Scholarlink

Diabetic retinopathy (DR), a leading cause of vision loss, can progress without warning, making continuous monitoring crucial. Artificial neural networks (ANNs) have demonstrated success in analysing medical images, particularly in the classification of multi-label data. DR tests help diagnose normal cases as well as various stages of DR, including mild, moderate, and severe non-proliferative DR (NPDR) and proliferative DR (PDR). Retinal fundus images can reveal multiple DR lesions simultaneously, enabling early detection, which is critical in preventing the worsening of DR. Automated systems for detecting DR can facilitate timely follow-up treatments, reducing the risk of retinal damage. In this study, we propose an accurate, robust, and effective automated methodology for DR detection. The approach involves two key steps: (1) Reconstruction and enhancement of blood vessels using customized algorithms, and (2) classification of DR severity using an artificial neural network to differentiate between healthy individuals, those with DR, and those with mild to moderate NPDR.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid Neural Network and Machine Learning Approaches for Accurate Diabetic Retinopathy Detection and Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rahul Vadisetty,
  • Anand Polamarasetti

摘要

Diabetic retinopathy (DR), a leading cause of vision loss, can progress without warning, making continuous monitoring crucial. Artificial neural networks (ANNs) have demonstrated success in analysing medical images, particularly in the classification of multi-label data. DR tests help diagnose normal cases as well as various stages of DR, including mild, moderate, and severe non-proliferative DR (NPDR) and proliferative DR (PDR). Retinal fundus images can reveal multiple DR lesions simultaneously, enabling early detection, which is critical in preventing the worsening of DR. Automated systems for detecting DR can facilitate timely follow-up treatments, reducing the risk of retinal damage. In this study, we propose an accurate, robust, and effective automated methodology for DR detection. The approach involves two key steps: (1) Reconstruction and enhancement of blood vessels using customized algorithms, and (2) classification of DR severity using an artificial neural network to differentiate between healthy individuals, those with DR, and those with mild to moderate NPDR.