Survey of Explainable AI Techniques in Healthcare

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Explainable Artificial Intelligence (XAI)

Background:

• Deep learning models are increasingly used in medical imaging and healthcare.
• AI decisions in medicine require transparency and interpretability, similar to human clinical judgment.
• The 'black-box' nature of deep learning hinders trust and adoption in critical medical applications.

Purpose of the Study:

• To survey recent Explainable AI (XAI) techniques applied to healthcare and medical imaging.
• To categorize XAI types and highlight algorithms enhancing interpretability in medical image analysis.
• To identify challenges and provide guidelines for developing better AI interpretations in clinical settings.

Main Methods:

• Comprehensive literature review of recent XAI techniques in medical imaging and healthcare.
• Categorization of XAI methods based on their approach to model interpretability.
• Analysis of algorithms used for increasing transparency in deep learning models for medical applications.

Main Results:

• Summary and categorization of current XAI techniques relevant to medical imaging.
• Identification of key algorithms that improve the interpretability of AI models in healthcare.
• Discussion of prevalent challenges and proposed solutions for XAI in medical contexts.

Conclusions:

• XAI is essential for the safe and effective integration of AI in medical decision-making.
• Guidelines are provided to improve the development and application of interpretable AI in medical imaging.
• Future research directions are outlined to advance XAI in clinical practice and medical image analysis.