Quantification of diabetic maculopathy by digital imaging of the fundus

Abstract

Measurement of the extent of diabetic retinopathy is an essential part of assessing the efficacy of local or systemic treatment regimens. Current clinical studies use empirical grading of retinopathy which is performed by a trained observer using standard photographs. This method is relatively arbitrary, as well as time consuming and vulnerable to observer error. We have developed a digital fundus imaging system and image processing programs which provide objective, quantitative measures of macular oedema, retinal exudates, and microaneurysms in diabetic retinopathy. Using fluorescein angiograms, the degree of macular oedema is quantified both in terms of area of fundus involved and severity of oedema by analysis of the temporal changes in intensity of fluorescence. Fluorescein angiograms are also used for the detection and counting of microaneurysms, by a combination of shade correction, matched filtering, and shape algorithms. For detection and measurement of retinal exudates, a colour transparency projected through a red free filter is analysed using a combination of shade correction and thresholding techniques. The system described is in clinical use, and has potential for a wide variety of applications. With further development, digital analysis of fundus images should supercede the currently used manual semi-quantitative methods, providing faster, more accurate, objective quantitative results.