Towards Transparent Healthcare: Advancing Local Explanation Methods in Explainable Artificial Intelligence

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Machine Learning

Background:

• AI is increasingly used in healthcare for diagnosis, prognosis, and personalized medicine.
• Transparency and interpretability are crucial for trust and adoption of AI in clinical settings.
• Local Explainable AI (XAI) methods offer case-specific explanations for AI model predictions.

Purpose of the Study:

• To explore the application and impact of local XAI methods in healthcare.
• To highlight the importance of interpretability in AI-driven medical decision-making.
• To review key contributions of local XAI in medical contexts.

Main Methods:

• Focus on Local Rule-Based Explanations (LORE) as a specific local XAI technique.
• Review of existing literature and case studies on local XAI in healthcare.
• Analysis of the trade-offs between model performance and interpretability.

Main Results:

• Local XAI methods, such as LORE, provide granular, understandable insights into AI predictions.
• These methods enhance physician and patient comprehension of AI-driven medical outcomes.
• Local XAI contributes to improved clinical decision-making, fairness, and regulatory compliance.

Conclusions:

• Local XAI is vital for transparent and trustworthy AI applications in healthcare.
• Implementing local XAI can significantly improve the utility and acceptance of AI tools in medicine.
• Further research into local XAI is essential for advancing AI in clinical practice.