AI in imaging: the regulatory landscape

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Regulatory Science

Background:

• Artificial intelligence (AI) applications in medical imaging have grown substantially in recent years.
• Despite advancements, a significant portion of AI medical imaging literature exhibits methodological weaknesses.
• Existing AI tools often lack the rigorous validation required for clinical integration.

Purpose of the Study:

• To highlight the increasing number of AI-enabled medical devices and publications.
• To address the identified weaknesses in the current AI medical imaging literature.
• To discuss the impact of evolving regulatory requirements on AI medical device development and validation.

Main Methods:

• Systematic reviews of AI in medical imaging literature.
• Analysis of evolving regulatory guidance for AI-enabled medical devices.
• Discussion of risk mitigation strategies, including bias identification and clinical validation.

Main Results:

• A significant increase in AI medical imaging publications and devices.
• Identification of substantial weaknesses in a proportion of the existing literature.
• Emergence of proactive regulatory frameworks demanding higher standards for AI device development.

Conclusions:

• Stricter regulatory requirements, while potentially lengthening development times, are crucial for ensuring the trustworthiness and clinical meaningfulness of AI medical devices.
• Addressing risks like bias and ensuring validation in realistic clinical settings are paramount.
• Aligning academic research with regulatory frameworks will improve literature quality and facilitate the translation of AI tools into clinical practice.