Harnessing artificial intelligence for enhanced veterinary diagnostics: A look to quality assurance, Part II External...

Area of Science:

• Veterinary medicine
• Artificial intelligence
• Diagnostic technology

Background:

• Artificial intelligence systems (AIS) are increasingly valuable diagnostic tools in veterinary medicine.
• While AIS demonstrate high accuracy, quality assurance for these systems is still developing.
• Part I of this study focused on AIS development and technical validation.

Purpose of the Study:

• To explore external validation (in silico testing) as the next critical step for AIS development.
• To emphasize the importance of in silico testing for maintaining high-quality veterinary diagnostics.
• To outline a comprehensive quality assurance process for evaluating AIS in veterinary medicine.

Main Methods:

• Investigating sources of bias in AIS.
• Applying calibration methods and predicting uncertainty.
• Implementing safety monitoring systems and assessing repeatability and robustness.
• Conducting in silico testing with unseen data to ensure accuracy and precision.

Main Results:

• External validation through in silico testing is essential for AIS intended for medical use.
• Rigorous quality assurance processes are necessary for reliable AIS.
• Testing with unseen data confirms the accuracy and precision of AIS outputs.

Conclusions:

• In silico testing is a critical quality assurance component for veterinary AIS.
• A multi-faceted quality assurance process ensures the reliability and safety of diagnostic AIS.
• Continued validation is key to integrating AI effectively into veterinary diagnostics.