Scholarlink

<p>Diabetic retinopathy (DR) is a leading cause of blindness worldwide, and its early detection is vital to prevent vision loss. Manual examination of retinal fundus images is often time-consuming and may lead to inconsistencies, highlighting the need for reliable automated solutions. In this work, we propose an Enhanced Bag of Visual Words framework combined with Grey Wolf Optimization (EBGWO) for DR detection. The method begins by extracting local features using the Scale-Invariant Feature Transform (SIFT), after which Grey Wolf Optimization is applied to construct a discriminative visual dictionary. Each image is then represented as a bag-of-words histogram and classified using a Random Forest model. Experiments were conducted on two benchmark datasets, MESSIDOR and APTOS 2019, achieving accuracies of 93.9% and 85.9%, respectively. Comparative evaluations with other metaheuristic approaches demonstrate that the proposed EBGWO consistently outperforms alternative methods, confirming its effectiveness for automated DR classification.</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

Diabetic retinopathy detection using enhanced bag of visual word framework with grey wolf optimization

  • 期刊论文

  • 发表日期：

    2026年4月4日

查看全文
  • Sachin Bhandari,
  • Sunil Pathak,
  • Sonal Amit Jain,
  • Yerzhan Kerimbekov

摘要

Diabetic retinopathy (DR) is a leading cause of blindness worldwide, and its early detection is vital to prevent vision loss. Manual examination of retinal fundus images is often time-consuming and may lead to inconsistencies, highlighting the need for reliable automated solutions. In this work, we propose an Enhanced Bag of Visual Words framework combined with Grey Wolf Optimization (EBGWO) for DR detection. The method begins by extracting local features using the Scale-Invariant Feature Transform (SIFT), after which Grey Wolf Optimization is applied to construct a discriminative visual dictionary. Each image is then represented as a bag-of-words histogram and classified using a Random Forest model. Experiments were conducted on two benchmark datasets, MESSIDOR and APTOS 2019, achieving accuracies of 93.9% and 85.9%, respectively. Comparative evaluations with other metaheuristic approaches demonstrate that the proposed EBGWO consistently outperforms alternative methods, confirming its effectiveness for automated DR classification.