Automatic Detection and Classification of Retinal Disease using Improved Firefly Algorithm-based Gated Recurrent Unit

Optical Coherence Tomography (OCT) is employed to detect and classify retinal disease. However, manual detection of images by ophthalmologists leads to subjectivity and error. Therefore, different automation technique is established but it requires enhancement in classification accuracy. To solve this problem, the Improved Firefly Algorithm-based Gated Recurrent Unit (IFA-GRU) is proposed to automatically detect and classify retinal disease utilizing OCT images. FA is improved by establishing better advancement execution. Initially, the OCT2017 is employed as a benchmark dataset for accessing the IFA-GRU technique. The Gaussian-blur-based sharpening and Contrast Limited Adaptive Histogram Equalization (CLAHE) are established for image quality. The modified ResNet-50 is performed to extract the features by removing the final layer of Fully Connected (FC). Then, IFA is utilized to choose the appropriate features. Finally, GRU is established to detect and classify retinal disease automatically using Deep Learning (DL). The IFA-GRU achieves better accuracy of 99.96% compared to Multi-view Deep-Board Learning Network (MDBL-Net), Multi-Scale-Denoising Residual Convolutional Network (MS-DRCN), and Hybrid ConvNet-Transformer Network (HCTNet) respectively.