Pancreatic cancer, known as the "king of cancer", is the cancer with the lowest five-year survival rate of all malignant tumors. It is difficult to detect pancreatic cancer with common plain CT. Now, with the help of artificial intelligence (AI), plain CT is expected to play a huge potential in large-scale early screening of pancreatic cancer. Recently, relevant research results have been published as original works in the international medical journal Nature Medicine. Dr. Cao Kai, attending physician of the Radiodiagnosis Department at the First Affiliated Hospital of Naval Medical University (Shanghai Changhai Hospital), is the co first author, and Professor Shao Chengwei, Professor Lu Jianping, and others are the co corresponding authors. As a digestive tract tumor with poor prognosis, pancreatic cancer has clinical characteristics such as difficult early diagnosis, low resection rate, easy recurrence and metastasis after surgery. "The early symptoms of pancreatic cancer are hidden, and there is still a lack of biomarkers or imaging methods that are highly sensitive and specific enough to be widely used in large-scale population screening." Cao Kai introduced, "This leads to a low early diagnosis rate of pancreatic cancer. Once diagnosed, 80% of pancreatic cancer patients are in the middle and late stages." Cancer treatment focuses on early screening and early treatment, and improving the early screening rate of pancreatic cancer is very important to improve the prognosis of pancreatic cancer patients. Early detection and treatment of pancreatic cancer has always been the focus of medical attention. Chest plain CT is simple and easy to perform, and is currently widely used in lung nodule screening. It has become an internationally recognized and suitable early lung cancer screening method with a high detection rate. So, can chest plain CT appropriately expand its application scope to screen pancreatic cancer? Under the leadership of the Shanghai Institute of Pancreatic Diseases, the research team, together with Ali Dharma Institute, the First Affiliated Hospital of Zhejiang University Medical College and other institutions, has constructed a unique in-depth learning framework, which will eventually be trained as a pancreatic cancer detection model (PANDA). Cao Kai introduced that PANDA firstly uses a segmentation network (U-Net) to locate the pancreas, secondly uses a multi task network (CNN) to detect lesions, and thirdly uses a dual channel transformer module to distinguish pancreatic cancer from other pancreatic lesions. Simply put, the "three-step method" is the use of AI to amplify and recognize subtle pathological features in plain CT images that are difficult to recognize by the naked eye. The pancreatic tumor CT training set constructed by this model currently includes 3208 surgical cases, which have been validated by multiple centers in 10 hospitals worldwide, with a sensitivity of 92.9% (accuracy in identifying the presence of pancreatic tumors) and a specificity of 99.9% (accuracy in identifying the absence of tumors). In the retrospective trial of 20530 real cases, the model found 31 clinically misdiagnosed lesions, of which 2 patients with early pancreatic cancer had completed surgery. "The research results fully show that the use of 'plain CT+AI' for large-scale early screening of pancreatic cancer has great potential. This will provide new support for optimizing the diagnosis and treatment guidelines for pancreatic cancer screening, and will also have a positive impact on the future diagnosis and treatment process, treatment decisions, treatment costs, etc. of pancreatic cancer." Shao Chengwei said. "This study clinically confirmed the reliability of the 'plain CT+AI' cancer screening technology path, and provided innovative strategies for the treatment of pancreatic cancer." Cao Kai said that with the continuous maturity and promotion of the technology“