Training AI on Rare Cells: Scopio CTO Erez Naaman on the Diagnostics Digest Podcast

Building an AI that reliably identifies common blood cells is one challenge. Building one that catches the rare malignant cell that appears once in every 10,000 – the cell that changes a diagnosis – is an entirely different problem.

In this episode of the Diagnostics Digest podcast, Scopio CTO and co-founder Erez Naaman discusses what it takes to train clinical-grade AI for hematology: years of annotating massive datasets with clinical partners, the infrastructure required to process multi-gigabyte full-field scans, and why the “long tail” of rare and abnormal cells is where the real diagnostic value lies.

Erez also explains why Scopio’s approach to AI is built around decision support rather than replacement – shifting the lab professional’s workflow from manually searching for cells to verifying what the system has found – and why that distinction matters in a field facing a global staffing shortage.

The conversation covers the origin of Scopio’s computational photography approach, how Full-Field imaging creates the data foundation that makes meaningful AI possible, and where the technology is heading next.

Hosted by Nathan Sharpe on the Diagnostics Digest podcast.

Listen on Diagnostics Digest