Scholarlink

Glaucoma is a disease which can cause permanent blindness. It can be detected using Fundus Image features like Optic Cup (OC), Optic Disc (OD), Fovea, Colour Intensity, Blood Vessels (BV) and Optic Nerve Head (ONH). OC and OD are used to calculate Cup-to-Disc Ratio (CDR), which is a significant feature for Glaucoma DiagnosisQuery. To avoid vision loss, accurate and early diagnosis is necessary. This study presents an automated approach to forecast CDR to diagnose Glaucoma using combination of ResNet-50 features and CDR with XGBoost classifier. ResNet-50, a pre-trained Convolutional Neural Network (CNN), is utilized for feature extraction, capturing high-level spatial and structural information from the images. The extracted features are then fed into the XGBoost classifier, a robust and efficient machine learning model, to predict CDR and classify patients into 3 different classes – Early Stage, Moderate Stage and Late Stage. The proposed model was experimented on SIGF retinal fundus image datasets, achieving high accuracy, sensitivity, and specificity. The result demonstrates the effectiveness of combination of manual feature with ResNet-50 features using XGBoost classifier for glaucoma forecasting.

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

Forecasting CDR to Diagnose Glaucoma Using ResNet-50 Features and XGBoost Classifier from Retinal Fundus Images

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kartik Thakkar,
  • Ravi Gulati

摘要

Glaucoma is a disease which can cause permanent blindness. It can be detected using Fundus Image features like Optic Cup (OC), Optic Disc (OD), Fovea, Colour Intensity, Blood Vessels (BV) and Optic Nerve Head (ONH). OC and OD are used to calculate Cup-to-Disc Ratio (CDR), which is a significant feature for Glaucoma DiagnosisQuery. To avoid vision loss, accurate and early diagnosis is necessary. This study presents an automated approach to forecast CDR to diagnose Glaucoma using combination of ResNet-50 features and CDR with XGBoost classifier. ResNet-50, a pre-trained Convolutional Neural Network (CNN), is utilized for feature extraction, capturing high-level spatial and structural information from the images. The extracted features are then fed into the XGBoost classifier, a robust and efficient machine learning model, to predict CDR and classify patients into 3 different classes – Early Stage, Moderate Stage and Late Stage. The proposed model was experimented on SIGF retinal fundus image datasets, achieving high accuracy, sensitivity, and specificity. The result demonstrates the effectiveness of combination of manual feature with ResNet-50 features using XGBoost classifier for glaucoma forecasting.