AI spots esophageal cancer in routine noncontrast CT scans: 80,612-patient study hits 98.5% specificity
pattyneta · x · 2026-10-01
A Nature Medicine study shows an AI model can detect esophageal cancer and precancerous lesions from ordinary noncontrast chest CT scans — previously considered impossible.
- Covered 80,612 patients across 12 centers
- External cohorts: 98.5% specificity and 90.0% sensitivity for cancer
- 52.5% sensitivity for precancerous lesions
Open access paper suggests routine CT scans could be repurposed for large-scale, near-zero-cost esophageal cancer screening.
More from Research
- Stanford theorist: LLMs now approach Millennium Prize Problems; math community in turmoil — burny_tech · 2026-10-01
- Evo-Φ36 cryo-EM structure featured in PDB's 55th anniversary Molecule of the Month — BrianHie · 2026-10-01
- Why you should be suspicious of newly released public datasets — yenkel · 2026-10-01
- Boston Dynamics gives Atlas 13-DOF dexterous hands built for sim-to-real RL and real work — chris_j_paxton · 2026-10-01
- Karl Friston Podcast: Beyond the Free Energy Principle, A Life Driven by Curiosity — burny_tech · 2026-10-01
- LSD Online Distillation Cuts RL Post-Training Length Tax from 19% to -3.7% — Xu Wan · 2026-10-01