Optimized ICA-Based Approach for Accurate Retinal Vessel Segmentation

Retinal vessel segmentation is a critical task in medical diagnostics, yet it remains challenging due to noise, low contrast, and complex backgrounds in retinal images. Traditional methods often fail to accurately delineate vessels, leading to errors that can compromise clinical outcomes. While Independent Component Analysis (ICA) has been explored for vessel enhancement, its effectiveness is hindered by the fixed selection of independent components (ICs), which can result in under- or over-decomposition, reducing segmentation accuracy. This paper proposes a novel ICA-based adaptive binarization method that addresses these limitations. The key innovation lies in the integration of Principal Component Analysis (PCA) to dynamically determine the optimal number of ICs, ensuring precise decomposition and enhanced feature extraction. Adaptive binarization is then applied to accurately separate vessels from the background, while morphological operations refine vessel structures and eliminate noise. Evaluated on the DRIVE, STARE, and HRF datasets, the proposed method demonstrates superior performance compared to traditional ICA-based techniques, achieving higher accuracy and reliability. This advancement offers a robust and efficient tool for automated retinal image analysis, with significant potential to improve medical diagnostics and support clinical decision-making.