Can a Better Stain Improve AI in Pathology? | with Ingrid Carlbom, CEO and Founder of CADESS.AI AB

Can a Better Stain Improve AI in Pathology? | with Ingrid Carlbom, CEO and Founder of CADESS.AI AB

In this episode, I talk with Professor Ingrid Carlbom, founder of CADESS.AI, about one of the most persistent challenges in prostate pathology: grading variability.

We explore why classical H&E staining makes segmentation difficult, how a Picrosirius red + hematoxylin stain improves tissue contrast, and why grading at the gland and cellular level leads to more reproducible results. Ingrid walks through clinical testing, explainable user interfaces, and the realities of developing both stain and AI software as a single system.

This episode offers a fresh perspective on digital pathology—one that prioritizes data quality, transparency, and clinical trust.

KEY HIGHLIGHTS:

  • 30–40% interobserver variability in Gleason grading
  • Optimized stains enable precise gland segmentation
  • AI as decision support—not replacement
  • Explainability builds pathologists’ trust
  • Regulatory complexity of combined stain + AI systems
  • Broader applications beyond prostate cancer

Episode Resources:

SUPPORT THE SHOW

Content Episodes You will enjoy