Diabetic Retinopathy (DR) can degenerate into a lot of complications if not properly managed in its initial stages. Although the symptoms of diabetes are not normally obvious at the early stages, the patient only gets to know of it at the later stage, thus, when the condition is in its severe stages. Therefore, an efficient and robust Diabetic Retinopathy screening system is required at the initial stages, to prevent vision loss and other closely related complications. In this paper, we develop a system with the highest efficacy for Diabetic Retinopathy detection in the early and later stages. The system mainly comprises five modules, which include: (i) Pre-processing, (ii) Segmentation of Blood Vessel, (iii) Segmentation of Exudate (iv) Feature Extraction based on texture analysis and (v) Diabetic Detection. The experimental findings show that the suggested method achieves a more accurate diabetic detection model.

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

Convolution Neural Network-Based Automatic Detection of Blood Vessels in Retinal Images for Diabetic Retinopathy

  • Prosper Dongdomokawono Bagyo,
  • Pankaj Prusty,
  • Mihir Narayan Mohanty,
  • Bibhuprasad Mohanty

摘要

Diabetic Retinopathy (DR) can degenerate into a lot of complications if not properly managed in its initial stages. Although the symptoms of diabetes are not normally obvious at the early stages, the patient only gets to know of it at the later stage, thus, when the condition is in its severe stages. Therefore, an efficient and robust Diabetic Retinopathy screening system is required at the initial stages, to prevent vision loss and other closely related complications. In this paper, we develop a system with the highest efficacy for Diabetic Retinopathy detection in the early and later stages. The system mainly comprises five modules, which include: (i) Pre-processing, (ii) Segmentation of Blood Vessel, (iii) Segmentation of Exudate (iv) Feature Extraction based on texture analysis and (v) Diabetic Detection. The experimental findings show that the suggested method achieves a more accurate diabetic detection model.