Recognition and Detection of Diabetic Retinopathy Using Densenet-65 Based Faster-RCNN

• DOI: 10.32604/cmc.2021.014691
• Publication Date: 2021-02-23T02:30:14Z
• Authors: Saleh Albahli; Tahira Nazir; Aun Irtaza; Ali Javed
• Keywords: Robustness; Feature Engineering

Diabetes is a metabolic disorder that results in retinal complication called diabetic retinopathy (DR) which one of the four main reasons for sightlessness all over globe. DR usually has no clear symptoms before onset, thus making disease identification challenging task. The healthcare industry may face unfavorable consequences if gap identifying not filled with effective automation. Thus, our objective to develop an automatic and cost-effective method classifying samples. In this work, we present custom Faster-RCNN technique recognition classification lesions from images. After pre-processing, generate annotations dataset required model training. Then, introduce DenseNet-65 at feature extraction level compute representative set key points. Finally, localizes classifies input sample into five classes. Rigorous experiments performed on Kaggle comprising 88,704 images show introduced methodology outperforms accuracy 97.2%. We have compared state-of-the-art approaches its robustness term localization classification. Additionally, cross-dataset validation APTOS datasets achieved remarkable both training testing phases.