Explainability in Generative Medical Diffusion Models: A Faithfulness-Based Analysis on MRI Synthesis

2026-02-10

AI Summary

This study addresses the opacity of decision-making in generative diffusion models for medical MRI synthesis by introducing fidelity evaluation into the interpretability analysis of such models. The work proposes a novel interpretable framework that integrates denoising trajectories with prototype-based networks—including ProtoPNet, EPPNet, and ProtoPool—to elucidate the model’s underlying reasoning mechanisms. Experimental results demonstrate that the proposed approach, particularly when instantiated with EPPNet, achieves a fidelity score of 0.1534, significantly outperforming baseline methods. This advancement provides more reliable and transparent explanations of the generative process, thereby enhancing the safety and trustworthiness of medical AI systems.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This study investigates the explainability of generative diffusion models in the context of medical imaging, focusing on Magnetic resonance imaging (MRI) synthesis. Although diffusion models have shown strong performance in generating realistic medical images, their internal decision making process remains largely opaque. We present a faithfulness-based explainability framework that analyzes how prototype-based explainability methods like ProtoPNet (PPNet), Enhanced ProtoPNet (EPPNet), and ProtoPool can link the relationship between generated and training features. Our study focuses on understanding the reasoning behind image formation through denoising trajectory of diffusion model and subsequently prototype explainability with faithfulness analysis. Experimental analysis shows that EPPNet achieves the highest faithfulness (with score 0.1534), offering more reliable insights, and explainability into the generative process. The results highlight that diffusion models can be made more transparent and trustworthy through faithfulness-based explanations, contributing to safer and more interpretable applications of generative AI in healthcare.

Problem

• explainability
• generative diffusion models
• MRI synthesis
• faithfulness
• medical imaging

Innovation

• faithfulness-based explainability
• generative diffusion models
• MRI synthesis
• prototype-based explanation
• denoising trajectory