Scholarlink

<p>Diabetic retinopathy (DR) poses a substantial risk for vision loss, which drives innovation in its detection. This study analyses DR research’s approaches and geographic distribution, concentrating on datasets such as MESSIDOR-2, IDRiD, DDR, and EyePACS. It also examines the competition between deep learning algorithm applications with DenseNet, SqueezeNet, Vision Transformers, versus traditional ML architectures, focusing on transfer learning’s impact on model efficacy. Results reveal that DenseNet-169 holds the highest accuracy among all models tested for DR detection, and class-wise performance evaluation yielded a high-performing model alongside balanced success across detection classes. This review evaluates hybrid optimized approaches focused on improving automated systems, highlighting opportunities for further development in integrated frameworks, emphasizing system intelligence.</p>

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

Advances in Deep Neural, Transformer Learning, and Kernel-Based Methods for Diabetic Retinopathy Detection: A Comprehensive Review

  • 期刊论文

  • 发表日期：

    2025年10月2日

查看全文
  • Tathagat Banerjee,
  • Davinder Paul Singh,
  • Pawandeep Kour

摘要

Diabetic retinopathy (DR) poses a substantial risk for vision loss, which drives innovation in its detection. This study analyses DR research’s approaches and geographic distribution, concentrating on datasets such as MESSIDOR-2, IDRiD, DDR, and EyePACS. It also examines the competition between deep learning algorithm applications with DenseNet, SqueezeNet, Vision Transformers, versus traditional ML architectures, focusing on transfer learning’s impact on model efficacy. Results reveal that DenseNet-169 holds the highest accuracy among all models tested for DR detection, and class-wise performance evaluation yielded a high-performing model alongside balanced success across detection classes. This review evaluates hybrid optimized approaches focused on improving automated systems, highlighting opportunities for further development in integrated frameworks, emphasizing system intelligence.