalmologist could advise the camera operator, as is being deployed in medicine (Bai et al, 2007; Wu et al, 2010).
The use of reference manuals to identify abnormalities in retinal photographs is an accepted practice in human medicine (Castro et al, 2007); however, even medical specialists require further training to correctly grade diabetic retinopathy to reduce variability in opinion (Ruamviboonsuk et al, 2006) and automated image reading software is now being deployed as it is better at determining whether images are good or very good quality compared to ophthalmologists (Bartling et al, 2009).
Various factors could explain why feline retinal images scored higher than canine images, including the large difference in sample sizes (eight dogs to 23 cats) and minimal variation between cats in skull size, shape and nose length compared to large variations seen between dogs of different breeds.
Identifying retinal abnormalities was not a primary objective of this study, but abnormalities were seen in this cohort of older animals, confirming ophthalmic examination is useful in a geriatric screen and that digital images obtained, even if only of limited diagnostic quality, allow new diagnoses to be made.
Further veterinary studies are needed to validate a grading system for the diagnostic quality of captured retinal images, to determine whether routine mydriatic use improves image quality, and to directly compare findings from direct ophthalmoscopy with digital image interpretation. A survey of first opinion vets to assess their confidence in ophthalmoscopy would be of interest and a study into the reasons for such wide inter-observer variation in image interpretation by veterinary ophthalmologists could result in the development of protocols to reduce this.
The camera and software were easy to use and generated at least partially diagnostic images for 69 per cent of dogs and cats in this study without the use of a mydriatic, and for all bar one case within 15 minutes. However, 31 per cent of images considered of adequate diagnostic quality by the lead author were not thought so by independent ophthalmologist observers so clear guidelines are needed for inexperienced imagers to define diagnostic quality.
There was wide variability in interpretation of image quality and recognition of abnormalities present by the ophthalmologists, suggesting further training or standardisation of image reporting may be needed. First opinion vets cannot rely on standard interpretation of images distributed through a telemedicine system so should consider sending images to more than one ophthalmologist for opinion.
Retinal abnormalities were identified in 13/31 (42 per cent) of cases, confirming ophthalmic examination is useful in a geriatric screen and that digital images obtained, even if only of limited diagnostic quality, allow new diagnoses to be made.
This study suggests retinal imaging may be a useful tool in geriatric screening, but further studies are needed to determine whether imaging is more sensitive than direct or indirect ophthalmoscopy in practice.
The authors would like to thank Oakham Veterinary Hospital and its clients, the University of Nottingham School of Veterinary Medicine and Science, Heidi Featherstone, Jane Sansom, Elaine Holt, Peter Wilson, Greg Firth, the RSPCA and Cats Protection.
Figure 1. Images representing the three levels of diagnostic image quality as scored by six veterinary ophthalmologists.
Figure 2. Comparison of canine and feline retinal image diagnostic quality. The chart shows the average scores for diagnostic quality given by each veterinary ophthalmologist (VO 1 to VO 6) and the average of all scores for the two species (VO average).
Figure 4. This chart shows the average score diploma holding and certificate holding veterinary ophthalmologists gave the images.
Figure 5. How six veterinary ophthalmologists (VO 1 to VO 6) scored the images for diagnostic quality. One (blue) indicates an image of no diagnostic quality, two (red) indicates an image of limited diagnostic quality and three (green) indicates an image of good diagnostic quality.
