An accurate unsupervised extraction of retinal vasculature using curvelet transform and classical morphological operators
Feudjio Ghislain,Saha Tchinda Beaudelaire,Tchiotsop Daniel
DOI: https://doi.org/10.1016/j.compbiomed.2024.108801
IF: 7.7
2024-06-26
Computers in Biology and Medicine
Abstract:Background Many ophthalmic disorders such as diabetic retinopathy and hypertension can be early diagnosed by analyzing changes related to the vascular structure of the retina. Accuracy and efficiency of the segmentation of retinal blood vessels are important parameters that can help the ophthalmologist to better characterize the targeted anomalies. Method In this work, we propose a new method for accurate unsupervised automatic segmentation of retinal blood vessels based on a simple and adequate combination of classical filters. Initially, contrast of vessels in retinal image is significantly improved by adding the Curvelet Transform to commonly used Contrast-Limited Adaptive Histogram Equalization technique. Afterwards, a morphological operator using Top Hat is applied to highlight vascular network. Then, a global threshold-based Otsu technique using minimum of intra-class variance is applied for vessel detection. Finally, a cleanup operation based on Match Filter and First Derivative Order Gaussian with fixed parameters is used to remove unwanted or isolated segments. We test the proposed method on images from two publicly available STARE and DRIVE databases. Results We achieve in terms of sensitivity, specificity and accuracy the respective average performances of 0.7407, 0.9878 and 0.9667 on the DRIVE database, then 0.7028, 0.9755 and 0.9507 on the STARE database. Conclusions Compared to some recent similar work, the obtained results are quite promising and can thus contribute to the optimization of automatic tools to aid in the diagnosis of eye disorders.
computer science, interdisciplinary applications,engineering, biomedical,biology,mathematical & computational biology