How long does it take for a world-leading medical research to be applied to ordinary people's daily medical treatment?3 months. existLishui, Zhejiang, Alibaba Damo Academy’s medical AI team and Alibaba Charity gave such an answer-on November 20, 2023, the top international medical journal NatureMedicine published the latest research results of Damo Academy,PANDA(PancreaticCancerDetectionwithAi) model passes "Plain CT+AI"The combination of "can accurately identify early pancreatic cancer lesions.
Three months later, Alibaba"Medical AI Multi-cancer Early Screening Charity Project"It was launched in Lishui, Zhejiang and completed deployment in Lishui Central Hospital and Jingning County People's Hospital.
Starting from the screening of pancreatic cancer and osteoporosis, it will also cover the detection of liver cancer, esophageal cancer, gastric cancer, colorectal cancer and other lesions that account for 38.1% of new cancers, as well as a total of 13 disease types such as fatty liver and pulmonary nodules. With the help of the latest medical AI technology, the people of Lishui can conduct early screening for multiple cancers and diseases during daily examinations.
Lu Le is the head of the medical AI team of Damo Academy, he hopes that this research can help more doctors and cure more patients in Lishui.
My own relatives also have cancer patients, and they also use our AI tool to determine whether the tumor is resectable.
PANDA is coming to Lishui
Dr. Zhou from the Radiology Department of Lishui Central Hospital, recent work has ushered in subtle but noticeable changes.
The image transmission system he uses daily is loaded with the PANDA model of Alibaba Damo Academy. In addition to the familiar imaging interface, he also has a powerful "AI assistant" that can assist in displaying inspection results and providing risk warnings.
With the help of PANDA, Dr. Zhou conducted screening and verification of past cases, and two cases of pancreatic cancer were discovered. In the actual examination environment in the past, the two patients were diagnosed with pancreatitis through plain scan CT. Pancreatic cancer was diagnosed only after subsequent enhanced CT or enhanced magnetic resonance for other reasons. He is very satisfied with the sensitivity of the "AI assistant".
pancreatic cancer, known as the "King of Cancer". Although it is not as common as gastric cancer, liver cancer and other cancers, it has no obvious symptoms in the early stage and is difficult to detect. When patients come to the hospital with symptoms, they are usually in an advanced stage and are extremely difficult to cure. The average 5-year survival rate is only 10%.
"Aiming at the hidden location of pancreatic cancer and no obvious representation in plain CT images, we constructed a unique deep learning framework and finally trained it as a pancreatic cancer early detection model."Guo Jianfei, product expert of DAMO Academy’s medical AI team and relevant person in charge of the PANDA projectIt is said that the model is directly facing the early screening of pancreatic cancer. Its principle is to use AI to amplify and identify those subtle lesion features in plain CT images that are difficult to identify with the naked eye.
The sensitivity perceived by Dr. Zhou is behind the extensive clinical verification that PANDA experienced before it was launched in Lishui.
The PANDA model constructed a training set containing 3,208 real patients, and ultimately passed multi-center verification with 6,239 patients from 10 hospitals/universities including the Shanghai Institute of Pancreatic Diseases and the First Affiliated Hospital of Zhejiang University School of Medicine, as well as real-world clinical verification with approximately 20,000 consecutive patients. Its pancreatic cancer identification sensitivity increased by 34.1% compared to manual diagnosis.
In addition, before the release of PANDA, no one knew that plain CT could be used to detect cancer. In the past, the detection and diagnosis of tumors mainly relied on enhanced CT, which required the injection or oral administration of contrast agents to patients and was not suitable for large-scale screening.
Sensitivity and convenience are the reasons why NatureMedicine rated PANDA.
PANDAof daily users exceptRadiology,besidesPhysical examination departmentGuo Jianfei said that it not only has the high popularity of CT examinations, but also takes into account the high sensitivity of artificial intelligence. It can screen for cancer and chronic diseases during patients' daily physical examinations and hospital examinations. It can screen multiple cancer types early at once, which will effectively improve the efficiency of cancer screening and the early cancer detection rate in asymptomatic people, thereby improving the prognosis of cancer patients.
For the people of Lishui, they may not have any obvious perception, but changes are already happening. The world's cutting-edge medical research results are the first to be implemented in Lishui. When they go to relevant hospitals for examinations and physical examinations, they can go from a single, shallow examination to predicting risks early without spending extra money, greatly improving the cure rate of diseases, and also greatly saving personal and family medical expenses and national medical insurance funds.
A golden age is ready
When DAMO Academy's PANDA-related paper was published, NatureMedicine magazine distributed an editorial titled"Tumor Screening Based on AI and Medical Imaging: A Golden Era is Ready"(AIandimaging-basedcancerscreening:gettingreadyforprimetime).
The Golden Age not only praises the cutting-edge nature of this research, but also is full of expectations for its large-scale implementation in the future. It is a "very promising method."
The same is true for DAMO Academy and Ali Public Welfare. They hope to achieve inclusive benefits and health equity through technological advancement, and use public welfare to promote advanced achievements in the medical field more widely and sustainably.
This is exactly why PANDA landed in Lishui. Lishui is located in a mountainous area in Zhejiang and is a key support area for Alibaba's public welfare rural revitalization.
One of the two pilot hospitals is Lishui Central Hospital, a municipal-level tertiary hospital responsible for the city's main critical and acute health protection work, and the other is Jingning County People's Hospital, a county-level secondary hospital. Jingning is also a county where Alibaba directly dispatches rural correspondents to help. They basically represent the level of the medical and health system in the Lishui area.
This means that we have truly stepped out of the ideal environment of the laboratory and into the grassroots medical system.
Guo Jianfei believes that PANDA has previously cooperated with some of the best pancreatic disease hospitals in China such as Shanghai, Beijing, and Hangzhou, with advanced equipment and standardized diagnosis and treatment procedures. Therefore, the quality of the data used for research is very high, and the construction and training of the model are relatively controllable.
But when it comes to grassroots hospitals, Guo Jianfei made an analogy, "It's like a high-definition TV turning into a snowflake black-and-white TV." They don't have the same advanced equipment as front-line hospitals, and many real-life situations are uncontrollable. However, under such ordinary medical standards, although it puts forward higher requirements for the PANDA model, it will be of great benefit to its wider popularization.
The more important significance is that after the PANDA model in Lishui is successfully verified, it can be replicated in batches and on a large scale in the future. Any hospital can copy the PANDA model at a low cost. This is the greatest significance. "Guo Jianfei emphasized.
As Lu Le, head of the medical AI team at DAMO Academy, said, "Because if the things we are doing are done well, they can be made available to the public, so that they can truly help thousands of patients."
A “dream team” dedicated to philanthropy and exploring medical AI
Lu Le has a PhD in computer science from Johns Hopkins University. He has been engaged in research and development of medical imaging and clinical informatics for more than ten years before leading the medical AI team of DAMO Academy. He hopes to build a research team that can help doctors and cure more patients.
My own relatives also have cancer patients, and they also used our AI tool to determine whether the tumor is resectable. The doctors at the local tertiary hospital thought that the chance of surgery had been lost. Later, it was proved that the judgment of our AI was consistent with the expert judgment of the top specialty hospital in Shanghai. The patient had a successful operation half a year ago and has now exceeded the estimated survival period of the local tertiary hospital.
After the PANDA model was published, Lu Le couldn't help but sigh, "We have been conducting research on pancreatic cancer for a long time, but we never thought that it would be used by our family members. Life is a dream, but it is also destiny."
Guo Jianfei, an engineer at DAMO Academy, is also "destined" to be associated with PANDA.
Twenty years ago, when he was studying for a bachelor's and master's degree at China Medical University, he developed a strong interest in how medical imaging is combined with computer technology. He also participated in the first computer-aided diagnosis textbook in the field of medical imaging in China.
There are many such dream chasers in DAMO Academy’s medical AI team. For problems that have yet to be solved, they have no set direction to pursue. They can only seek breakthroughs by relying on scientific research accumulation and bold assumptions in this field, allowing AI to learn and iterate through a large amount of relevant image data.
Guo Jianfei said frankly that the smooth advancement of PANDA in Lishui relied on more than 20 partners from Damo Academy. He also became a good back-to-back partner with Zi Luo from the Alibaba Promote Shared Wealth Working Group. Together with the staff from various departments and hospitals in Lishui, they formed a real "dream team."
The combination of this lineup has stepped out of Lishui Speed and created a Lishui model. Tumor screening based on AI and medical imaging is ready for a golden era.