Artificial Intelligence in Digital Pathology for Prostate Cancer Detection: FDA Clearance and Real-World...

Area of Science:

• Digital Pathology
• Surgical Pathology
• Artificial Intelligence

Background:

• Prostate cancer diagnosis in surgical pathology can suffer from interpretation variability.
• Artificial intelligence (AI) offers potential to improve diagnostic accuracy and consistency.

Purpose of the Study:

• To review FDA-cleared AI tools for prostate biopsy interpretation (Paige Prostate Detect and Ibex Prostate Detect).
• To examine regulatory indications, performance, and integration of these AI tools.
• To discuss real-world implementation, risks, and benefits for pathologists.

Main Methods:

• Review of regulatory indications and clinical validation studies for Paige Prostate Detect and Ibex Prostate Detect.
• Analysis of diagnostic performance, including sensitivity and specificity trade-offs.
• Consideration of implementation factors like technical inputs, data analysis levels, and user expertise.

Main Results:

• AI tools demonstrate potential for enhanced diagnostic consistency, particularly for general pathologists.
• Clinical validation studies show trade-offs between sensitivity and specificity in AI-assisted assessments.
• Risks include domain shift and potential inequitable performance in under-represented populations.

Conclusions:

• FDA clearance is a critical first step, but not sufficient for safe AI deployment.
• Local validation and ongoing performance monitoring are essential for equitable and effective integration of AI tools.
• AI can be a valuable adjunct in prostate cancer diagnosis, but requires careful consideration of its limitations and impact on workflows.