Comparative Analysis of Superpixel and Gabor Methods for Exudate Feature Extraction in Diabetic Retinopathy Fundus Images

Diabetic retinopathy (DR) is a prevalent health issue associated with long-term diabetes, often leading to various complications. One characteristic manifestation of DR is the development of exudates in the retinal region. Fundus image analysis serves as a common method for diagnosing DR, relying on the expertise of ophthalmologists. Recently, computer-assisted diagnostic tools have enhanced the speed of diagnosis, provided diverse perspectives, and evaluated treatment outcomes. Consequently, numerous studies have focused on identifying diabetic retinopathy lesions in fundus images. This research aims to extract exudate features using the superpixel and Gabor methods in fundus retinal images. A combination of image processing with the superpixel algorithm is compared with the Gabor feature extraction method. The findings reveal that the superpixel method outperforms the Gabor method in accurately extracting exudate characteristics. The achieved results indicate a 90% accuracy for the superpixel method and 78.3% accuracy for the Gabor method. These outcomes underscore the superior performance of the superpixel method in exudate extraction from fundus images. This study contributes to advancing the field of diabetic retinopathy analysis and highlights the potential of the superpixel method in improving diagnostic accuracy and efficiency.