Sandpiper Optimization Algorithm With Region Growing Based Robust Retinal Blood Vessel Segmentation Approach

Retinal blood vessel examination is commonly utilized for retinal disease diagnosis by ophthalmologists. The automated retinal vessel segmentation process becomes an essential tool to identify disease. Several retinal vessel segmentation models suffer from a lack of high generalization abilities and low accuracy due to the presence of complex symmetrical and asymmetrical patterns. Robust vessel segmentation of fundus images is needed to improve diagnostic performance including vein occlusion and diabetic retinopathy (DR). In this aspect, this study concentrates on the design of a sandpiper optimization algorithm with region growing based robust retinal blood vessel segmentation (SPORG-RBVS) approach. The proposed SPORG-RBVS technique involves different stages of pre-processing such as grayscale conversion, Z-score-based data normalization, and multi-scale vessel enhancement filtering. The SPO approach addresses the intricate challenges modeled by difficult symmetrical and asymmetrical patterns in retinal vessel segmentation. This method has been specifically designed to improve the generalization capabilities and accuracy of retinal vessel segmentation manners, vital for the precise detection of retinal diseases like vein occlusion and DR. Through phases of preprocessing comprising grayscale conversion, Z-score-based data normalization, and multi-scale vessel enhancement filtering, the SPORG-RBVS model ensures robust segmentation of fundus images. Particularly, the automated segmentation approach employing SPORG incorporates primary seed point generation and threshold determination using the SPO method, contributing to the overall performance of disease detection. A wide-ranging experimental analysis is executed and the outcomes are examined on three benchmark databases such as Digital Retinal Images for Vessel Extraction (DRIVE), Structured Analysis of the Retina (STARE), and CHASE_DB1 (CHASE). The comparative study stated the supremacy of the SPORG-RBVS method over existing techniques with maximum accuracy of 98.68%, 98.14%%, and 98.34% under DRIVE, STARE, and CHASE datasets, respectively.