Diabetic Retinopathy Detection in Fundus Images Using Deep-Learned Features

Diabetic Retinopathy (DR) is a condition that is marked by deterioration of retinal blood vessels due to diabetes, serving as a primary factor for vision impairment. Diabetic Macular Edema (DME) is a consequential complication affecting the macula. To streamline the screening process, a deep-learning model has been developed to evaluate diabetic retinopathy across its early to late stages. The existing model concurrently assesses DR and DME, employing an attention network to investigate the underlying correlation between these conditions. The Indian Diabetic Retinopathy Image Datasets (IDRiD) is utilized for this research, widely recognized in the DR-related research field. The proposed methodology begins with data pre-processing, involving resizing and denoising, followed by feature extraction. The obtained features are then input into attention modules for grading. The performance of the model is evaluated through metrics such as loss and accuracy, sheds light on its effectiveness, providing valuable insights for potential enhancements to optimize its efficiency.