Medical imaging demand is compounding exponentially, while the supply of radiologists remains strictly bottlenecked. How does a critical healthcare discipline overcome a 10x order-of-magnitude mismatch between surging scan volumes and available interpreting capacity?

In this episode of Rethink Imaging, host Chris St. John is joined by Dr. Nina Kottler, a leading authority on clinical AI integration and Associate Chief Medical Officer of Clinical AI at Radiology Partners. Dr. Kottler breaks down the decade-long evolution of healthcare AI, from programmatic machine learning in 2016 designed to enhance diagnostic sensitivity, to the modern agentic foundation models required to solve today's crushing operational backlogs. She details the heavy cognitive load radiologists face when managing fragmented software systems across multiple monitors, defines what constitutes a true "agentic" AI system, and clarifies the distinction between model drift and shifting input data. Finally, Dr. Kottler presents a strategic case for reform: urging health systems and payers to shift focus from reimbursing standalone software applications to funding robust clinical AI governance.

• The 10x Capacity Mismatch: Why imaging volume growth is outstripping radiologist workforce capacity by an order of magnitude, making unread scans the single greatest quality risk.

 • The Evolution from Narrow to Agentic AI: How medical AI is advancing from binary single-finding detection tools to multimodal foundation models capable of autonomous workflow prioritization. 

• The Basal Ganglia Workflow Friction: Why modifying software interfaces creates immense cognitive load for clinicians and how to execute safe, human-centered change management

• Demystifying AI Drift: How performance changes stem from shifts in clinical input data rather than degradation of the core model itself.

 • Reimbursing Clinical AI Governance: A policy proposal advocating that CMS and private payers fund clinical oversight and safety governance rather than isolated software algorithms. 

01:18 - Classifying AI: Tracing the evolution from early machine learning to multimodal foundation models. 

04:39 - The 2016 vs. 2026 Paradigm Shift: Quality vs. Capacity Moving from lesion detection quality to capacity crisis management.

11:37 - The 10x Capacity-Volume Mismatch: Analyzing why health systems cannot staff their way out of surging scan volumes.

12:57 - The "Matrix" Workflow: Navigating disconnected EMRs, RIS, and PACS across multiple monitor setups.

16:35 - Human + AI Symbiosis & Validating AI Outputs: Validating narrow AI outputs, quantitative metrics, and human oversight.

22:19 - Defining True Agentic AI: Breaking down high-level goal setting, tool integration, and autonomous self-monitoring.

26:18 - AI Drift vs. Shifting Data & Generalizability: Training large foundation models on broad unlabeled clinical datasets.

32:58 - Patient Trust, Privacy & Precision Health: Evolving from historical diagnosis to predictive patient care.

37:06 - Managing Change Fatigue & Cognitive Strain: Balancing basal ganglia habits with frontal lobe cognitive demands.

43:48 - Rethinking AI ROI: Reimbursing Clinical Governance: Why payers should reimburse clinical governance over individual software