Reliable XAI Explanations in Sudden Cardiac Death Prediction for Chagas Cardiomyopathy

Logic-based explanations provide guaranteed accuracy: Unlike common AI explanation methods that use heuristics without verification, this study applied a logic-based approach that achieved 100% explanation fidelity, meaning the explanations for why the AI made its predictions were mathematically guaranteed to be correct—a critical requirement for medical decision-making.

Transparency can coexist with high performance: The AI model maintained over 95% accuracy and recall while providing reliable, understandable explanations for its predictions. This demonstrates that making AI systems more transparent doesn't require sacrificing their ability to correctly identify patients at risk of sudden cardiac death.

Current explanation methods may be unreliable in healthcare: The paper found that state-of-the-art heuristic explanation methods showed inferior consistency and robustness compared to the logic-based approach, suggesting that many AI tools currently used or proposed for medical applications may provide explanations that cannot be trusted without verification.

Trustworthy AI could expand life-saving care to underserved regions: By addressing the "black box" problem that hinders AI adoption in clinical practice, this approach could facilitate deployment of cardiac risk prediction tools in endemic regions where Chagas disease is prevalent but medical resources are limited, potentially saving lives where they're needed most.