Pancreatic cancer, known as the "King of Cancers", has an average five-year survival rate of less than 10%, making it the malignant tumor with the lowest survival rate in China and even the world. For early-stage pancreatic cancer, surgery can achieve a resection rate of 90% to 100%, and the 5-year survival rate can reach 50%. However, 80% of pancreatic cancers are at an advanced stage once discovered. This is because the location of pancreatic cancer is hidden and has no obvious features in plain CT images. Current clinical guidelines lack effective screening methods.
Alibaba officially issued a statement saying,The latest research in the top international medical journal "Nature Medicine" shows that through "plain CT+AI", humans have for the first time a large-scale screening method for early pancreatic cancer.
Alibaba Damo Academy (Lakeside Laboratory) has teamed up with more than ten top medical institutions in the world to use AI for pancreatic cancer screening of asymptomatic people in physical examination centers, hospitals, etc.With only the simplest plain CT scan, 31 clinically missed lesions were discovered among more than 20,000 real-world consecutive patients, of which 2 patients with early-stage pancreatic cancer have been cured by surgery.
"NatureMedicine" published a special commentary article on this: "Cancer screening based on medical imaging AI is about to enter a golden age."
The research team built a unique deep learning framework that uses AI to amplify and identify subtle lesion features in plain CT images that are difficult to identify with the naked eye, achieving efficient and safe early pancreatic cancer detection and overcoming the problem of high false positives in previous screening methods.
It is understood that this study has constructed the largest pancreatic tumor CT training set to date (including 3208 real patients).Finally, it passed multi-center verification in more than ten hospitals around the world and measured a sensitivity of 92.9% (the accuracy of judging the presence of pancreatic lesions) and a specificity of 99.9% (the accuracy of judging the absence of disease).
Up to now, this technology has been used more than 500,000 times in hospitals, physical examinations and other scenarios, with only one false positive every 1,000 times. In the future, multi-center prospective clinical verification will continue, with a view to rewriting the pessimistic argument that "screening for pancreatic tumors is not recommended".
At present, this work has made phased progress in seven types of high-incidence cancers, including pancreatic cancer, esophageal cancer, lung cancer, breast cancer, liver cancer, gastric cancer, and colorectal cancer. The research results have been published in medical journals such as "NatureMedicine" and "NatureCommunications" and top AI conferences such as CVPR/MICCAI/IPMI.
Attached is the link to the article "NatureMedicine": https://www.nature.com/articles/s41591-023-02640-w