Scholarlink

<p>Retinal Detachment (RD) is an ophthalmic disease that leads to sight impairment or permanent blindness if not recognized early. Early screening of RD can enhance the rate of successful treatment and visual outcomes. Therefore, RD requires a quick intervention to avert irreversible visual acuity. The manual screening of RD is labor-intensive and time-consuming. Therefore, it is extremely challenging to do biomedical studies on a massive scale. Due to the extensive subject exper- tise, it is particularly difficult for computer-aided diagnosis tools and ophthalmic professionals. With this view, we proposed a deep analysis-based model named representation learning-based convolutional neural network for automated RD detection. In the first step, the retinographic fundus images undergo the pre-processing process to analyze the original images deeply. Afterward, representations or features of the authentic retinal fundus images are extracted to feed into the representation learning-based convolutional neural network model. Finally, the representation learning convolutional neural network is used as a classifier for the classification and prediction of RD. The proposed model is tested on 1627 images collected from online repositories. The proposed model provides a classification accuracy of 98.36%, a sensitivity of 99.04%, a specificity of 96.05%, and an Area under the curve (AUC) of 1, respectively. It notify that our proposed model achieved 3.36% classification accuracy improvement compared to various baseline methodologies for early diagnosis of RD in fundus images.</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

RL-RD-Net: A latent features integration through representation learning for retinal detachment diagnosis with retinographic color fundus images

  • 期刊论文

  • 发表日期：

    2025年8月2日

查看全文
  • Sonal Yadav,
  • R. Murugan,
  • Nishant Sharma,
  • Naveen kumar Roy,
  • Tripti Goel

摘要

Retinal Detachment (RD) is an ophthalmic disease that leads to sight impairment or permanent blindness if not recognized early. Early screening of RD can enhance the rate of successful treatment and visual outcomes. Therefore, RD requires a quick intervention to avert irreversible visual acuity. The manual screening of RD is labor-intensive and time-consuming. Therefore, it is extremely challenging to do biomedical studies on a massive scale. Due to the extensive subject exper- tise, it is particularly difficult for computer-aided diagnosis tools and ophthalmic professionals. With this view, we proposed a deep analysis-based model named representation learning-based convolutional neural network for automated RD detection. In the first step, the retinographic fundus images undergo the pre-processing process to analyze the original images deeply. Afterward, representations or features of the authentic retinal fundus images are extracted to feed into the representation learning-based convolutional neural network model. Finally, the representation learning convolutional neural network is used as a classifier for the classification and prediction of RD. The proposed model is tested on 1627 images collected from online repositories. The proposed model provides a classification accuracy of 98.36%, a sensitivity of 99.04%, a specificity of 96.05%, and an Area under the curve (AUC) of 1, respectively. It notify that our proposed model achieved 3.36% classification accuracy improvement compared to various baseline methodologies for early diagnosis of RD in fundus images.