Scholarlink

<p>As the world’s population ages, cataracts are becoming one of the leading causes of visual loss. To avoid blindness, cataracts must be diagnosed and treated as soon as possible. Deep learning models aren’t often used to help diagnose cataracts, though, for a variety of reasons, including restricted data labelling, sample size, unequal distribution and field generalization. The deep-learning network used in this study helps diagnose cataract fundus images by overcoming the aforementioned problems and limitations. For fundus image feature extraction, the network also presents the Sequential Convolutional Network with Similarity Network (SCNet-SNet) for prediction. This enhances the generalizability to novel disease samples. Ultimately, this study uses ablation experiments to confirm the function of each model component, particularly the SCNet-SNet contributions to improving the execution. The experiment outcomes demonstrate that the suggested SCNet-SNet model performs exceptionally well in identifying cataract fundus images with an accuracy of 98.8%, an F1 score of 98.7%, and an AUC value of 99.8%.This SCNet-SNet model’s development offers a valuablecataract auxiliary diagnosis tool and a fresh approach to using deep learning in ocular illness identification.</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

Modeling a Sequential and Similarity Network Model for Cataract Prediction Using Learning Approaches

  • 期刊论文

  • 发表日期：

    2025年12月24日

查看全文
  • Arveti Mallikharjuna Rao,
  • Suma Kamalesh Gandhimathi

摘要

As the world’s population ages, cataracts are becoming one of the leading causes of visual loss. To avoid blindness, cataracts must be diagnosed and treated as soon as possible. Deep learning models aren’t often used to help diagnose cataracts, though, for a variety of reasons, including restricted data labelling, sample size, unequal distribution and field generalization. The deep-learning network used in this study helps diagnose cataract fundus images by overcoming the aforementioned problems and limitations. For fundus image feature extraction, the network also presents the Sequential Convolutional Network with Similarity Network (SCNet-SNet) for prediction. This enhances the generalizability to novel disease samples. Ultimately, this study uses ablation experiments to confirm the function of each model component, particularly the SCNet-SNet contributions to improving the execution. The experiment outcomes demonstrate that the suggested SCNet-SNet model performs exceptionally well in identifying cataract fundus images with an accuracy of 98.8%, an F1 score of 98.7%, and an AUC value of 99.8%.This SCNet-SNet model’s development offers a valuablecataract auxiliary diagnosis tool and a fresh approach to using deep learning in ocular illness identification.