A 5-step best practice to clinically evaluate AI models
Date & Time
Thursday, May 30, 2024, 2:15 PM - 2:30 PM
Nina Kottler

In the rapidly evolving field of medical technology, the integration of Artificial Intelligence (AI) in clinical settings presents both groundbreaking opportunities and significant challenges. One of the challenges is the absence of standardized practices for evaluating AI's role in healthcare. This discussion will outline a 5-step process \ for appraising the clinical accuracy of AI models in radiology. The dialogue will review established accuracy metrics, highlight those of paramount importance from the clinician's vantage point, and introduce novel metrics designed to refine the assessment and integration of AI in clinical practice. This approach underscores the criticality of incorporating the clinician's perspective into the assessment of AI models.