医学
深度学习
人工智能
验光服务
疾病
眼病
计算机视觉
计算机科学
特征(语言学)
眼动
眼科
作者
Vishant P,Suki Bharath M,Sandhya Sasidharan,K. P. Revathi,Senthil Pandi S,Subbalakshmi S
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-04-24
卷期号:: 1-6
标识
DOI:10.1109/icsft67370.2026.11566680
摘要
This paper presents SmartEyeX, a unified multi-modal deep learning framework for automated retinal disease diagnosis using fundus and Optical Coherence Tomography (OCT) images. The proposed system integrates multi-class classification, disease-specific segmentation, quantitative clinical metric extraction, and explainable artificial intelligence within a single deployable architecture. Convolutional Neural Networks and Vision Transformers are employed for classification, while encoder-decoder segmentation networks provide pixel-level localization of pathological regions. Clinically relevant measurements, including Cup-to-Disc Ratio (CDR) and lesion area percentage, are computed to support severity grading. Grad-CAM–based visual explanations enhance model interpretability and clinical trust. Experimental evaluation on 8,000 retinal images demonstrates an overall classification accuracy of 94.44% and mean Dice coefficient of 0.89. The system is deployed using a FastAPI-based inference framework for real-time clinical application. The results confirm that integrating classification, segmentation, and explainability improves diagnostic reliability compared to standalone models.
科研通智能强力驱动
Strongly Powered by AbleSci AI