Diabetic Macular Edema Detection and Classification Using Advanced Convolutional Neural Networks

Diabetic Macular Edema (DME) is a serious complication of Diabetic Retinopathy (DR) and one of the leading causes of vision loss globally. Early detection is crucial for effective management and prevention of blindness. This research investigates the performance of three latest advanced Convolutional Neural Networks (CNNs)-Xception, NaSNetMobile, and ConvNeXt for the automated detection of DME using transfer learning. We utilized the IDRiD dataset of retinal fundus images, which was preprocessed through several techniques, including resizing, normalization, data augmentation, and class balancing, to enhance model performance. The architectures were fine-tuned to maximize their effectiveness on the classification task. The results of this work indicate that NaSNetMobile outperforms Xception and ConvNeXt in terms of accuracy, sensitivity, and overall efficiency, making it a promising tool for DME detection. The findings provide valuable insights into the comparative strengths of these architectures, contributing to the development of more accurate and efficient deep learning models for medical image classification tasks.