Scholarlink

Uneven blood glucose levels are a hallmark of diabetes, a chronic disease. Diabetic retinopathy (DR) can develop in either type 1 or type 2 diabetes. The retina’s blood vessels (BV) are impacted by DR, a retinal condition. DR can be categorized as either non-proliferative diabetic retinopathy (NPDR) or the severe post stage (PDR), which can cause blindness. Early detection and treatment are key to avoiding this. Although fundus images are primarily utilized for detecting purposes, they can occasionally be impacted by issues such as uneven contrast, poor lighting, and artifacts. Low light conditions are the primary cause of this in medical images. This study presents the results of image enhancement using various preprocessing steps which acts to improve the image quality and prediction of DR utilizing the VGG-16 model on APTOS 2019 dataset. To begin, we create a green channel image by extracting the original image’s green component. Preprocessing was done by combining the green channel and CLAHE with Gaussian blur algorithm. After the preprocessing step the VGG-16 model was implemented for the prediction of DR. The proposed model achieved a test accuracy of 93.02% for APTOS dataset.

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

An Integrated Preprocessing and Deep Learning Method for Diabetic Retinopathy Prediction Using VGG-16

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Prithwiraj Debbarma,
  • Hrijuta Datta,
  • Muthusivaramapandian Muthuraj

摘要

Uneven blood glucose levels are a hallmark of diabetes, a chronic disease. Diabetic retinopathy (DR) can develop in either type 1 or type 2 diabetes. The retina’s blood vessels (BV) are impacted by DR, a retinal condition. DR can be categorized as either non-proliferative diabetic retinopathy (NPDR) or the severe post stage (PDR), which can cause blindness. Early detection and treatment are key to avoiding this. Although fundus images are primarily utilized for detecting purposes, they can occasionally be impacted by issues such as uneven contrast, poor lighting, and artifacts. Low light conditions are the primary cause of this in medical images. This study presents the results of image enhancement using various preprocessing steps which acts to improve the image quality and prediction of DR utilizing the VGG-16 model on APTOS 2019 dataset. To begin, we create a green channel image by extracting the original image’s green component. Preprocessing was done by combining the green channel and CLAHE with Gaussian blur algorithm. After the preprocessing step the VGG-16 model was implemented for the prediction of DR. The proposed model achieved a test accuracy of 93.02% for APTOS dataset.