Retinal Vessel Segmentation by Pixel Optimization using the BAT Algorithm calculations and A Deep Neural Network
DOI: 10.46335/IJIES.2024.9.1.1
Authors : Prof. Shubhangi Chaware , Dr.Mohd. Zuber
Abstract – Non-uniform illumination and low contrast of blood vessels are challenging tasks in retinal segmentation. Non-accurate blood vessel segmentation degrades the efficacy of automatic blood vessel segmentation. Recently, several authors proposed deep learning and optimization-based blood vessel segmentation. This paper proposes optimized pixel-based segmentation using deep neural networks. For the extraction of pixel features, the LoG feature extractor is employed. The lower content of noise affects the segmentation process, which now applies the BAT optimization algorithm. The BAT optimization algorithm reduces the lower content of noise and improves the training process of deep neural networks. The optimized pixel features improve segmentation accuracy and sensitivity. The proposed algorithm is implemented in MATLAB 2014. For the analysis of the proposed algorithm, it employs two reputed datasets, DRIVE and STARE. The suggested approach has been shown to be successful and effective in increasing the sensitivity of blood vessel segmentation, outperforming other innovative methods.
