Counterfactual Explainability Framework With CycleGAN And Counterfactual-Classifier Alignnment Score for Retinal Disease Classification

2026-07-23

AI Summary

Current deep learning models for retinal disease classification suffer from limited interpretability and an inability to reliably link predictions to clinically relevant lesion regions, hindering their clinical deployment. To address this, this work proposes CounterFundus, a novel framework that leverages CycleGAN to generate healthy counterfactual images corresponding to pathological fundus images. Lesion localization is achieved through difference maps between original and counterfactual images. The study introduces CCAS—a unified evaluation metric combining Spearman correlation, Intersection over Union (IoU), and pointing accuracy—to quantitatively assess the spatial alignment between counterfactual explanations and classifier saliency maps. Experiments demonstrate that the generated counterfactuals exhibit strong alignment with classification evidence across all CCAS dimensions, and that CCAS-guided counterfactual data augmentation significantly enhances downstream classification performance.

Abstract

Automated detection of vision impairing retina-based ocular conditions from fundus images is important for early screening, timely referral and reducing dependency on specialist-only assessment, for which neural network-based deep learning (DL) models have been widely utilized. However, explainability of the DL frameworks remains a major bottleneck for clinical adoption, particularly when model decisions are not linked to retinal regions that are clinically meaningful. To address this issue, this study presents CounterFundus, a novel CycleGAN-driven counterfactual explainability framework, integrating EfficientNet-B5-based retinal disease detection with visually interpretable disease-to-normal fundus image translation. For each pathological image, the counterfactual yielded by the CycleGAN generator represents an estimated healthy counterpart and the resultant difference map is utilized to localize disease-associated retinal changes. Unlike conventional post-hoc saliency methods, CounterFundus provides counterfactual explanations through visually plausible disease-to-normal retinal translation. Thereafter, to quantify the spatial agreement between counterfactual difference maps and classifier saliency, the Counterfactual-Classifier Alignment Score (CCAS) is introduced, embedding Spearman correlation, binary IoU and pointing accuracy into a single assessment protocol. To this end, EigenCAM-aligned evaluation demonstrates that the generated counterfactual explanations remain spatially consistent with classifier-relevant retinal evidence across all CCAS dimensions. Along with that, ablation studies further confirm that CCAS-filtered counterfactual augmentation improves the downstream classification performance in fundus images, establishing CounterFundus as a clinically-grounded, explainable artificially intelligence (XAI) framework for retinal disease detection.

Problem

• Explainability
• Retinal Disease Classification
• Counterfactual Explanation
• Clinical Adoption
• Deep Learning

Innovation

• Counterfactual Explainability
• CycleGAN
• Retinal Disease Classification
• CCAS
• XAI