Radiology Reimagined: Advancing Clinical Practice Through AI Innovation (2025)-"Proving Value with Qualitative and Quantitative Metrics"

"Proving Value with Qualitative and Quantitative Metrics" – Dr. Nina Kottler

Pdf Summary

This presentation by Dr. Nina Kottler focuses on evaluating the real-world clinical value of artificial intelligence (AI) in radiology, emphasizing the need for standardized best practices and measurable metrics. It outlines the typical innovation adoption cycle—from hype to practical productivity—and stresses the importance of clinicians leading best practice development for AI integration.<br /><br />Central to evaluating AI's impact is the Enhanced Detection Rate (EDR), a theoretical metric quantifying how often AI identifies additional clinical findings compared to radiologists alone. However, EDR can overestimate benefits because it assumes radiologists always accept correct AI results and ignore incorrect ones. To address this, Dr. Kottler introduces Enhanced Detection Gain (EDG), the difference between pre-deployment (theoretical) and post-deployment (observed) EDR, reflecting actual realized clinical benefit.<br /><br />Examples from Radiology Partners show significant gaps between theoretical EDR and observed EDR after AI model deployment for conditions like intracranial hemorrhage and pulmonary embolism, indicating that many promised benefits remain unrealized in practice. This underscores the necessity of careful post-deployment monitoring, bias mitigation, cost management, and continuous evaluation across clinical, technical, and business dimensions in the AI lifecycle.<br /><br />The take-home message is that AI can add value but only if integrated with best practices and realism about clinical adoption. Measuring actual clinical impact through metrics like EDG ensures technology benefits are tangible and justifies AI as a standard of care tool. Dr. Kottler invites clinicians to take an active role in shaping AI frameworks and to leverage ongoing education and collaboration to optimize AI utility in healthcare.<br /><br />Contact and additional resources are provided for further engagement with AI evaluation at Radiology Partners and upcoming educational forums such as RSNA.

Keywords

Artificial Intelligence

Radiology

Clinical Value

Enhanced Detection Rate (EDR)

Enhanced Detection Gain (EDG)

Innovation Adoption Cycle

Post-Deployment Monitoring

Bias Mitigation

AI Integration Best Practices

Clinical Impact Metrics