An Integrated Deep Learning Framework for Early Prediction of Diabetic Retinopathy

Diabetic Retinopathy (DR) is a serious severe microvascular complication of diabetes that affects the eyes, leading to progressive damage to the retina and potential vision loss. Timely intervention and detection are crucial for preventing irreversible damage. With the advancement of technology, Deep Learning (DL) has emerged as a powerful tool in the field of medical diagnostics, offering a promising solution for the early prediction of diabetic retinopathy. This study compares four convolutional neural network architectures, DenseNet201, ResNet50, VGG19, and MobileNet V2. for predicting diabetic retinopathy prediction: DenseNet201, ResNet50, VGG19, and MobileNet V2. The evaluation is based on both accuracy and training time data. MobileNet V2 outperforms other models with a validation accuracy of \(\:78.22\text{\%}\). ResNet50 has the shortest training time (15.37 seconds). The findings emphasize the trade-off between model accuracy and computational efficiency, stressing MobileNet V2's potential applicability for Diabetic Retinopathy prediction due to its balance of high accuracy and reasonable training time. Making a 5-fold cross-validation with 100 repetitions, MobileNetV2 exhibits a validation accuracy of \(\:77.4\text{\%}\), showcasing its performance under the area under of the receiver operating characteristic curve (AUC) metric. This suggests its competence in effectively classifying data and highlights its robustness across multiple validation scenarios. Moreover, the clustering approach can find our damaged locations in the retina using the IRI method which achieves almost \(\:90\text{\%}\) accuracy. These findings are useful for researchers, and healthcare practitioners and researchers looking to investigate efficient and effective powerful models for predictive analytics to diagnose Diabetic Retinopathy.