(Original title: Medical AI on the air outlet: giant high-profile competition industry is full of fog)
Our reporter Xiao Meili reports from Guangzhou
Reading
At the same time, medical artificial intelligence is full of pain points from R&D, landing to industrialization. The "information isolated islands" of domestic hospitals have made the medical data threshold long-term high. The acquisition of medical big data is difficult and expensive, and the quality of data is poor. This has become the focus of Tucao, and it also directly affects technology research and development and landing.
Medical AI (artificial intelligence) is really hot.
Tencent Yingying’s debut on the debut came with a touch of playfulness. At the meeting that Tencent created, the first hour and 55 minutes mainly stated Tencent’s achievement in the “Internet+” field. Near the last five minutes, artificial intelligence broke through the microphone and entered the eardrum. Appearance.
For such debut, Chang Jia, the head of Tencent's “Internet+” medical service, told 21st Century Business Herald reporter: “The leading department of itself is Tencent’s Internet+ Business Unit, which emphasizes solving user problems in real situations and has no intentions. Take a technology as the main line to advance the work."
Although the appearance time is short, the amount of information is enormous. This is Tencent's first artificial intelligence product applied in the medical field. It aggregates the technologies of Tencent's internal teams including AI Lab and Excellent Chart Lab. The most mature laboratory for intelligent screening system for esophageal cancer can achieve 90% accuracy. . At the same time, Tencent also announced the establishment of a joint laboratory for artificial intelligence medical imaging, the Cancer Hospital of Zhongshan University (Guangdong Provincial Institute of Esophageal Cancer), the Second People's Hospital of Guangdong Province, and People's Hospital of Nanshan District of Shenzhen City became the first cooperative hospitals.
It's not just Tencent who shoots high-profile. In the past six months, technology giants such as Baidu, Ali, Google, Microsoft, and Apple have spared no effort in deploying medical artificial intelligence. They are characterized by a focus on technology, huge amounts of money, frequent acquisitions, and conquered areas throughout the industry chain.
At the same time, there are also entrepreneurs and medical institutions, who often support R&D and even survival through financing. They mainly focus on areas such as medical imaging diagnosis, virtual doctors' assistants, and medical big data, which are rapidly becoming commercialized. Some rely on exclusive technology to obtain large scales. Financing entrepreneurs are also trying to break basic scientific research.
At the same time, medical artificial intelligence is full of pain points from R&D, landing to industrialization. Domestic hospitals' "information isolated islands" make the medical data threshold long-term high, directly affecting technology research and development and landing. At the industrialization level, there are problems such as the lack of industrial supervision and the lack of a supporting approval system. Whether medical artificial intelligence is a virtual fire or is still in its ascendant, it is still necessary to set aside heavy fog.
Tencent enters the game
Tencent Yingying includes 6 personal intelligent systems, involving diseases such as esophageal cancer, lung cancer, sugar net disease, cervical cancer and breast cancer. The early intelligence screening system for esophageal cancer is the most mature and belongs to medical image recognition. It is reported that this system has been used in preclinical experiments in less than 4 seconds when screening an endoscopy.
For the significance of early screening for esophageal cancer, Professor Jian-Hua Fu, Director of the Hospital Administration Department of Sun Yat-sen University and Director of the Guangdong Provincial Institute of Esophageal Cancer pointed out that early esophageal cancer can be discharged 3-5 days after endoscopic surgery, and the غير مجاز مي باشدt of surgery is only late. One-third of esophageal cancer treatment غير مجاز مي باشدts, postoperative complications are few. "However, due to the lack of sufficient cognition and effective early screening methods, the detection rate of early esophageal cancer in China is currently below 10%."
Chang Jia thinks: “In the medical field, there are few doctors at the grassroots level. Many diseases are relatively uncommon, and doctors have a long learning cycle. Therefore, in the early screening and more similar medical fields, artificial intelligence should enjoy a relatively large Development opportunities."
The 21st Century Business Herald reporter learned that in the cooperation between technology companies and medical institutions, technology companies generally play the role of technology providers. Medical institutions provide medical data, build a comprehensive knowledge base for clinical pathways, procedures, and disease diagnosis, and images continue. Collected systems, etc., in the development process, the degree of integration of the two parties directly affect the final results.
Prior to the three hospitals in hand, Tencent often shot in the medical field “buy, buy, buy,” and the field of medical artificial intelligence is no exception. For example, last year's carbon cloud intelligence for digital life management using gene big data completed A round of financing of nearly 1 billion yuan, and Tencent was one of the major investors. In early March of this year, Grail, an early cancer screening company founded by Illumina, a US giant, announced that it had received 900 million yuan. US dollar financing B round of financing, there have also appeared in the shadow of Tencent.
However, it is the first time to introduce medical artificial intelligence products personally. "In fact, the medical field is very difficult and the medical system is too complicated." Chang Jia said bluntly. "Tencent never mentioned subversive medical care. We are connected to medical care. Under the open system, we will invest in a lot of companies. But we have found some There is no way to completely separate Tencent's basic capabilities. We also need the underlying capabilities."
At the end of last month, Tencent announced the launch of an artificial intelligence accelerator, which will connect the capabilities from Tencent's AI Lab, Best Picture, Tencent Cloud, etc., and provide more than 20 artificial intelligence technologies.
At present, each AI team within Tencent has its own priorities: The TU Lab focuses on the study of machine learning, pattern recognition, and cognitive technology. The results have been applied in products such as P pictures every day; the WeChat AI team mainly focuses on speech recognition and pattern recognition. The core application is to provide voice input and text in WeChat; AI Lab mainly focuses on image recognition, speech recognition, natural language processing and machine learning.
After establishing a joint laboratory with the hospital, Tencent Yingying will continue to use these technologies for image recognition and deep learning. “Artificial intelligence is a type of grassroots technological capability that is open. It also means that we are starting to move closer to the depth of medical integration. Before the medical insurance payment, user service, and WeChat public account we laid out were all in the public and hospital service sector. Ecology will play a long-term strategic support role." Chang Jia said.
Giants competition
Compared with Tencent, Baidu and Ali's medical artificial intelligence achievements have already been published.
Just last month, Ali Health announced the joint medical imaging center of the Wanliyun Medical Imaging Center to release Doctor You, a clinical medical research diagnostic platform, a medical aids detection engine, and a physician ability training system. In addition to Ali's introduction of the ET medical brain in March of this year, and the "Future Hospital" program that was launched in 2014, Ali's penetration in the field of medical AI continues to deepen.
Baidu started to enter the medical industry in 2010 and has a layout in registration and hospital services. After announcing the abolition of the medical division, Baidu will shift its focus to medical artificial intelligence. Baidu’s medical brain has been launched. Li Zheng, general manager of Baidu Cloud Partners, told the 21st Century Business Herald reporter: “We are still working hard.”
BAT has entered the market so that the domestic medical artificial intelligence is very popular, but the industry generally believes that the technical gap between China and foreign countries in this area is still very long. In the commentary on the senior partner of the company: "The medical artificial intelligence has made breakthroughs in the basic research and technical aspects, and the basic business exploration is started by foreign companies."
Jin Lu, director of Dana Science and Technology Products, added: “There are many public databases in the world. After the artificial intelligence heats up, there are more and more public databases in the medical field. This is very good for the technology companies that carry out basic research and can be fast. To form their own technology, or to verify that the original technology can be transferred to the medical field can not make achievements. These basic R & D are generally tech giants do, foreign companies do a little earlier."
It is understood that the application of artificial intelligence in the medical field includes: auxiliary diagnosis and treatment, medical imaging, drug discovery, health management, emergency room and hospital management, wearable devices, nutrition management, virtual assistants, and so on. Its industry chain mainly includes the basic layer, technical layer, and application layer. The entry barriers and core advantages at each level are different. Participants, investment opportunities, and returns are also different.
Among them, the basic layer is mainly laid out by several technology giants, including IBM, Google, Microsoft, Amazon, Ali, Baidu and so on. The giants generally choose to enter the field with a large amount of computational requirements. Such enterprises basically belong to high input and high returns.
The technical layer is the infrastructure of the artificial intelligence large eغير مجاز مي باشدystem. It needs a certain scale of engineering teams to integrate with the industry to form a solution or a common technology platform. Or there are many algorithms, frameworks and tools to form an algorithm tool platform and an eco-platform for developers. These enterprises are suitable for mid- to long-term investment layout.
The typical case of these two levels is IBM's Watson Tumor Robot, which has been commercialized and is progressing to a Chinese hospital one year later. However, Tang Heng pointed out: “The cycle of basic R&D is already very long. If we want to really use it, we must combine it with the depth of the industry. Watson is now the foremost product of medical artificial intelligence, but it is also in a stage of technological growth. It is not comparable to clinicians. The results of other companies' R&D are even earlier."
The application layer is the channel for realizing the application scenario. According to incomplete statistics, there are currently more than 90 medical artificial intelligence startup companies in the application layer and technology layer. The types of applications include medical industry solutions and medical industry application platforms. The competition is even fiercer and business realisation is coming faster.
In January of this year, Arterys, headquartered in San Francisco, USA, announced that its product, Arterys Cardio DL, received FDA approval for the analysis of cardiac MRI images. This is the first FDA approved clinical medical artificial intelligence product.
Jin Lu believes that: "The real application of technology to the medical field requires a deep understanding of the industry. Many technologies are beyond humans in laboratory or competition verification. However, in the clinic, patients have more complex disease problems and more complex application scenarios. This is an opportunity for SMEs to respond to changes more flexibly and revise them."
Business problems
As the players who come into the game get together, medical big data is often the focus of Tucao. The contents of Tucao are nothing more than data acquisition difficult and expensive, and poor data quality.
Yin Heng, the chief dermatologist of Xiangya Second Hospital, focused on the dermatological artificial intelligence system of the hospital and Dingxiang Science & Technology Co., Ltd.. After introducing the initial model, Yin Heng still exclaimed: “The image resources of skin diseases are relatively easy to obtain. Other departments need to use a lot of image data. However, the system needs to collect a huge, massive database of dermatological resources. It is very difficult for a hospital to collect a large amount of dermatology resources, and it takes a lot of financial, material, and human resources.”
Yan Kezhou, a senior engineer of the Tencent Architecture Platform Division, is working on a breast cancer pathological image recognition project. The data problems he encountered came from doctors' habits. "We need more data for wave plate scanning. This data is very small. It needs to digitize the entire wave plate. However, this machine is not widely used in many hospitals. Many doctors are also more resistant to such machines. They have come into contact with the medical school. The training is done using a microscope."
Medical data is equal to the staple food that sustains survival and growth for artificial intelligence. With data, artificial intelligence machines can be trained. Application transformation also needs to be connected with clinical data. For domestic companies, the main channel for obtaining data is cooperation with public top-three hospitals. At this point, the lack of interconnection between hospitals and hospitals has become a major stumbling block.
Tang Heng pointed out: “Chinese hospitals all have private clouds, and all private clouds are independent, and data is not shared. Some data are now occupied by Internet registration platforms, but the number is small and the quality of data cannot reach the medical R&D level. The hospital connects the system and then grabs the data. It is generally selective crawling and cannot obtain the full version of the data resources."
In contrast, European and American countries have a complete electronic data flow for medical data, but the final result is that the data portal is occupied by giants, and the غير مجاز مي باشدt of purchasing data is not low. Take IBM Watson as an example. In 2015, IBM acquired three medical big data companies, Phytel, Explorys, and Merge Healthcare. Last February, IBM spent another US.6 billion to acquire Truven Health Analytics, a health big data company. At this point, IBM invested 4 billion US dollars in less than a year to obtain data.
Solved the data problem, medical artificial intelligence began to enter the industrialization, domestic approval and supervision in this field is still in a blank stage.
Tang Heng bluntly said: “The future is based on medical devices, systems or other methods of supervision has not yet been conclusive. The products that have been used in clinical use in China are basically scraping the ball. It is difficult to set prices based on the same standards for similar clinical projects. A relatively high price."
The U.S. practice is that the U.S. FDA formally established a new department consisting of software engineers and developers, AI technology, and cloud computing experts in May this year. It is dedicated to reviewing digital medical and AI technologies, and formulating examination and approval and supervision. Specifications and standards.
In addition, the five major US technology giants (Google, Facebook, Amazon, IBM, and Microsoft) have jointly established an AI cooperation organization to ensure that the industry can develop safely, transparently, and reasonably in the future.
"Actually, I don't have to worry about supervision. If I come out with a clinically useful product, I will definitely be able to get it approved." Jin Lu believes that "the current problem is that there are some conceptual technologies in China that are being launched, but few are able to land." less."
The business model is still to be explored. Tang Heng believes that at the current stage, the business model innovation of the medical artificial intelligence enterprise at the application level has encountered a bottleneck period. It basically relies on screening services and analytical reports to realize it. Whether or not the technology upgrade and data accumulation in the later period can bring about some qualitative changes still need to be verified.
"Looked at many projects, but the investment is very few, many projects are not profitable or even loss. The project's valuation is also relatively high, because now is a catch, many capitals want to invest in a successful project to achieve Market leader." He said.
2.3-24GHz Frequency Synthesizer
Modular Synthesizer,Wideband Synthesizer,Dual Frequency Synthesizer,Digital Frequency Synthesizer
Chengdu Zysen Technology Co., Ltd. , https://www.zysenmw.com
برچسب:
،
ادامه مطلب
بازدید:
