FGSO_FractalNet: Fractional Group Search Optimizer-Enabled FractalNet for Diabetic Macular Edema Detection Using OCT Image.

Abstract

Diabetic macular edema (DME) is a serious diabetes-related complication characterized by swelling in the macula. It arises when high blood sugar levels harm the retinal blood vessels, causing fluid leakage and accumulation. Consequently, early detection of DME is essential for appropriate treatment to prevent significant vision loss. Moreover, early and effective treatment is key to controlling the disease, keeping vision intact, and achieving better overall outcomes. Accordingly, in this paper, an innovative model named Fractional Group Search Optimizer-enabled FractalNet (FGSO_FractalNet) is introduced for DME detection. The detection process for diabetic macular edema begins by acquiring optical coherence tomography (OCT) images from the database. Next, layer segmentation is conducted on these OCT images using the E-Net algorithm. Following segmentation, image augmentation is performed to enhance the dataset. Then, the features are extracted from the augmented OCT images. Finally, DME is detected using the proposed FractalNet, which is trained using fractional group search optimizer (FGSO). The FGSO is developed by integrating fractional calculus (FC) and group search optimizer (GSO). The proposed FGSO_FractalNet method achieved notable results with an accuracy of 91.333%, a sensitivity of 90.174%, and a specificity of 90.560% for dataset 1, showcasing its strong performance. Similarly, the proposed method attained 89.506% accuracy, 88.371% sensitivity, and 88.750% specificity for dataset 2.