ExplainabilityTransparencyTrustworthinesscs.CV
TACE: Tumor-Aware Counterfactual Explanations
Eleonora Beatrice Rossi, Eleonora Lopez, Danilo Comminiello
September 19, 2024
arXiv.org
1 citations
Key Takeaways
-
AI transparency in medical imaging is critical for clinical adoption: The paper addresses a major barrier to using AI in healthcare—the "black box" nature of deep learning models—by developing explainable AI methods that help doctors understand why the AI made a specific diagnosis decision.
-
Targeted explanations improve reliability: TACE's key innovation is modifying only tumor-specific features while keeping the rest of the organ unchanged, which makes the explanations more faithful to how the AI actually works. This approach led to dramatic improvements in classification accuracy: 10.69% for breast cancer detection and 98.02% for brain tumor detection.
-
Better explanations can increase diagnostic confidence: By generating more reliable counterfactual explanations (showing what would need to change for a different diagnosis), the method helps clinicians understand the AI's reasoning, potentially increasing trust in AI-assisted diagnosis and reducing diagnostic errors.
-
Method-specific design matters for medical AI: Unlike generic explainability techniques, TACE was specifically designed for medical imaging by focusing on tumor regions rather than making arbitrary image changes, demonstrating that responsible AI in healthcare requires domain-specific solutions tailored to clinical needs.
Abstract
The application of deep learning in medical imaging has significantly advanced diagnostic capabilities, enhancing both accuracy and efficiency. Despite these benefits, the lack of transparency in these AI models, often termed "black boxes," raises concerns about their reliability in clinical settings. Explainable AI (XAI) aims to mitigate these concerns by developing methods that make AI decisions understandable and trustworthy. In this study, we propose Tumor Aware Counterfactual Explanations (TACE), a framework designed to generate reliable counterfactual explanations for medical images. Unlike existing methods, TACE focuses on modifying tumor-specific features without altering the overall organ structure, ensuring the faithfulness of the counterfactuals. We achieve this by including an additional step in the generation process which allows to modify only the region of interest (ROI), thus yielding more reliable counterfactuals as the rest of the organ remains unchanged. We evaluate our method on mammography images and brain MRI. We find that our method far exceeds existing state-of-the-art techniques in quality, faithfulness, and generation speed of counterfactuals. Indeed, more faithful explanations lead to a significant improvement in classification success rates, with a 10.69% increase for breast cancer and a 98.02% increase for brain tumors. The code of our work is available at https://github.com/ispamm/TACE.