Scholarlink

<p>Diabetic retinopathy (DR) is the prime reason for blindness in adults, and its early detection is important for avoiding visual blindness. The ophthalmologists generally require rich experience and skills for effectively diagnosing different retinal diseases from fundas images. The article attempts to develop a deep learning (DL) and wild horse optimization (WHO) based diagnostic tool for processing and classifying fundas images into five different levels of DR as No DR, Mild DR, Moderate DR, Severe DR and Proliferative DR. The WHO, a metaheuristic algorithm, performs multi-level segmentation of colour fundas images, and the segmented image is then used for obtaining colour and texture features. The DL network is then trained by WHO for performing classification. The developed diagnostic tool is studied on two fundas image databases, APTOS and EyePACS containing 3662 and 35,126 images respectively, and the results are compared with existing diagnostic tools. The study reveals that the proposed method offers an accuracy, recall, precision and F1 score of (89.05%, 80.95%,97.25%, 80.39%, and 80.21%) for APTOS, and (92.07%, 83.39%, 98.03%, 76.89% and 79.64%) for EyePACS respectively, which are much better than those of the existing diagnostic tools.</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

Wild horse optimization and deep learning based computer aided diagnostic tool for retinal diseases

  • 期刊论文

  • 发表日期：

    2025年7月3日

查看全文
  • Janani Priya Mohan,
  • Yamuna Govindarajan

摘要

Diabetic retinopathy (DR) is the prime reason for blindness in adults, and its early detection is important for avoiding visual blindness. The ophthalmologists generally require rich experience and skills for effectively diagnosing different retinal diseases from fundas images. The article attempts to develop a deep learning (DL) and wild horse optimization (WHO) based diagnostic tool for processing and classifying fundas images into five different levels of DR as No DR, Mild DR, Moderate DR, Severe DR and Proliferative DR. The WHO, a metaheuristic algorithm, performs multi-level segmentation of colour fundas images, and the segmented image is then used for obtaining colour and texture features. The DL network is then trained by WHO for performing classification. The developed diagnostic tool is studied on two fundas image databases, APTOS and EyePACS containing 3662 and 35,126 images respectively, and the results are compared with existing diagnostic tools. The study reveals that the proposed method offers an accuracy, recall, precision and F1 score of (89.05%, 80.95%,97.25%, 80.39%, and 80.21%) for APTOS, and (92.07%, 83.39%, 98.03%, 76.89% and 79.64%) for EyePACS respectively, which are much better than those of the existing diagnostic tools.