Written by Chen Dengxin
Editor/ Li Wenjie
排版/ Annalee
China's AI has once again attracted the attention of the other side of the ocean.
A few days ago, Stanford University released the "2024 AI Index Report", revealing that the investment in artificial intelligence in the United States in 2023 will be 67.2 billion US dollars, about 8.7 times that of China.
This means that China has come to the forefront of the world in AI cancer hunting.
How important is AI in the "new battlefield" of early cancer screening? Obviously, the United States is the pioneer in this way, why is China catching up? Where is AI landing?
The blood drop test for cancer is fake, and the AI cancer search is real
"This war, humanity has been going on for thousands of years. ”
The book "The King of All Diseases: The Biography of Cancer" revealed: "In the face of cancer, no one can say that it can be cured, and in order to catch up with the pace of this disease, human beings have repeatedly created and learned new knowledge and abandoned old strategies. ”
As a result, AI is expected to be high.
Then came the "blood drop test for cancer", a company called Theranos, which claims to use high-tech means, including AI, to quickly detect cancer by collecting a few drops of blood from the fingertip and automatically generating a diagnosis report of the patient's condition.
With the aura of "blood test for cancer", Theranos became a unicorn in Silicon Valley, with a valuation of $9 billion at one point, and its founder was once called "the female version of Jobs".
However, the paper could not contain the fire, and the truth of the scam was finally revealed, and the founder was also imprisoned.
Top 10 countries with the highest number of new cancer cases
Although there is a lot of chaos, the long-cherished desire of human beings to overcome cancer with the help of AI has not changed, and compared with the "blood drop cancer test", AI cancer search is faster and more stable.
Among them, Google can be called the global "pioneer" of AI cancer hunting.
IN 2016, WHEN GOOGLE PARTICIPATED IN ISBI'S REGIONAL CANCER CELL DETECTION COMPETITION, ITS AI TOOL LYNA BECAME FAMOUS FOR ITS EFFECTIVE TRACKING OF BREAST CANCER, AND LATER RELEASED AN ARTIFICIAL INTELLIGENCE DETECTION SYSTEM FOR BREAST CANCER, WHICH OUTPERFORMED EVEN MEDICAL EXPERTS.
Despite this, Google's AI model has not been implemented on a large scale.
Ruth Porat, Google's chief investment officer and CFO, once admitted: "Google's use of AI in healthcare – from predicting patients' illnesses through analysis and research of electronic medical records to aiding in the detection of diseases such as lung cancer, while we are still in the early stages of technology development, the results are promising." ”
In layman's terms, Google's AI cancer search is a prospective study that is in the stage of scientific research and experiments, and there is still a long way to go in large-scale clinical practice.
As a result, global cancer screening relies heavily on traditional approaches.
The problem is that the accuracy of traditional methods is closely related to the quality and experience of doctors, resulting in high cancer detection costs and low prevalence of early cancer screening.
Large-scale early screening, China came out on top
Correspondingly, China is about to take the lead in entering the golden age of AI cancer hunting.
In November 2023, the medical AI team of the Damo Academy, together with a number of medical institutions around the world, released the pancreatic cancer detection model PANDA, which uses AI to magnify and identify those subtle lesion features in non-contrast CT images that are difficult to identify with the naked eye, so as to achieve early pancreatic cancer detection and overcome the problem of high false positives in previous screening methods.
In a word, relying on the method of "non-contrast CT + AI", the purpose of large-scale early screening of pancreatic cancer can be achieved.
More importantly, the accuracy rate of judging the presence of pancreatic lesions is as high as 92.9%, and the accuracy rate of judging the absence of disease is as high as 99.9%, and it will not bring additional radiation and economic burden to patients.
AI cancer finding accuracy is high
Professor Gu Yajia, Director of the Department of Diagnostic Radiology at Fudan University Cancer Hospital, said: "Imagine that we can find out whether pancreatic cancer can be detected by doing the simplest non-contrast CT scan when we go for a physical examination, which will help many pancreatic patients and reduce the occurrence of tragedy." ”
Based on the research results of PANDA, the DAMO Academy is exploring AI multi-cancer screening technology, which is expected to cover 8 types of fatal cancers such as pancreatic cancer, esophageal cancer, lung cancer, breast cancer, liver cancer, and gastric cancer, and 5 chronic diseases such as osteoporosis and cardiovascular disease.
THERE ARE TWO REASONS BEHIND THE PANDA MODEL.
On the one hand, the "Healthy China" strategy is strong.
In 2016, the Healthy China 2030 Plan was promulgated, and the 5-year cancer survival rate in China increased from 40.5% in 2015 to 43.7% in 2022, while the Healthy China Action - Cancer Prevention and Control Action Implementation Plan (2023-2030) proposed to further improve the construction of the cancer prevention and control system.
As we all know, early screening is the key to cancer prevention.
It can be seen that the PANDA model not only has a wide range of application scenarios, but also has the urgency of time, which creates conditions for large-scale clinical practice, and has been first deployed in Lishui Central Hospital and Jingning County People's Hospital in February 2024.
Lv Le, head of the medical AI team of Damo Academy, said: "We are working with many top medical institutions around the world to explore new methods of low-cost and efficient multi-cancer screening using AI technology, hoping to allow people to detect a variety of early-stage cancers through a single non-scan CT. ”
On the other hand, Google is distracted.
An industry insider told Zinc Scale: "In fact, the underlying logic of Google and Damo Academy's AI empowerment is the same, which can improve the accuracy of cancer screening and reduce the waiting time of patients, but Google's efficiency is difficult to describe." ”
The above-mentioned industry insiders further said that AI has always been Google's background color, and it has pressed the fast forward button of AI early, but it has not made much progress on the ground, and the same is true for "AI + medical care", with a more comprehensive layout and profound technical heritage, but there is less "one foot in the door".
In fact, Google disbanded its independent Google Health division in 2021, as evidenced by this.
What's worse is that in 2022, Google once became the sixth largest short-selling target on Wall Street, and had to reduce staff and increase efficiency to give an explanation to the capital market, and in 2023, it will be "stolen from the tower" by OpenAI and chased to become Google's keyword.
In this way, AI cancer hunting has not become Google's focus, and it is not surprising that it has been overtaken by the corner.
Where has AI landed?
It can be seen from the above that the implementation of AI is more important than technology.
At present, AI has entered the "era of large models", and the industry can be described as a hundred schools of thought, including volume computing resources, volume parameters, volume application scenarios, volume business marketing, volume data lists, and ......
However, the large model has moved from early adopter to practical, and the real winner in the future must be the enterprise that is deeply engaged in application scenarios, which puts forward higher requirements for players.
Taking "AI + healthcare" as an example, AI has penetrated into multiple scenarios such as medical treatment, treatment, follow-up, and physical examination, assisting doctors to improve work efficiency, helping patients to improve medical efficiency, helping various medical institutions reduce costs and increase efficiency, and alleviate the shortage of medical resources......
For example, voice AI is used to intelligently identify the content of conversations between doctors and patients, and then input the data into electronic medical records to automatically create clinical records, thereby improving the effectiveness of doctors' diagnosis.
Of course, there is no shortage of challenges in landing.
Medical data is naturally sensitive, and how to obtain it in a compliant, complete, sustainable, and high-quality manner is related to the accuracy, reliability, and legitimacy of AI.
Economy is another difficulty that cannot be ignored.
A securities analyst said in an interview with the 21st Century Business Herald: "We believe that the economic benefits of quantitative AI technology in long-term operation can be evaluated by analyzing the effects of AI technology in improving diagnostic accuracy, reducing the misdiagnosis rate, and optimizing treatment plans, and can also be combined with the practice of DRG/DIP in medical lean management to further improve the implementation of DRG/DIP policies in the future." ”
China leads the way in the number of AI patents
In addition, applications have become the "keyword" for AI landing.
Alibaba launched "Taobao Ask", JD.com launched "Jingyan", both focusing on e-commerce shopping, ByteDance launched "Doubao" and "Little Wukong", focusing on search and video editing, Tencent launched "Xiaoqin", "Unaccompanied" and "AI Listen Together", focusing on social networking and music, Baidu reshaped its products and launched AI native applications such as "Baidu GBI" and "Cloud Yiduo......
According to Gartner, more than 80% of enterprises are expected to use generative AI APIs (APIs) or deploy applications that support generative AI by 2026.
All in all, the competitive highland of AI is the industry, and only by taking root in the industry and solving the practical problems of the industry can it blossom and bear fruit, release greater potential energy, and promote the rapid development of new quality productivity.
It is not difficult to see that Chinese AI is running forward.