Dual-Model Approach for Diabetic Retinopathy and Macular Edema Detection

Diabetic Retinopathy (DR) and Diabetic Macular Edema (DME) are major causes of vision loss in diabetic patients. Early detection is crucial for preventing permanent vision impairment. This research paper proposes a novel dual-model architecture for detecting Diabetic Retinopathy and Diabetic Macular Edema. The proposed framework integrates preprocessing techniques such as circular cropping and Ben’s preprocessing, followed by data augmentation. The Diabetic Retinopathy model is trained on the APTOS2019 dataset, while the Diabetic Macular Edema model is trained using the IDRid dataset. InceptionV3 outperforms other models with an accuracy of 88% for Diabetic Retinopathy classification, while MobileNetV2 outperforms other models with an accuracy of 83% for DME classification. The dual-model architecture enables simultaneous evaluation of retinal images for DR and DME diseases. The Experimental results demonstrate promising outcomes, suggesting the potential effectiveness of the proposed dual-model architecture in diagnosing diabetic eye diseases.