Automated detection of proliferative retinopathy in clinical practice

Audrey Karperien, Herbert Jelinek, J.J.G. Leandro, J.V.B Soares, Jr. R.M. Cesar, A. Luckie

Research output: Contribution to journalArticlepeer-review

18 Downloads (Pure)

Abstract

Timely intervention for diabetic retinopathy lessens the possibility of blindness and can provide considerable cost savings to the health system. To aid automated detection of proliferative retinopathy we introduced here a method that relies on automated segmentation of retinal blood vessels using the continuous wavelet transform and determination of feature parameters using fractal analysis. Our method is the first to show that automated segmentation combined with fractal dimension as a feature parameter can distinguish between images with nonproliferative characteristics and proliferative changes in the retinal vasculature associated with diabetes, where images showing greater pathology had a significantly greater local connected fractal dimension (p=0.05). Overall, our results suggest that the local connected fractal dimension is a good index of proliferative retinopathy, and that, when coupled with automated segmentation, may prove to be an important part of developing accessible diabetic retinopathy assessment and screening.
Original languageEnglish
Pages (from-to)109-122
Number of pages14
JournalClinical Ophthalmology
Volume2
Issue number1
Publication statusPublished - 2008

Fingerprint Dive into the research topics of 'Automated detection of proliferative retinopathy in clinical practice'. Together they form a unique fingerprint.

Cite this