AI medical floor hospital acceleration | Image data easier | 'Entrance'

Pharmaceutical Network March 7 hearing in Guangzhou, a top three hospitals in the imaging industry for 8 years, a doctor, with a look at the door to describe his work, we need to go through the historical image contrast, quantitative analysis and other steps in order to Patient finished film to make a basic diagnosis and treatment ', and then, to decide what kind of patient needs treatment.

In the medical diagnosis, the value of the image is irreplaceable.90% of the medical data are images, such as CT, X-ray, MRI, ultrasound, PET, etc. If a patient with cancer is to undergo surgery, Tumor status, the degree of vascular stenosis and other information as the basis to determine the surgical options, medication programs and follow-up risk.

Advances in artificial intelligence (AI) technology and image recognition technology have given this work an effective assistant to the artificial intelligence medical image analysis system.

With the same two-dimensional medical image, doctors need to spend ten minutes to observe and reason, and artificial intelligence can be 'read' in tens of seconds after deep learning training. Artificial intelligence is expected to increase diagnostic speeds with sufficient big data support. 10 times, and thus greatly reduce the cost of diagnosis and treatment.

For doctors, efficient analysis can help them save the time of reading, reduce the misdiagnosis rate and provide a richer historical image comparison. The hospital is also willing to see artificial intelligence digital imaging of medical imaging to facilitate the construction of medical databases , Thereby reducing the cost of treatment programs.

The combination of capital, technical and medical data is the three most popular artificial intelligence medical imaging take-offs in the world. Is this "reading" assistant still in the internship unique?

Compared with 2016, the artificial intelligence medical industry in 2017 is even more hot, but also craves the fruits of landing.

'Everybody is tired of talking. After all, no one can live on 'chicken blood'. 'European think tank health industry analyst Shang Hao told the 'Financials' reporter, 'Is there an application level can be called 'hands down' What? I think it is artificial intelligence medical image analysis.

Compared with other areas of artificial intelligence medicine, the advantage of artificial intelligence medical imaging is in the data. The imaging data is not like a medical record. It contains medical history, patient information, symptoms, treatment methods, and more recovery. Information, its own high degree of information integration - a medical pathological film contains a large amount of high-value information. Therefore, compared with other medical data, image data processing is less difficult and processing value is higher.

'Original medical images are very high-dimensional and complex, and artificial intelligence transforms high-dimensional data into a low-dimensional, easier-to-handle problem.' 'Wang Xiaozhe, chief architect of Zero Krypton Technology, Reporters said that medical imaging data itself well fit the artificial intelligence characterization model algorithm.

The unified imaging data standards make it easier for algorithms to enter and aid in the construction of diagnostic models.

Each hospital's image data, not only in a radiology department, involving almost every clinical department, which also means that the image data does not exist in a single information area. Peking University Cancer Hospital Information Department of Heng counter-anti-repair "Finance" Reporter, medical imaging data in the hospital is the largest amount of data, and is standardized, more convenient for machine reading, it is very important.

Many people in the field of medical imaging brain. As early as 2003, Philips health Zhou Zhenyu, senior director of science clinical sciences, and his mentor had the idea of ​​creating an imaging big data platform, but 'many of the challenges that we were unable to overcome at that time, such as imaging quality, data and computer mismatch, diagnostic logic Nor is the way of thinking, which all led to the fact that we could not achieve true wisdom and medical treatment 15 years ago. "Zhou Zhenyu told Caijing reporter.

Just as an imaging physician needs to read a large amount of clinical images, 'feeding' the pathological image data is the most important way for an AI system to learn, and compared to the time when Zhou Zhenyu and his mentor started to think, pathological image data that are currently 'fed' More and more adequate, artificial intelligence analysis ability to grow.

Due to the relative abundance of data, developers are able to gather in the vertical field, and in 2016 the Beth Israel Medical Demonstration Center (BIDMC) and Harvard Medical School announced a partnership to develop an artificial intelligence platform for breast imaging. Pathological images, the system to complete the film on the identification of cancer cells and health areas, and in-depth learning technology framework to complete self-improvement and improve the recognition accuracy and efficiency.The platform leader Andrew Beck said the platform of the patient's breast image Analysis accuracy can reach 92%, combined with the pathologist analysis accuracy of up to 99.5%.

The uniqueness of medical treatment forced AI companies to have to cooperate with hospitals from the outset. Because in the health system, hospitals are relatively independent and the data is held independently, all AI companies have the opportunity to enter more. In 2017, targeting medical care. Artificial intelligence companies in the field frequently publish cooperation projects with hospitals. From the data published to the public, not only have the capabilities of medical image processing and analysis been improved, but also more clinical applications have been made.

An example of tracking lung nodules is through an artificial intelligence system that not only tells the patient where there are pulmonary nodules but also predicts the probability of malignancy by using the original data analysis and suggests probable patient screening reviews by probabilities or Not need to do biopsy, or the corresponding genotype of the examination of these clear information not only allows patients to become more aware of the disease, doctors and patients easy to communicate, but also to pay for weight loss, which is also the government happy to see.

Independent third-party medical imaging platform Huiying Hui Ying CEO Chai Xiang Fei provided to the "Caijing" reporter's data show that Hui Yi Hui Ying's reading volume has more than one million.This is accelerating its cut into the actual application scenarios are inseparable from the current Huiyi Huiying has access to more than 500 primary hospitals and more than 200 top three hospitals.

Airdoc, another prominent artificial intelligence medical company with a focus on ophthalmological imaging, collected hundreds of thousands of fundus photographs by domestic and foreign hospitals. In the second half of this year, Airdoc, together with Zhejiang Eye Hospital, Shanghai North Hospital Have reached a cooperation agreement, the establishment of a artificial intelligence ophthalmic image analysis related technology base.

According to incomplete statistics, currently entering the field of artificial intelligence medical imaging Entrepreneurship Company, has reached more than 40. In addition to vertical artificial intelligence medical company, the Internet giant's action becomes more and more obvious.

Given the long-standing data advantages of Internet giants, their involvement may directly affect the future pattern of change in this area, and the blessing of the policy has magnified the influence.

November 15, Ministry of Science and Technology in the next generation of artificial intelligence development planning and major science and technology project start-up meeting announced the first batch of national new generation of artificial intelligence open innovation platform list: relying on Baidu company to build automatic driving, relying on Aliyun company building city brain, Relying on Tencent to build medical images and relying on IMC to build smart voice, this is an artificial intelligence medical image, for the first time being listed as a separate category of artificial intelligence in the government level.

Tencent 'airborne' artificial intelligence medical imaging market, but in the list released 3 months ago in August, Tencent launched the first AI medical products 'looking film', mainly used in early esophageal cancer screening.Currently, The Cancer Hospital Affiliated to Sun Yat-sen University, Second People's Hospital of Guangdong Province, People's Hospital of Nanshan District, Shenzhen has joined this cooperation project.

Whether it be the Internet giant or a more vertical AI medical company, the results of the sinking to the hospital in 2017 are all significant - ensuring adequate access to food for artificial intelligence, discovering problems encountered in applications, More recently, there are more commercial prospects. These are all capitals are happy to see.

Excluding the artificial intelligence technology itself, the medical imaging market also reached a new critical point in 2017.

Between the growing market demand and the medical imaging resources, the fracture becomes significant. The booming growth of the medical imaging market has attracted artificial intelligence companies and they hope to become the earliest 'stitching glue' in this crack. At the same time, becoming the first 'miners' to arrive.

According to Yang Hongfei, CEO of Flint Creations, from the perspective of the current market size of medical imaging, patient-side growth is rapid, and imaging inspection revenue accounts for more than 10% of total hospital revenues. drug Revenue share. Flint to create June released the "medical imaging market spectrum and industry development analysis," pointed out that in accordance with the past five years, China's overall medical spending by 2020, China's medical imaging market will reach about 600 billion to 800 billion yuan.

It also lifted the capital's 'appetite' .According to incomplete statistics, as of press time, the domestic artificial intelligence medical imaging market to complete Angel Wheel financing enterprise There are three, Pre-A round of financing 2, A round of 7, B round of 3, C + round of a.

Among them, the largest amount of the three financings was concentrated in May and the second half. In May, IntuTech announced the completion of 380 million yuan in Series C financing, led by Gaochun Capital Group, Yunfeng Fund, Sequoia Capital, and Gaochun Capital. , Real funds with investment; In September, it was assumed that the technology announced the completion of 120 million yuan of round B financing, led by Qiming Venture Capital, Yuansheng Capital, Redwood China joint investment; In October, Huihui Huiying announced the completion of 'hundreds of millions' Yuan's B round of financing, the investor is Tat Tai Capital and 2 other investment institutions.

In the opinion of Wang Yuquan, president of consulting company Frost & Sullivan China, it is a very reasonable result for domestic capital to get together with artificial intelligence medical images. 'Improvement of the medical image of artificial intelligence imaging is an inevitably happening matter. This improvement can be seen, it is also the moment. The most popular, it is expected that the capital enthusiasm will not be turned cold by 2018. ' Wang Junquan told the "Finance" reporter.

However, it seems that it is not easy to cut a huge piece of cake. Even the leading IBM Watson, its Watson for Oncology has not yet reported its profitability.

Oncology Aid Solutions Watson for Oncology is working with medical image analysis to provide physicians with treatment recommendations and generate treatment plans to improve the accuracy of doctor care backed by IBM's own information analytics expertise and market acceptance Watson for Oncology entered China, the United States, the Netherlands, South Korea, Thailand and India, etc. However, the current profitability of IBM Watson for the time being did not appear in the public report.

Compared to Watson, a domestic artificial intelligence medical imaging company is still in the stage of application of disease screening, that is, to determine whether there is a certain type of disease in the image. No 'three steps, most of the artificial intelligence medical imaging company is still stuck in the first exploration.

For now, the achievements of AI are far from expected. 'Shanghai Long March hospital Wei Ruili, director of ophthalmology, told Caijing reporters: 'The AI ​​is mainly used for screening. When it is actually used, the doctor will review it again, just as if the patient took the diagnostic report of the local hospital, and we have seen it again.'

In addition, domestic companies are still concentrated in the field of diseases where medical image analysis requires relatively simple and relatively low value. Taking the lungs as an example, lung cancer recognition is a popular area of ​​artificial intelligence medical images. This is due to the natural contrast of lung image recognition. , It is easier to overcome the direction, but the specific symptoms of lung cancer does not have in-depth analysis capabilities.

Liu Zaiyi, professor of radiology at Guangdong Provincial People's Hospital, felt this way. Most cases of lung cancer in our hospital are under review. There are many metastases in the lungs of patients in stage III and IV, combined exudation and atelectasis. It is difficult to automate the comparison of these features, and these assisting physicians' products may indeed reduce some workload in clinical practice, but they are less helpful to doctors and have less application scenarios.

Get together more easily break through the field, means that the competition is more intense, the risk of being squeezed by the giants are also higher.And in the "Caijing" reporter interviewed by industry insiders, profitability and profitability problems is still their brains rely on After the problem.

In November, Everbright Securities analysts pointed out that service-oriented medical imaging downstream industries are in urgent need of innovation in service mode, and enterprises that quickly obtain sufficient resources in telemedicine imaging diagnosis and independent imaging centers will have greater advantages in future intelligent diagnosis of images.

Judging from the market reaction of the giants, IBM has been trying to open the Chinese market for many times, and Ali and Tencent, which have been operating frequently, have a very strong impact on the market once they are shot. "Reporter, 'According to my contacts in capital, it is expected that the giants will step up their actions in 2018. A round of 'big fish eats small fish' will begin immediately.'

Wang Yuquan, president of consulting firm Frost & Sullivan in China, believes that even though the giants will not aggressively integrate the artificial intelligence medical imaging market so quickly, the fights between small companies will be fiercer. Who can beat the unpredictable and may even be like sharing the bicycle industry In the same way, 沦 is the representative of the capital game.

As the earliest and most competitive "battlefield" in 2017, the artificial intelligence and medical imaging industry has encountered many problems that are difficult to find an empirical reference, and this also reflects its distance from tasting Artificial Intelligence Medical This delicious 'first soup' is the closest.