Objective:

To address the validation gap in AI tools used in pathology and provide a framework for safe implementation.

Approach:

• Recommendation Statement: The Digital Pathology Association (DPA) published a recommendation statement to guide laboratories in validating and deploying AI tools in pathology.
• Focus on Validation: The recommendations emphasize the importance of validating both scanners and AI algorithms separately to ensure accuracy and reliability.
• Practical Steps: The guidance outlines practical steps for laboratories, including early planning, pathologist oversight, and establishing quality control processes.

Key Findings:

• Many hospitals deploy AI tools without confirming validation for the specific scanners used.
• AI models can lose accuracy when analyzing images from different scanners than those used during training.
• Validation of scanners and AI algorithms separately is crucial to identify performance issues effectively.

Interpretation:

The recommendations aim to provide a framework for safe and consistent AI implementation in pathology, addressing both performance variability and patient safety.

Limitations:

• The article does not provide specific examples of AI tools or case studies.
• It does not discuss the potential costs or resource implications of implementing the recommendations.

Conclusion:

The DPA's recommendations serve as a roadmap for laboratories to implement AI in a way that supports high-quality patient care.

Sources:

• Digital Pathology Association