Scholarlink

This study presents an integrated framework combining Convolution Neural Networks (CNN) and a large language model (LLM) to improve the classification of retinal diseases: coronal neovascularization (CNV), diabetic macular edema (DME), and Drusen, using optical coherence tomography (OCT) images. Three CNN-based models: Simple, Complex, and Pre-Trained (InceptionV3) were evaluated; whereas InceptionV3 (utilizing transfer learning) was found to have achieved superior performance, delivering over 95% accuracy alongside strong performance metrics. A Flask-based web interface was developed to facilitate seamless communication between CNN and LLM components, providing a user-friendly platform for diagnosis and decision support. This research underscores the potential of AI-driven solutions in improving early detection and treatment of retinal diseases, with scalability to other medical conditions, ultimately fostering better patient outcomes and greater trust in technology-enhanced diagnostics.

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

Automated Retinal OCT Image Classification and Disease Interpretation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sahar Hamayun,
  • Faran Mushtaq,
  • Muhammad Usman Akram,
  • Saad B. Ahmed

摘要

This study presents an integrated framework combining Convolution Neural Networks (CNN) and a large language model (LLM) to improve the classification of retinal diseases: coronal neovascularization (CNV), diabetic macular edema (DME), and Drusen, using optical coherence tomography (OCT) images. Three CNN-based models: Simple, Complex, and Pre-Trained (InceptionV3) were evaluated; whereas InceptionV3 (utilizing transfer learning) was found to have achieved superior performance, delivering over 95% accuracy alongside strong performance metrics. A Flask-based web interface was developed to facilitate seamless communication between CNN and LLM components, providing a user-friendly platform for diagnosis and decision support. This research underscores the potential of AI-driven solutions in improving early detection and treatment of retinal diseases, with scalability to other medical conditions, ultimately fostering better patient outcomes and greater trust in technology-enhanced diagnostics.