DAMO GRAPE - An early gastric cancer identification model jointly developed by Alibaba DAMO Academy and Zhejiang Cancer Hospital
DAMO GRAPE is the world's first AI model for identifying early-stage gastric cancer based on plain CT scans, jointly launched by Zhejiang Cancer Hospital and Alibaba DAMO Academy. DAMO GRAPE breaks through the limitations of traditional imaging, using deep learning to analyze non-contrast CT scans...
What is DAMO GRAPE?
DAMO GRAPE, jointly developed by Zhejiang Cancer Hospital and Alibaba DAMO Academy, is the world's first AI model for identifying early gastric cancer based on plain CT scans. Breaking through the limitations of traditional imaging, DAMO GRAPE utilizes deep learning to analyze non-contrast CT images, achieving highly efficient screening for gastric cancer. In large-scale clinical studies, DAMO GRAPE demonstrated a sensitivity of 85.1% and a specificity of 96.8%, significantly outperforming human radiologists. The model can detect early gastric cancer lesions up to six months in advance, providing a new and efficient method for the early diagnosis and treatment of gastric cancer, and is expected to significantly improve the survival rate of gastric cancer patients.
DAMO GRAPE's main functions
- Early gastric cancer screeningDAMO GRAPE, based on the analysis of non-contrast CT images, identifies early gastric cancer lesions, significantly improving the early detection rate of gastric cancer.
- Auxiliary diagnosisIt provides diagnostic support to radiologists, helping to improve the accuracy and efficiency of diagnosis and reduce the possibility of missed diagnoses and misdiagnoses.
- risk assessmentTo assess the risk of gastric cancer in patients and identify high-risk individuals, which facilitates further diagnostic procedures such as gastroscopy.
- Early warningEarly detection of potential gastric cancer lesions before patients show obvious symptoms can buy valuable time for early treatment.
DAMO GRAPE's technical principles
- Deep learning algorithmsDAMO GRAPE is based on deep learning technology and is trained using a large amount of CT image data of gastric cancer and non-gastric cancer to learn the characteristics and patterns of gastric cancer lesions.
- Multicenter datasetBased on the world's largest multicenter dataset of gastric cancer plain CT images (6720 cases), covering data from different regions and different equipment, the model's generalization ability is improved.
- Image segmentation and classificationThe model combines segmentation and classification networks. First, it segments the gastric region from CT images, then performs tumor detection and classification on the segmented region, and outputs a gastric cancer risk score and segmentation mask.
- Feature extraction and recognitionAnalyzing subtle changes and patterns in CT images, such as gastric wall thickness and gastric mucosal heterogeneity, can help identify early gastric cancer lesions, breaking through the limitations of traditional imaging techniques.
- Grad-CAM visualizationBased on Grad-CAM technology, the decision-making process of the model is visualized, which helps doctors understand the basis of the model's judgment and enhances the interpretability of the model.
DAMO GRAPE's project address
- Technical Papers: https://www.nature.com/articles/s41591-025-03785-6
Application scenarios of DAMO GRAPE
- Large-scale population screeningIn health check centers and primary hospitals, gastric cancer screening is conducted on a large number of people to identify potential patients in advance and improve the early detection rate of gastric cancer.
- Assisting doctors in diagnosisIt provides radiologists with auxiliary diagnostic tools to help identify gastric cancer lesions more accurately, reduce missed diagnoses and misdiagnoses, and improve diagnostic efficiency.
- High-risk population monitoringRegular screenings are conducted for residents in areas with a high incidence of gastric cancer, as well as high-risk groups such as those with a family history of gastric cancer or chronic gastritis, to detect lesions in advance.
- Early warning and interventionEarly detection of potential gastric cancer lesions when patients have no obvious symptoms allows for timely treatment, improving patient survival rates and quality of life.
- Medical resource optimizationIn a tiered healthcare system, medical resources are allocated rationally, medical efficiency is improved, and data and tools are provided to support medical research and teaching.