On the Interpretability of Artificial Intelligence in Radiology: Challenges and Opportunities

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Artificial intelligence (AI) is increasingly used in clinical radiology.
• Ensuring AI systems function correctly and gain expert trust is essential.
• Interpretability methods aim to enhance understanding of complex AI algorithms.

Purpose of the Study:

• To provide insights into the current state of the art of interpretability methods for radiology AI.
• To discuss radiologists' opinions on AI interpretability.
• To suggest trends and challenges for streamlining interpretability in clinical practice.

Main Methods:

• Review of current interpretability methods for AI in radiology.
• Discussion of radiologists' perspectives on AI interpretability.
• Identification of future trends and challenges.

Main Results:

• Interpretability methods are gaining attention to foster trust in AI for radiology.
• Radiologists' opinions on interpretability are crucial for adoption.
• Several trends and challenges exist in implementing these methods.

Conclusions:

• Effective interpretability methods are vital for the successful integration of AI in radiology.
• Addressing radiologist concerns and practical challenges is necessary for widespread adoption.
• Future research should focus on streamlining interpretability for clinical practice.