Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Providing full screening of diabetic patients for retinopathy throughout the community would significantly reduce the incidence of blindness in this group of patients. However, although panretinal photocoagulation is of proved value in reversing or preventing neovascular complications and careful macular treatment prevents further visual loss, the resources or the infrastructure to detect appropriate patients for treatment are not universally available.1 The importance of the situation has been highlighted by the declaration of St Vincent, a European directive on the need to provide screening.2 3 An effective screening and treatment programme would ultimately reduce the burden on health and social service budgets because fewer of these relatively young patients would require long term support.4-6
Locally based efforts indicate that the key to diabetic retinopathy screening is efficiently obtaining images of the retina for classification. This raises two problems—firstly, the target population must be reached at minimum inconvenience to the patient and, secondly, the images must be assessed effectively. Care facilities for diabetic patients are spread throughout the community and so it is most appropriate that their retinas be examined there. Unfortunately, the expertise for assessment of diabetic retinopathy is not always available locally. Therefore systems involving fundus photography with later assessment of the image by a specialist have been employed but these use relatively expensive photographic prints, slides, or Polaroids with inefficient manual transport of images.
The technological revolution has provided us with new methodologies that we should exploit to solve these problems. For example, it is now possible to digitise images and send these through telemedicine—for example, email, with the advantages of computerised storage and image manipulation. Although commercial fundus cameras can be used to obtain and store digitised image data their database, storage systems do not easily lend themselves to data analysis (a problem that the companies involved should address otherwise investigators will be forced to design their own systems). Digital cameras are available and are becoming cheaper. Also, competition in the market for scanning laser ophthalmoscopes is increasing. Cost savings are likely if the cameras can be placed in the community7 with transfer of the digitised images through various methods. Compression of the images allows faster data transfer and reduces storage requirements but the methods are diverse requiring investigators to decide carefully the best ones. This is a difficult area because as anyone who has bought the latest PC computer knows your “pride and joy” is out of date within a few years such is the speed of development of computer technologies. The data must also be confidential and so not all methods of data transfer are appropriate.
Potentially, digitised data can be analysed by computer either at the site of acquisition of the image or after transfer to a centralised “sorting” area. If the images are processed by computer the need for assessment by technical or ophthalmological staff is removed with associated reductions in cost.8 The ultimate goal would involve automated acquisition of the photographic image, digitisation, and analysis without any intervention by technical staff. The computer would then indicate to the patients that they require to attend an ophthalmology department. Currently, artificial neural network (ANN) computer programs are capable of discriminating normal fundi from those with retinopathy, potentially reducing the numbers of images requiring expert examination by 70% or more.9 10 The application of such programs in the field has yet to be assessed but the rapid progress in PC technology is likely to increase the ability to diagnose automatically fundus images. Continued development of ANN programs is likely to be a fruitful area of research especially if the tasks are relatively simple—for example, whether diabetic retinopathy is present or not. More complex image analysis such as grading of retinopathy into background, preproliferative, and proliferative will be much more difficult to achieve. New methods involving parallel ANNs,11 real time “on chip” video signal processing,12 or the use of workstation computers may help.
Although services for diabetic screening are fragmented existing facilities could be adapted to digitising methods with relatively small initial investment. Pilot schemes should be commenced investigating the most appropriate technologies available while allowing some adaptability for any new developments that may occur. Already, departments in Europe are involved in this area of research. These should be brought together to share their experiences to date, to reduce duplication of effort, and, hopefully, to produce a coordinated approach to the problem. If digitisation of diabetic images proves practical, as is likely, investment on a national scale may be necessary to produce the infrastructure required. Remember that care of diabetic patients already consumes approximately 8% of the UK national healthcare budget.13 Effective computerised screening could potentially reduce that burden while bringing major benefits to patients.
Merely browsing the technical sections of the broadsheet newspapers shows how fast digital cameras, data storage, and data transfer are progressing, largely fuelled by the needs of the video and music industry. Progress is relentless and provides us with opportunities to solve practical problems in the management of our patients. Surely on the back of such developments we will be able to find a solution for a logistical problem such as diabetic retinopathy screening.
Supported by British Diabetic Association Grant No x88202.