Advancing explainable AI in healthcare: Necessity, progress, and future directions

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Precise liver geometric modeling is crucial for treatment planning to minimize harm to surrounding tissues and vessels.
• Machine learning and computer vision techniques offer automated and effective liver image segmentation.
• The ambiguity of AI systems hinders their adoption in critical healthcare domains.

Purpose of the Study:

• To focus on the interpretability of machine learning methods in healthcare.
• To present a literature review and taxonomy of explainable AI (XAI) approaches.
• To serve as a reference for theorists and practitioners in the field.

Main Methods:

• Systematic literature review of explainable AI (XAI) approaches.
• Analysis of XAI methods published between 2019 and 2023.
• Development of a comprehensive taxonomy of XAI method traits and aspects.

Main Results:

• Identified and categorized various explainable AI (XAI) techniques.
• Provided a structured overview of XAI methods for diverse user groups.
• Highlighted the potential of explainable modeling for trustworthy AI in healthcare.

Conclusions:

• Explainable AI (XAI) is essential for the safe and effective integration of AI in healthcare.
• Thorough validation, data quality, cross-validation, and regulation are critical for trustworthy AI.
• The developed taxonomy can guide researchers and practitioners in selecting and applying appropriate XAI methods.