Scholarlink

<p>Glaucoma is a progressive optic neuropathy that leads to irreversible vision loss if not detected and treated in its early stages. It is primarily associated with increased intraocular pressure (IOP), which causes damage to the optic nerve. Traditional diagnostic methods, such as tonometry and fundoscopic examinations, require frequent clinical visits and specialized expertise making early detection challenging in remote or underserved areas. Recent advancements in deep learning and medical imaging have introduced automated approaches to assist in glaucoma screening. This study leverages fundus images of the eye to develop a simplistic deep learning-based segmentation model for detecting glaucoma indicators, such as the optic cup-to-disc (C/D). The developed model is based on the UNet architecture, which effectively performs pixel-wise segmentation to isolate the optic cup and disc and utilize a customized loss function for effective performance. With this model the accuracy obtained was 99.5% and IoU score was 97.72%.</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

Segmentation of optic cup and disc regions in optical fundus images for the detection of glaucoma using UNet and a modified loss function

  • 期刊论文

  • 发表日期：

    2025年11月3日

查看全文
  • Vidya Rao C A,
  • Shivani H R,
  • Shree Sushma R,
  • R. V. Manjunath

摘要

Glaucoma is a progressive optic neuropathy that leads to irreversible vision loss if not detected and treated in its early stages. It is primarily associated with increased intraocular pressure (IOP), which causes damage to the optic nerve. Traditional diagnostic methods, such as tonometry and fundoscopic examinations, require frequent clinical visits and specialized expertise making early detection challenging in remote or underserved areas. Recent advancements in deep learning and medical imaging have introduced automated approaches to assist in glaucoma screening. This study leverages fundus images of the eye to develop a simplistic deep learning-based segmentation model for detecting glaucoma indicators, such as the optic cup-to-disc (C/D). The developed model is based on the UNet architecture, which effectively performs pixel-wise segmentation to isolate the optic cup and disc and utilize a customized loss function for effective performance. With this model the accuracy obtained was 99.5% and IoU score was 97.72%.