Aberdeen Biomedical Imaging Centre
Research Area: Ophthalmic Imaging
The programme has two main research streams. The first concentrates on the development of instrumentation and the second on the automated analysis of retinal images.
The instrumental work is mainly concentrated on the development of the scanning laser ophthalmoscope (SLO). We have shown how this instrument can be used for full colour imaging of the fundus, by imaging the retina with several lasers simultaneously. In addition to work on the clinical value of colour, we are also exploring the measurement of the physiological status of retinal tissue by spectral imaging, recording the amount of light reflected at different wavelengths.
A recent development is the adaptation of the SLO to perform fluorescence lifetime imaging (FLIM). With this technique we look at how the characteristics of the chromophore is affected by its environment. Pilot work has shown that we can produce images with the SLO using FLIM in which we accurately measure the changes in fluorescent lifetime. In the future we aim to develop the device to perform in vivo imaging, which will require the development of suitable chromophores for use with humans.
The SLO has also been developed to image blood cell dynamics in the retina of rats and mice. This allows quantitative measurements to be made of cell velocity in arteries, veins and capillaries, the extent of cellular adhesion to vessel walls in terms of cell rolling and sticking efficiencies. In a collaboration with the university department of ophthalmology, the technique has produced many papers primarily exploring the regulation mechanisms of leukocytes in inflammatory eye disease. Research is currently being done to develop a new instrument to image the dynamics of several different populations of cells simultaneously.
In the second strand of our research programme we have been developing software to analyse retinal images for the presence of early signs of diabetic retinopathy. In 2003 the SEHD decided to set up a national programme in Scotland to screen all people with diabetes for early signs of eye disease. We have completed a trial to study the effectiveness of using the software routinely in the screening programme. This evaluation involved the largest trial to date with 14000 images from consecutive patients referred for screening. This has demonstrated that software can be used to first quality control the fundus images and then analyse them for the presence of eye disease, reducing the workload of manual screeners by 60%. The health services analysis carried out as part of the trial showed an annual saving in operating costs of £200,000. Discussions are in progress to commercialise the software. A new project has started to develop the software to recognise different types of eye disease. This will further reduce the number of images that need analysing by humans.
We are now using our expertise in software development for retinal image analysis on projects in two related areas. Digital photography for the detection of sight-threatening macular disease, as conducted in a screening programme, has its limitations. Only 10% of patients with photographic criteria for referral are subsequently found to have sight-threatening macular oedema. As a consequence scarce ophthalmology service resources are being wasted. The emerging gold standard for macular oedema is optical coherence tomography (OCT), but this imaging modality is impractical for screening large numbers of patients. We are developing novel software to identify the most specific digital photographic markers likely to predict sight threatening macular oedema, using optical coherence tomography as the cross reference standard.
Stroke is the third most common cause of death in the developed world. A recent US population-based study attributed approximately 40% of all ischaemic strokes to the effects of diabetes alone or in combination with hypertension. The retina shares many anatomical and physiological characteristics with the cerebral circulation, and changes in the retina have been shown to be an independent stroke risk factor. Retinal arteriolar abnormalities, such as arteriovenous nicking and focal arteriolar narrowing, may reflect arteriolar damage from diabetes, hypertension or a combination of the two. Detecting retinal arteriolar abnormalities may enable clinicians to identify people with known risk factors, such as diabetes and hypertension, who are at higher risk of stroke and institute preventive measures. We aim to develop an automated system for detecting these retinal arteriolar abnormalities.