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.
The assay is indicated to identify patients with advanced melanoma who have BRAF V600E or BRAF V600K variants and may benefit from FDA-approved targeted therapies.
