Johns Hopkins Uses AI to Build $1,000-Per-Specimen Cancer Maps for Faster Discovery
Updated
Updated · The Cancer Letter · Aug 7
Johns Hopkins Uses AI to Build $1,000-Per-Specimen Cancer Maps for Faster Discovery
3 articles · Updated · The Cancer Letter · Aug 7
Summary
Johns Hopkins researchers say AI now turns labor-intensive pathology work that once required hand-drawn tumor annotations into minutes-long processing, producing “AI-ready” cancer data sets for research.
AstroPath adapts infrastructure from the Sloan Digital Sky Survey to map tumor microenvironments at microscopic scale, showing how tumor and immune cells are arranged and interact.
Those spatial maps are being linked to genomic and molecular data to find biomarkers that predict immunotherapy response, an urgent need because only about 30% of solid-tumor patients now benefit.
Each stained and analyzed specimen costs about $1,000, but the team argues that is modest against cancer drugs that can cost $150,000 if better matching cuts side effects and wasted treatment.
Using public melanoma slides from The Cancer Genome Atlas and its AstroID privacy system, the group says it could scale proof-of-concept work to large production within a year if funding arrives.