Shulou(Shulou.com)11/24 Report--

This may be the first time that mankind has taken the initiative in the struggle against the "King of Cancer".

On November 21, the latest research in Nature Medicine, a top international medical journal, shows that through plain scan CT+AI, humans have for the first time a large-scale screening method for early pancreatic cancer. Ali Damo Hospital (Lakeside Laboratory), in conjunction with more than ten top medical institutions around the world, uses AI to screen pancreatic cancer for asymptomatic people in physical examination centers, hospitals and other asymptomatic people. with the simplest plain scan of CT, 31 clinical missed lesions have been found in more than 20,000 consecutive patients in the real world, of which 2 patients with early pancreatic cancer have been cured by surgery. Nature Medicine published a special commentary on this: "Cancer screening based on medical imaging AI is about to enter a golden era."

Pancreatic cancer, known as the "king of cancer", has an average five-year survival rate of less than 10%, making it the lowest survival rate in China and even in the world. Once 80% of pancreatic cancer is found to be advanced, it is ferocious and difficult to cure. Current clinical guidelines lack effective screening methods because they are prone to missed diagnosis or misdiagnosis. The contrast of plain CT images commonly used in physical examination and hospital is low, so it is difficult to identify early pancreatic lesions.

In the latest paper "Large-scale pancreatic cancer detection via non-contrast CT and deep learning", the medical AI technology based on Aridamo Hospital Shanghai Institute of Pancreatic Diseases, the first affiliated Hospital of Zhejiang University Medical College, Shengjing Hospital affiliated to China Medical University, Fudan University affiliated Cancer Hospital, the first affiliated Hospital of Charlie University in Prague, Xinhua Hospital affiliated to Shanghai Jiaotong University Medical College, Johns Hopkins University and other institutions proposed "plain scan CT+AI" for large-scale early screening of pancreatic cancer for the first time.

In view of the fact that the location of pancreatic canceration is hidden and there is no obvious representation in plain scan CT images, the research team constructed a unique deep learning framework, and finally trained the pancreatic cancer early detection model PANDA: one is to build a segmentation network (U-Net) to locate the pancreas, the second is to use multi-task network (CNN) to detect abnormalities, and the third is to use dual-channel Transformer to classify and identify the types of pancreatic lesions. In short, this technique uses AI to magnify and identify those subtle lesions in plain scan CT images that are difficult to identify with the naked eye, to achieve efficient and safe detection of early pancreatic cancer, and to overcome the problem of high false positive screening methods in the past.

According to Dr. Cao Kai of Shanghai Institute of Pancreatic Diseases, this study constructed the largest pancreatic tumor CT training set so far (including 3208 real patients), and finally passed multicenter verification in more than 10 hospitals around the world. 92.9% sensitivity (accuracy in judging the presence of pancreatic lesions) and 99.9% specificity (accuracy in judging disease-free) were measured. In the retrospective trial of more than 20,000 real cases, 31 cases of clinically missed lesions were found, and 2 patients with early pancreatic cancer were cured by operation.

Up to now, this technique has been used more than 500000 times in hospitals, physical examinations and other scenarios, with only one false positive for every 1000 times. Multicenter prospective clinical trials will continue in the future in order to rewrite the pessimistic argument of "pancreatic tumor screening is not recommended."

Professor Gu Yajia, director of the Department of Radiology diagnosis at the affiliated Cancer Hospital of Fudan University, said that this paper proposes a potential method for large-scale screening of pancreatic cancer, which will improve the detection rate without additional radiation and financial burden on patients. "imagine that when we do the simplest plain CT scan when we go for a physical examination, we will be able to detect pancreatic cancer, which will help many pancreatic patients and reduce the occurrence of tragedies."

Lu Le, head of the Dharma Hospital Medical AI team, said the study was an important milestone in clinically confirming the reliability of the "plain scan CT+AI" cancer screening technology path. The Dharma Hospital Medical AI team is working with several of the world's top medical institutions to explore new low-cost and efficient ways to screen for multiple cancers using AI technology, hoping to detect a variety of early cancers in a single plain CT scan. At present, this work has made phased progress in seven high incidence cancers, such as pancreatic cancer, esophageal cancer, breast cancer, liver cancer, gastric cancer, colorectal cancer, and so on. The research results have been published in medical journals such as "Nature Medicine" and "Nature Communications" and AI conferences such as CVPR / MICCAI / IPMI.

According to reports, the Dama Hospital Medical AI team has been committed to the research on the integration of AI and medical images for a long time, focusing on three major directions: accurate cancer diagnosis and treatment, accurate chronic disease diagnosis and treatment, and neurodegenerative disease pre-screening, and developed the whole process of cancer diagnosis and treatment technology, including scale screening, accurate diagnosis, prognosis treatment, and response evaluation. The team developed the CT image COVID-19 AI assistant diagnosis system in the early stage of COVID-19 's epidemic, which was rated as the national advanced group of science and technology in anti-epidemic by the Ministry of Science and Technology.

Link to "Nature Medicine" article attached

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.