Explainable AI in Clinical Decision Support Systems: A Meta-Analysis of Methods, Applications, and Usability...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems

Background:

• Artificial intelligence (AI) integration in clinical decision support systems (CDSSs) enhances diagnostics, risk stratification, and treatment planning.
• Explainable AI (XAI) is crucial for overcoming barriers to clinical adoption of AI models.
• Current AI applications show promise but require further development for widespread clinical use.

Purpose of the Study:

• To systematically analyze the application and challenges of XAI methods in CDSSs.
• To identify current trends, gaps, and future directions in XAI for clinical decision support.
• To provide recommendations for responsible and clinically relevant AI implementation in healthcare.

Main Methods:

• Systematic meta-analysis of 62 peer-reviewed studies published between 2018 and 2025.
• Examination of XAI methods in diverse clinical domains: radiology, oncology, neurology, and critical care.
• Focus on model-agnostic techniques like Gradient-weighted Class Activation Mapping (Grad-CAM) and attention mechanisms.

Main Results:

• Model-agnostic XAI techniques are prevalent in imaging and sequential data tasks.
• Significant gaps exist in user-friendly evaluation, methodological transparency, and ethical considerations.
• Lack of research on explanation fidelity, clinician trust, and real-world usability was observed.

Conclusions:

• XAI in CDSSs is vital for advancing healthcare AI.
• Addressing gaps in validation, design, and interpretability is essential for responsible AI.
• Future AI solutions must be transparent, ethical, and clinically relevant for successful integration.