Explainable Fundus Image Curation and Lesion Detection in Diabetic Retinopathy
Anca Mihai, Adrian Groza
TL;DR
The paper tackles the challenge of diabetic retinopathy fundus image quality and reliable lesion annotation by proposing an end-to-end data curation framework that filters low-quality images with an explainable classifier, enhances images, provides DL-based lesion suggestions, and measures inter-annotator agreement to ensure annotation reliability. It combines handcrafted feature descriptors with contrastive learning and vision-language model signals to assess quality, and uses lesion-specific DL models with CLAHE/gamma enhancement for improved detection, followed by post-processing and agreement checks. Datasets from multiple public sources are fused to train robust models, and results show improved lesion segmentation and annotation reliability, demonstrating readiness for real-world clinical deployment. The work also offers explainability via SHAP and provides code for reproducibility.
Abstract
Diabetic Retinopathy (DR) affects individuals with long-term diabetes. Without early diagnosis, DR can lead to vision loss. Fundus photography captures the structure of the retina along with abnormalities indicative of the stage of the disease. Artificial Intelligence (AI) can support clinicians in identifying these lesions, reducing manual workload, but models require high-quality annotated datasets. Due to the complexity of retinal structures, errors in image acquisition and lesion interpretation of manual annotators can occur. We proposed a quality-control framework, ensuring only high-standard data is used for evaluation and AI training. First, an explainable feature-based classifier is used to filter inadequate images. The features are extracted both using image processing and contrastive learning. Then, the images are enhanced and put subject to annotation, using deep-learning-based assistance. Lastly, the agreement between annotators calculated using derived formulas determines the usability of the annotations.