Explainable Fundus Image Curation and Lesion Detection in Diabetic Retinopathy

TL;DR

This paper presents a comprehensive framework for quality control, lesion detection, and explainability in AI-assisted diabetic retinopathy diagnosis using fundus images, emphasizing data quality and annotation reliability.

Contribution

It introduces a novel quality-control pipeline combining explainable features, image enhancement, and annotator agreement to improve AI training and evaluation in DR detection.

Findings

Effective filtering of low-quality images using explainable features

Enhanced annotation accuracy with deep-learning assistance

Reliable annotation usability assessment through agreement metrics

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.