<![CDATA[Eye diseases such as cataracts, diabetic retinopathy, and glaucoma require early detection to prevent permanent vision impairment. Previous studies have applied Convolutional Neural Networks for fundus image classification; however, optimization of preprocessing using Contrast Limited Adaptive Histogram Equalization in VGG16-based transfer learning remains limited, requiring further evaluation of its impact. This study aims to analyze the effect of CLAHE on VGG16 performance for classifying four fundus image classes: cataract, diabetic retinopathy, glaucoma, and normal. The method compared VGG16 without preprocessing and VGG16 with CLAHE using pre-trained weights and fine-tuning on classification layers. Performance was evaluated using accuracy, precision, and recall. Results showed that accuracy increased from 0.8229 to 0.8520, with an improvement of 0.0291. Improvements were also observed in glaucoma recall and normal precision, indicating better recognition of low-contrast disease patterns. These findings demonstrate that CLAHE enhances contrast, highlights features, and supports deep learning model generalization for medical image classification. Unlike previous studies that primarily focused on improving model architectures, this study specifically evaluates the impact of CLAHE on the sensitivity and generalization capability of VGG16 in multiclass fundus image classification]]>
Copyrights © 2026
