The Medical AI Podcast Episode 53: Radiology AI – the Roadblocks and how to get past them, with Professor Paul Chang
4 snips
Mar 8, 2025 
In this enlightening discussion, Professor Paul Chang, a radiology expert from the University of Chicago, dives into the hurdles of AI adoption in breast cancer detection. He highlights a significant study where AI outperformed human radiologists, yet only 10-20% of mammograms utilize this technology. Chang explores trust issues, the need for compelling use cases, and the complexities of data integration. He also addresses the regulatory challenges the FDA faces in overseeing these innovations, calling for a fresh perspective and better collaboration in healthcare.
AI Snips
Chapters
Books
Transcript
Episode notes
Radiology Faces Unsustainable Pressure
- Radiology is under intense pressure due to growing data and complexity, causing burnout and quality variability.
- AI can augment radiologists to handle workload and enable precision medicine through better image analysis.
AI Follows Healthcare Hype Cycle
- Healthcare is traditionally a late adopter of technology despite using high-end equipment.
- AI adoption suffers from overhyped expectations followed by disillusionment but will mature like past tech.
Trust Issues in Deep Learning AI
- Deep learning AI's "black box" nature causes lack of trust since we don't fully understand how it works.
- Algorithm drift reduces real-world performance, reflecting implementation flaws, needing continuous model updates.
