Fractal analysis of region-based vascular change in the normal and non-proliferative diabetic retina.

PURPOSE Evaluation of normal and abnormal vascular pattern in the human retina using a novel method: quantitative region-based fractal analysis. METHODS Binary (black/white) vascular patterns of the human retina originating at the optic disc were obtained by semi-automatic computer processing of digital images from 60-degree fundus fluorescein angiography of 5 normal eyes and 5 eyes with non-proliferative diabetic retinopathy (NPDR). As determined by image resolution, vascular patterns included vessels with diameters >or=50 microm and excluded small vessels and capillaries. The density of linearized (i.e., skeletonized) vascular patterns in the macular region versus paramacular region (termed "region-based" linearized vascular pattern) was quantified with the fractal dimension (D(f)) and confirmed by grid intersection (rho(v)). RESULTS By region-based quantification, D(f) and rho( v) were significantly higher in the normal macular region than in the NPDR macular region (p = 0.008 and p = 0.019, respectively). However, differences in D(f) and rho(v) between the normal and NPDR paramacular regions were not strongly significant (p = 0.168 and p = 0.337, respectively). CONCLUSIONS Results from the retrospective analytical study demonstrate the feasibility of using quantitative region-based fractal analysis of early-stage vascular disease in the human retina. The results are encouraging for a broader study of diverse patient populations.