<![CDATA[Diabetic retinopathy (DR) is a diabetes-related condition that can cause vision impairment or vision loss. Accurately identifying the level of DR from retinal fundus images is crucial for early detection. However, poor image quality often degrades classification performance. This study proposes an approach that integrates gamma correction-based image enhancement with a Batch Normalization–Visual Geometry Group (BN-VGG) architecture for multiclass DR severity classification. Gamma correction is applied to improve image contrast, while BN-VGG enhances training stability and feature representation. The proposed method categorizes DR into five classifications: normal, mild, moderate, severe, and proliferative. The enhanced images achieved PSNR of 30.85 and SSIM above 0.86, indicating improved visual quality. The model achieved accuracy at 0.97, sensitivity at 0.92, specificity at 0.98, F1-score at 0.92, Cohen's Kappa at 0.90, and G-Mean at 0.97. The innovative aspect of this study is the incorporation of gamma correction with BN-VGG architecture, demonstrating that image enhancement can significantly improve multiclass DR classification performance without increasing model complexity. The study's results indicate the proposed method's effectiveness for accurate & reliable DR severity classification]]>
Copyrights © 2026
