Developing a Qualification and Verification Strategy for Digital Tissue Image Analysis in Toxicological Pathology

Area of Science:

• Digital pathology
• Computational pathology
• Toxicologic pathology

Background:

• Digital tissue image analysis extracts quantitative data from whole-slide images.
• Quality control is essential for reliable results in digital pathology workflows.
• Toxicologic pathologists must ensure the accuracy of digital image analysis data.

Purpose of the Study:

• To describe common digital tissue image analysis endpoints and error sources.
• To outline quality assurance approaches for digital image analysis in toxicology.
• To adapt FDA regulatory frameworks for AI/ML software to toxicologic studies.

Main Methods:

• Review of common digital tissue image analysis endpoints.
• Identification of potential sources of analysis errors.
• Adaptation of FDA regulatory framework for AI/ML software modifications.

Main Results:

• Digital image analysis requires stringent quality control for accurate toxicologic data.
• Common endpoints and error sources in digital pathology are identified.
• Recommended quality assurance approaches are provided for classical and ML-based analysis.

Conclusions:

• A well-designed quality control system is vital for digital pathology.
• Toxicologic pathologists play a key role in ensuring the quality of image analysis.
• Proposed approaches enhance the reliability of digital image analysis in toxicopathologic studies.