Get Started
Home
Topics
Search
Library
Research questionHow can retinal fundus models remain interpretable and accurate when pathology alters vessel appearance?Fundus classifiers may produce accurate predictions without showing whether disease-relevant retinal regions drove them. Pathological changes can also reduce vessel-segmentation performance, making automated outputs harder to trust.
AI
Computer Vision
Health
Image & Video Processing
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.Explainable Convolutional Neural Networks for Retinal Fundus Classification and Cutting-Edge Segmentation Models for Retinal Blood Vessels from Fundus ImagesThe evidence uses FIVES for classification and FIVES and DRIVE for segmentation, with ImageNet-pretrained CNN classifiers, five gradient-based explanation methods, and CNN-, attention-, and transformer-based U-Net variants. Explanation comparisons are qualitative, and the reported results are dataset-specific; no clinical validation or deployment evidence is provided.research paper · Sep 3, 2026
Related questions
How can retinal vascular measurements support interpretable, label-free phenotyping for Alzheimer’s research?How can medical image models segment lesions despite complex backgrounds and varied morphologies?How can histopathology models withstand adversarial perturbations without erasing diagnostic tissue structures?How can computational pathology distinguish disease signal from acquisition shortcuts in reactive brain tissue?