3D ReX: Causal Explanations in 3D Neuroimaging Classification

2025-02-14

AI Summary

In 3D neuroimaging classification, the lack of interpretability of AI models undermines clinical trust. Method: This paper proposes the first post-hoc explanation framework grounded in actual causality theory, moving beyond conventional attribution methods that merely capture statistical correlations. By modeling the causal responsibility of input voxels for model decisions via causal reasoning—and integrating 3D gradient-constrained optimization with responsibility map generation—the method localizes causally critical brain regions driving classification outcomes. Contribution/Results: The framework is compatible with mainstream segmentation and classification models. Evaluated on stroke detection, it successfully identifies lesion-associated causally sensitive regions, significantly enhancing clinical interpretability and decisional credibility of explanations. It establishes a novel paradigm for causal interpretability research in medical AI.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality

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 learning to rank, online learning, and counterfactual learning for ranking

Abstract

Explainability remains a significant problem for AI models in medical imaging, making it challenging for clinicians to trust AI-driven predictions. We introduce 3D ReX, the first causality-based post-hoc explainability tool for 3D models. 3D ReX uses the theory of actual causality to generate responsibility maps which highlight the regions most crucial to the model's decision. We test 3D ReX on a stroke detection model, providing insight into the spatial distribution of features relevant to stroke.

Problem

• Explainability in medical imaging AI
• Causality-based 3D model explainability
• Spatial feature distribution in stroke detection

Innovation

• Causality-based explainability tool
• Generates responsibility maps
• Focuses on 3D neuroimaging classification