Article
Lesion-Guided Hybrid Deep Learning Framework for Accurate Diabetic Retinopathy Detection
Diabetic retinopathy is one of the leading causes of vision impairment among people with diabetes, making early diagnosis essential for preventing permanent blindness. Manual examination of retinal fundus images is time-consuming and depends heavily on the expertise of ophthalmologists. This study presents a lesion-guided hybrid deep learning framework for the automatic detection of diabetic retinopathy from retinal images. The proposed approach combines the feature extraction capability of Convolutional Neural Networks (CNN) with the discriminative strength of a hybrid deep learning architecture to identify retinal lesions such as microaneurysms, hemorrhages, and exudates. Before training, the retinal images are preprocessed and enhanced to improve image quality and reduce noise. Data augmentation is applied to address class imbalance and improve the model's generalization ability. Experimental results demonstrate that the proposed framework achieves higher classification accuracy, precision, recall, and F1-score compared with conventional deep learning models. The developed system offers a reliable and efficient computer-aided solution for early diabetic retinopathy screening, supporting ophthalmologists in making faster and more accurate clinical decisions.
Full Text Attachment
