Medical Network July 4th Medical AI is very hot, but the actual application of the application is very small, the current improvement in imaging and diagnosis, coupled with the true artificial intelligence talent has not yet poured into the medical field, the medical AI road is still long.
The east wind of artificial intelligence has already blown into the medical field.
Unbalanced supply and demand of quality medical resources, long training period for doctors, high rate of misdiagnosis, rapid changes in disease spectrum, rapid technological changes, increased aging of the population, growth of chronic diseases, increased attention to health, and the birth of medical AI (artificial intelligence) development of.
Jiang Xiaodong, founder of Changling Capital, told the 21st Century Business Herald: 'Why are we so optimistic about the application of artificial intelligence in China's medical system? Because in China's medical system, medical quality is not sustainable, medical resources, whether it is grassroots or Head, unbalanced distribution is common, and the problem of chaos caused by misalignment mechanism is also very common.
In response to these problems, artificial intelligence may play a key role in leading the industry change. However, the high barriers and specialities of the medical industry are also destined to be a rough road to AI empowerment.
Although the trials of medical AI products are diverse, they have not really landed, and products that can meet the clinical use scenarios are still absent. Most of the current research and development in hospitals is scientific research cooperation and experimentation.
'风口' empowerment
The slogan of AI medical treatment is blowing to the reconstructive medical system.
Specifically, based on big data, AI empowers the head hospital's medical capabilities to provide primary care, develop complementary medical treatments for different diseases, and allow primary hospitals to share the medical skills of the head hospital, ultimately distributing medical resources evenly. At each level.
Promote the development of artificial intelligence, such as cases, images, genes, and establish verifiable, repeatable medical standards by processing large amounts of high-quality medical big data. This enables patients to be diagnosed before, during, after, or both in and out of hospital. Enjoy standardized medical services.
'The future of artificial intelligence is not a substitute for doctors, it is not just a supplementary doctor, but a necessary cornerstone for helping to reconstruct the medical system and establish a new infrastructure.' Jiang Xiaodong said.
The road to empowerment of medical AI is based on deep learning of big data.
If artificial intelligence is divided into three dimensions: algorithm, computing power and data, the main opportunities in the industry now focus on data and application. The core of competition lies in the quality and quantity of data. Data is high-quality medical clinical or life data. The data is the sharpening stone. 'Without a good sharpening stone, there is no good knife.' Jiang Xiaodong told reporters including the 21st Century Business Herald.
Compared with humans, the biggest advantage of AI is computational efficiency, especially in data-intensive, knowledge-intensive, brain-intensive industries.
Back to the medical field, from the world Start a business In terms of company practice, specific applications include insight and risk management, medical research, medical imaging and diagnosis, lifestyle management and supervision, mental health, nursing, emergency room and hospital management, drug excavation, virtual assistants, wearable devices, etc.
The changes in artificial intelligence for China's new healthcare system infrastructure include: new medicine, ie disease screening and prediction, drug screening in patients, tumor diagnosis and treatment, drug discovery, new targets, discovery of new markers And new medical services, including data structuring, standardization, multi-source heterogeneous data aggregation mining, assisted diagnosis, optimized treatment of lesion delineation, health management, etc.
Among the many subdivisions, the field of auxiliary diagnosis represented by the imaging department has taken the lead.
'The imaging doctor writes hundreds of reports every day, and the fatigue leads to a drop in quality. If a good AI assistant solves this part of the work, it is meaningful to liberate the doctor and invest more time in the disease research and direct communication with the patient.' Liu Shiyuan, a professor of radiology and nuclear medicine at Changzheng Hospital of the Second Military Medical University, said: 'The disputes received by the hospital for more than 7,000 yuan of PET-CT are less than 120 yuan for a normal CT, precisely because of the actual operation of PET-CT. In the case of patients, they have more direct communication with doctors. In the future, the path of AI assisting doctors should also restore the value of doctors.
Medical imaging AI has emerged from the laboratory as a typical auxiliary diagnostic field, and is about to usher in a wave of commercialization. This year, there are also frequent highlights. Some intelligent imaging diagnosis enterprise Entering the approval process for the approval of the three types of equipment certificates, it is expected to formally enter the commercialization stage. The speed of AI imaging products is accelerating, and the product performance maturity is continuously improved.
On the other hand, in addition to the field of auxiliary diagnosis, investment trends have also begun to tilt towards AI+ drug discovery applications in recent years. The application of deep learning technology to preclinical research on drugs can quickly and accurately mine and screen suitable compounds or organisms; Cycle, reduce the cost of new drug research and development, and improve the success rate of new drug research and development. In 2015, Atomwise based on the existing drug candidate application, the artificial intelligence algorithm successfully found two candidate drugs to control Ebola virus within one day.
In the field of medical AI, capital forces are racing around, and Internet giants are rushing to the beach.
2016 is considered to be the first year of artificial intelligence + medical investment in the country, a total of 27 companies in 2016 to finance, of which 16 companies financing more than 10 million yuan.
Various domestic companies are vying to arrange medical AI tracks. The development of medical artificial intelligence is rapid. More than 28 startup companies have received financing last year, totaling more than 1.7 billion yuan.
On November 15, 2017, Tencent entered the first batch of national new generation artificial intelligence open innovation platform list. Previously, Tencent first single hospital As a breakthrough, then cooperate with the hospital through the establishment of a coalition or the establishment of a joint laboratory.
Traditional Internet giants with strong artificial intelligence talents and technology reserves are also deploying medical artificial intelligence. Ali health Launched medical AI product 'Doctor You' with Wanliyun, Tencent launched medical imaging AI products, Xunfei Medical launched image-assisted diagnostic system, intelligent medical assistant.
Li Zhifeng, general manager of Tencent's Internet Business Unit, told the 21st Century Business Herald: 'Tencent's goal is not to cut into too many fields at once, but to deepen the technology first, and then slowly expand. The medical AI track has a lot of expansion. Space, the industry is still in its infancy, Tencent hopes to attract more companies to the film platform, we also hope to cooperate with companies that have already gained advantages in this field, and choose to cooperate with Zero Technology because of its advanced data. Advantage. '
Zero Technology has entered into a partnership with more than 500 top three hospitals and established a data center. The Zero Health Medical Big Data Platform has gathered more than 2.8 million cases of effective cancer patients, with a single disease tumor penetration rate of more than 60%. In other words, new each year 60% of tumor-related cases will enter the system of Zero Technology.
According to the 21st Century Business Herald reporter, Zero Technology has completed the D round of financing in the first half of the year, and this round of financing reached 1 billion yuan, which is expected to become the first unicorn enterprise in the field of medical big data and artificial intelligence. According to close sources, Established by the State Council, one of the world's largest sovereign wealth funds, China Investment Corporation, is a very important investor in this round of financing.
Globally, investment turmoil appeared in 2014, mainly in the United States, the United Kingdom and India. Among them, cancer big data company Flatiron completed financing of 130 million in 2014, completed financing of 175 million US dollars in 2016 and 2.1 billion US dollars in 2018. The total price was acquired by Roche Pharmaceuticals and became the first company in the world for medical artificial intelligence companies to be acquired by pharmaceutical companies. Artificial intelligence medical treatment continues to heat up.
'virtual fire' cooling
According to statistics, the potential market for artificial intelligence + assisted diagnosis and treatment is huge, at least the scale of revenue above tera billion. However, artificial intelligence medical care is still a toddler.
Liu Shiyuan told the 21st Century Business Herald: 'The medical problem is very complicated, and there are many dimensions, different from artificial intelligence in other fields. For enterprises, what is needed is to dig deep into the clinical problems.'
For example, the problem of prominent spine, the imaging doctors have a great demand, but no one or the company has begun to do this subdivision. Because there are many influences involving the protrusion of the spine, the recurrence is high. There are cone dimensions, shoulder circumference Dimensions, saccular capsule dimensions, cervical vertebrae dimensions, and soft tissue dimensions, etc. Not to mention the types of diseases involved, cone morphology, density, MRI and other difficult challenges.
If a seemingly simple doctor reports that it falls to the machine level, it can only be realized by deep learning of big data of single disease and slowly integrating into a multi-task composite model.
'This journey is very long and needs innovation and breakthrough,' Liu Shiyuan said. 'At present, the application of AI in the imaging department is still only in the detection of lesions, and there is no application that meets the clinical scene. It is impossible to realize the detection from the nodules, to analyze and issue The integration process of diagnostic reports. '
There is no such thing. Qian Chaonan, deputy director of the Cancer Center of Sun Yat-sen University, and the vice president of the Cancer Hospital, also pointed out at the 2018 China (Guangzhou) VC roundtable summit last week: 'AI is very hot, and there are many bubbles. Computer experts who master artificial intelligence technology Companies with technology are excited and active, this is called a hot one. The other is cold, medical applications and clinical, this model is clearly not conducive to the development of the industry.
For the enterprise layout, the problem of homogenization of competition in the AI medical field is equally serious.
Take the most advanced imaging department of AI application as an example. The lungs and eye applications are relatively simple, and the enterprise has a lot of ground-breaking layouts. However, for example, the examination of complex parts such as the spine and abdominal diagnosis is relatively rare. At present, only the lungs are available. It is relatively mature with the fundus. The most widely used is the detection of pulmonary nodules, and some diseases of the lungs, including the differential diagnosis of pulmonary nodules. In the future, cerebral hemorrhage, fractures, coronary artery, the liver may also slowly enter Public view.
In the small-area AI medical landing, Liu Shiyuan pointed out that the clinical treatment is the most stressful, and the expectation of AI landing is also the highest. In the imaging department, the evaluation, the chest CT scan, the spinal MRI are all doctors work. Repetitive labor with high intensity, the demand for AI is naturally more urgent.
'The real field of AI is really small, and the single lung nodule detection is the earliest one year later. It will take at least three years to achieve the project that can report in line with the clinical scene,' Liu Shiyuan told the 21st Century Business Herald. , 'To achieve the landing of the clinical scene, what is needed is the composite learning and application of each single field.'
Compared with foreign medical AI, China is at the same level as foreign countries in the use of big data, and even has more application space. However, in the core algorithm competition of AI field, there is still a certain gap in China, which generally stays in foreign algorithms. The second innovation, the breakthrough in key areas is urgently needed. 'The core competition in the field of AI technology is mainly focused on the breakthrough of data, algorithms and neural networks.' Liu Shiyuan stressed.
Li Liping, co-founder of Zero Technology, chief clinical operation officer, told reporters in the 21st Century Business Herald: 'The threshold of medical AI is very high, algorithm experts and medical experts focus on different directions, medicine as a system science, data logic, modeling The layering is very complicated and rich. Only the deep integration of algorithm experts and medical experts can understand this problem by understanding each other's 'language'. So AI has no data, only data can't, there must be a large number of cases that can wake up silence. Compound type Talent In order to get out of the most critical step. '
In the actual operation of clinical research, the specific problem that needs to be solved is to rationally recruit patient samples and save clinical trial time. 'This requires an effective and feasible evaluation of the design plan in advance,' added Li Liping, 'simple The data collection software is very simple, but it is more difficult to collect long-term data and process the analysis results. The main development direction of Zero Technology is focused on big data + differentiation, forming closed-loop data logic from back-end systems to scientific research services. . '
In addition to the combination of big data and clinical, the conquerion of brain science problems also restricts the development of medical AI.
In view of the promotion of neural network to the development of AI, it can be understood that the neural network is a simulation of the brain system. The human brain is complex, and the current stage of imitation still remains in the surface of the nerve cells. This field, as the national key development direction Brain science, if you want to make a breakthrough, you will have to work hard.
In addition to the disconnection of clinical applications and the lack of deep-seated medical subdivisions, the development of universal AI is difficult. The founder of Yiyuan Intelligent, Wu Zhili also mentioned that in the field of AI, I have to admit that the top talents have not yet A large number of people are coming in, most of them are unmanned and in security, and the medical field is still very lonely. In addition, the AI algorithm talents are different from the medical talents knowledge system. How to combine their respective advantages to maximize their value is also worthy of thinking.
'The company specializes in AI or AI, everyone is developing in the field of artificial intelligence. Simple statistics, there are more than 120 AI in the medical field, and now the gap in the field of medical AI, especially Experts in the AI core algorithm may be the gap in the gap. ' Liu Shiyuan said.
