Alibaba has unveiled a new AI model, “Grape,” to help detect stomach cancer early through CT scans. Developed in partnership with Zhejiang Cancer Hospital, the tool promises more accurate diagnoses than human radiologists. With gastric cancer ranking among China’s deadliest cancers, early detection is often the difference between life and death.
Grape achieved a sensitivity of 85.1% and a specificity of 96.8% in clinical trials, outperforming traditional diagnostic approaches. The system is now being rolled out in hospitals across Zhejiang and Anhui provinces. For many, this marks a shift in how technology could help reduce the burden of invasive cancer screenings and improve survival rates.
Grape Uses AI to Make Stomach Cancer Detection Faster and Easier
Grape AI was built by Alibaba’s research arm, Damo Academy, using deep learning to analyze 3D CT scan data. The goal is to spot signs of early-stage gastric cancer that human eyes may miss. Working with Zhejiang Cancer Hospital, the team trained Grape on large sets of anonymized scan images. The model doesn’t replace doctors but assists them in identifying tumors earlier and more consistently.
“Tools like Grape can shorten diagnosis time and ease pressure on our radiology teams,” said Dr. Cheng Xiangdong of Zhejiang Cancer Hospital. Unlike endoscopies, which are invasive and often avoided by patients, CT scans reviewed by AI offer a non-invasive alternative. Grape AI follows Alibaba’s earlier work with a pancreatic cancer AI, which was granted a “breakthrough device” label by the US FDA. Its successful trials reflect a broader push to use AI in diagnostics across multiple cancer types.
Strong Results, But Patient and Doctor Trust Will Decide Impact
Despite Grape’s accuracy, its real-world success depends on whether people and hospitals are ready to trust and use it. Studies show that less than 30% of Chinese patients follow through with doctor-recommended endoscopies. That’s one of the barriers Grape aims to lower by offering a simpler, less invasive option. But even the best AI needs patient acceptance to be effective. Concerns over data privacy and the lack of human oversight still make many people uneasy about AI in healthcare.
Healthcare workers, too, may hesitate. Research shows that doctors worry about losing control or struggling to fit new tools into their workflows. To overcome this, Alibaba will need to provide training and involve frontline clinicians in deployment. Whether radiologists were part of Grape’s design process remains unclear. For hospitals in Zhejiang and Anhui to adopt it fully, the tool must not only work, but it must work for people.
Alibaba Advances AI-Led Screening with Grape Cancer Detection Tool
Grape is part of a larger movement to make cancer screening less stressful and more accessible using AI. Its success in detecting gastric tumors reflects a growing confidence in digital diagnostic tools. Earlier projects from Alibaba, like “Damo Panda” for pancreatic cancer, show a pattern: AI tools starting in one cancer type, then expanding.
Grape’s developers plan to extend its capabilities to detect other gastrointestinal cancers next. The AI’s rollout could also benefit from rising global acceptance. The FDA’s earlier fast-tracking of Alibaba’s pancreatic tool suggests similar approval paths may follow for Grape. Yet challenges remain. Privacy safeguards, human review, and clear patient communication will all play a role in its future success.
About The Author
