Fractal and multifractal analysis of two dimensional images of the retinopathy of prematurity: Morphologic parameters and classification

Retinopathy of prematurity is a disease caused by abnormal vessel regression of the retinal vasculature that affects premature infants with low birth weights. This paper explores the possibility of applying the concept of fractal analysis to images of retinal blood vessels for morphometric analysis and classification. More than 100 neonates were screening with RetCam 3 device by the experienced ophthalmologist who chose 101 suitable images and set them into three groups, according to the presence of the ROP and the stage of ROP development. Black and white binary images were carried out by box dimension quantifying the space-filling, shape and tortuosity of retinal blood vessels. Accordingly, grey-scale images were investigated by the lacunar and multifractal analysis, which quantifies heterogeneity and multifractal spectra, respectively. The results indicate that the gestational age of premature births did not influence the box dimension, homogeneity and multifractal spectra. The blood vessels in the retinae with severe retinopathy of prematurity are more heterogeneous, complex and have a higher degree of vessel aberration than those vessels with infants without retinopathy. Multifractal analysis, however, did show significant differences between only between those two stages of the retinopathy. We have demonstrated that objective parameters for morphometric analysis and classifying of retinopathy can be found by applying the fractal and multifractal analyses. The next step in developing might be to establish a clinical standard based on fractal and multifractal measures. Another application could be the development of software that can be implemented in standard ophthalmologic imaging.