KANEx: Translating Kolmogorov-Arnold Networks' Interpretability to Medical Explainability

2026-07-27

AI Summary

This work addresses the limited clinical trust in medical vision models due to their lack of interpretability. While existing vision-language models can generate natural language explanations, they fail to resolve the inherent black-box nature of underlying visual models. To bridge this gap, the authors propose KANEx, a novel framework that, for the first time, integrates the symbolic transparency of Kolmogorov–Arnold Networks (KANs) into medical explanation generation. By leveraging KAN’s interpretable function units as visual rationales, KANEx drives vision-language models to produce more trustworthy explanations. Furthermore, the authors introduce KAN-Map, a gradient-free heatmap method that jointly enhances the fidelity of visual attribution and textual explanations. Experiments on MIMIC-CXR demonstrate that the KAN architecture significantly improves semantic similarity and saliency map fidelity, yielding a 10% gain in both visual grounding and downstream reasoning quality.

Abstract

Computer vision models have become highly effective for medical applications, yet their black-box nature continues to undermine clinician trust. In clinical workflows, chest X-ray classifiers are increasingly paired with Vision-Language Models (VLMs) to generate natural-language explanations. However, these systems add linguistic fluency without addressing the underlying opacity of the visual model. With the emergence of Kolmogorov-Arnold Networks (KANs), whose spline-based components provide inherently interpretable functional units, we investigate whether this architectural transparency can be leveraged to produce more trustworthy textual explanations. We introduce KANEx, the first ever framework that leverages the symbolic transparency of KANs to ground VLM reasoning. This interpretability also made it possible to design KAN-Map, a novel heatmap generation method derived directly from KAN models rather than gradient approximations. We feed these grounded contexts into downstream VLMs for enhanced explainability. Benchmarked on the MIMIC-CXR dataset, we demonstrate that KAN-based architectures with ResNet/ViT baselines demonstrate improved semantic similarity while producing significantly more faithful saliency maps. KAN architectures improve visual localization and downstream reasoning quality by 10%. Our findings suggest that grounding linguistic explanations and visual attributions in mathematically interpretable units is a necessary step toward trustworthy medical AI.

Problem

• medical explainability
• black-box models
• visual interpretability
• trustworthy AI
• Kolmogorov-Arnold Networks

Innovation

• Kolmogorov-Arnold Networks
• Medical Explainability
• Vision-Language Models
• Interpretable AI
• Saliency Maps