Particle.news

Download on the App Store

Alibaba’s Grape AI Outperforms Radiologists in Gastric Cancer Detection as Deployment Nears

It provides a non-invasive CT scan alternative to endoscopy to boost early detection of gastric cancer.

Image
A doctor explaining the stomach cancer examination to an elderly patient.

Overview

  • A study involving nearly 100,000 participants across 20 hospitals underpinned the Nature Medicine paper that reported Grape’s clinical performance on June 24, 2025.
  • Grape delivered 85.1% sensitivity and 96.8% specificity in identifying gastric tumors from three-dimensional CT scans, exceeding radiologists’ accuracy by 21.8 and 14 percentage points.
  • By bypassing invasive endoscopy, which fewer than 30% of Chinese patients consent to, the AI could improve early detection rates for a cancer that often lacks symptoms.
  • In one instance the model flagged a late-stage tumor six months earlier than human specialists, highlighting its potential to accelerate diagnosis.
  • Following the published results, Alibaba’s Damo Academy and Zhejiang Cancer Hospital are preparing to roll out Grape across Zhejiang and Anhui provinces.