Could a simple scan image save thousands of lives—without pain, without a camera down the throat, and, most importantly, earlier than the most skilled doctors? That’s the promise of Grape, a new artificial intelligence model developed by Chinese tech giant Alibaba, capable of detecting signs of stomach cancer at an ultra-early stage simply by analyzing 3D medical images.
This significant scientific advancement, recently published in Nature Medicine, may represent a turning point in the fight against one of the most feared silent killers.
A Deadly Cancer, Often Invisible for Too Long
Gastric cancer, or stomach cancer, is the fourth leading cause of cancer death worldwide. Its greatest danger? It often progresses silently: symptoms appear late, sometimes when the disease is already advanced. Even when screening is possible, it can be burdensome and invasive. Early detection is crucial for improving survival rates, making this research particularly impactful.
In China, where cases are particularly numerous, less than 30% of patients agree to undergo an endoscopy. This procedure, which involves inserting a camera into the throat, is unpleasant, expensive, and rarely practiced as a preventative measure.
tens of thousands of cases are diagnosed too late, significantly reducing chances of survival.
Grape: The AI That Reads the Stomach
To address this challenge, Alibaba and the Zhejiang Cancer Hospital developed Grape, short for Gastric Cancer Risk Assessment Procedure. This artificial intelligence system is based on a simple, yet remarkably effective idea: using non-invasive 3D scanners (computed tomography) and having them analyzed by an AI model trained to detect micro-signals invisible to the human eye.
Traditionally, CT scans have been considered inadequate for examining hollow organs like the stomach due to shifting anatomy and the subtlety of early lesions. However, the research team overcame these limitations by building the world’s largest dataset of gastric cancer CT images, enabling the AI to recognize patterns that even experienced radiologists might miss.
Impressive Results
In a large-scale clinical trial involving nearly 100,000 patients across 20 hospitals in China, Grape outperformed human radiologists. It detected cancers with a sensitivity of 85.1% (true positive rate) and a specificity of 96.8% (very few false positives).
In simpler terms: it identified more actual cases, with fewer errors, and sometimes months before doctors, according to the study. This level of accuracy could significantly improve patient outcomes.
In one specific case cited by Alibaba, Grape identified a tumor six months before it was visible to radiologists, allowing the patient to begin life-saving treatment in time.
Simpler, Faster, More Acceptable
One of the most promising aspects of this technology is its non-invasiveness. Where endoscopy is off-putting due to its discomfort, a CT scan is quick, painless, and widely available. This paradigm shift could craft early screening much more common, particularly in rural or under-equipped areas. Increased accessibility to screening is a key factor in improving cancer survival rates.
The Grape model is currently being deployed in the provinces of Zhejiang and Anhui, two pilot regions in China. Eventually, it could be integrated into national screening programs.
Alibaba Looks Further
This isn’t the first breakthrough from Alibaba’s Damo Academy research hub. In 2023, another of their models, Damo Panda, specializing in pancreatic cancer detection, was already recognized by the U.S. Food and Drug Administration (FDA) as a “breakthrough device.” This validation could pave the way for international export of these AI solutions.
Alibaba is also collaborating with the World Health Organization (WHO) to deploy these technologies in emerging countries, where medical resources are limited, but where public health gains could be immense.
Toward Predictive and Accessible Medicine?
This new breakthrough demonstrates that artificial intelligence is not just a laboratory gadget or a productivity tool. It’s poised to become a pillar of the medicine of the future, capable of spotting the invisible, streamlining procedures, and democratizing access to screening for serious diseases.
And with cancers as silent and deadly as stomach cancer, this ability to detect before the first symptoms could make all the difference.
