Google Health倒下后，医疗AI商业化的前途何在？

After Google Health, What's next for medical AI commercialization?

医疗AI要如何应对商业化未果的窘境？“医学界智库”采访了数家头部公司的高管，以及相关从业者、投资人和医生。

How will medical AI cope with the failure to commercialize? Medical Think Tank interviewed executives at several leading companies, as well as practitioners, investors and doctors.

“内网升级导致AI系统不能用，今天还有这么多患者要体检，怎么办？”人工智能（AI）辅助医疗影像诊断公司医准智能创始人兼董事长吕晨的朋友圈里，出现一条体检中心主任的抱怨。这个处在人口密集南方城市的体检中心，在AI的帮助下，两个大夫一天就可以完成超过100个患者的影像诊断。这条讯息再次验证了吕晨的判断许多影像科医生早已习惯运用AI去辅助诊疗工作。

"The AI system cannot be used due to the Intranet upgrade, and so many patients need physical examination today. What should we do?" A complaint from the director of a medical examination center appeared in the friend circle of Lyu Chen, founder and chairman of AI-assisted medical imaging diagnosis company. With the help of AI, two doctors at the check-up center in a densely populated southern city can screen more than 100 patients a day. This message reaffirmed Lv's judgment that many radiologists have long been accustomed to using AI to assist in diagnosis and treatment.

对吕晨来说，医疗AI正迎来发展的好时机：全球范围内，医疗AI大约在10年前开始起步；2018年医生与AI的“对抗”比赛盛行，招致业内反感；2019年医疗AI进一步被影像科医生认可；2020年在新冠疫情的影响下，行业发展速度明显加快，国家药监局给相关产品亮了绿灯，商业化发轫。

For Lu, medical AI is coming at a good time: Globally, medical AI started to take off about 10 years ago; The popularity of doctor-AI "vs" competitions in 2018 has caused resentment in the industry; In 2019, medical AI was further approved by radiologists; In 2020, under the influence of the COVID-19 epidemic, the development speed of the industry was significantly accelerated, and the National Food and Drug Administration gave the green light to relevant products, and commercialization began.

这是一个有巨大想象空间的万亿级市场，覆盖医学影像、制药、健康管理、数据集成、预约问诊等等，AI在医疗领域能做的事情太多，上述历程在吕晨看来都是“符合医疗行业发展的规律”的。不过，远在大洋彼岸的“全球AI第一大厂”谷歌可能不会这么想：

This is a trillion-level market with huge room for imagination, covering medical imaging, pharmaceutical, health management, data integration, appointment and consultation, etc. AI can do too many things in the medical field. In Lu Chen's opinion, the above process is "in line with the law of the development of the medical industry". But across the pond, Google, the world's biggest AI company, might not think so:

8月底，谷歌旗下成立不到三年的健康部门Google Health宣布解散，其中专注于AI医疗影像的团队将被划入谷歌搜索和AI部门。而此前6个月，IBM脱手健康业务Watson Health，成为科技巨头折戟医疗AI领域的第一起标志性事件。

At the end of August, Google Health, the company's Health division, announced that it was disbanding after less than three years, with the team focused on AI medical imaging being moved into Google Search and AI. In the previous six months, IBM had sold off its Health business, Watson Health, in the first sign of a tech giant's demise in medical AI.

GoogleHealth部门负责人David Feinberg宣布离职。

David Feinberg, head of GoogleHealth, announced his resignation.David Feinberg, the head of GoogleHealth, has announced his departure.

在国内，阿里巴巴、、百度等互联网科技大厂们早早在医疗AI领域有所布局，商业化进程却不甚明朗。国家药监局在2020年初就开始陆续颁发一批医疗AI器械三类证，始终不见大厂踪影，直到今年9月，医疗才宣告成为国内“首个拿到三类证的互联网公司”。

In China, Alibaba, Baidu and other big Internet technology companies have made some layout in the medical AI field early, but the commercialization process is not very clear. At the beginning of 2020, the State Food and Drug Administration began to issue a batch of class III certificates for medical AI devices, and there was no trace of large manufacturers. It was not until September this year that medical care announced that it had become the "first Internet company to obtain class III certificates" in China.

如果说大厂尚能把投入于医疗的资源“抽回去”，那么对于国内一大批新兴的医疗AI公司来说，商业化的是与否、快与慢很可能直接决定它们的生死。

If it is said that the large enterprises can "draw back" the resources invested in medical treatment, then for a large number of emerging medical AI companies in China, the commercialization, speed and speed may directly determine their life and death.

医疗领域的投资者薛皓（化名）向“医学界智库”直言，目前医疗AI尚未探索出一条行之有效的、稳定的商业模式，他个人持相对保守态度，“这些公司被大型的设备厂商收购，可能是一个比较好的结局。”

Xue Hao (not his real name), an investor in the medical field, told the "medical think tank" that at present, medical AI has not yet explored an effective and stable business model. He personally holds a relatively conservative attitude. "It may be a better outcome for these companies to be acquired by large equipment manufacturers."

大厂之痛

The Pain of Big Factory

“谷歌医疗AI商业化提速！JeffDean亲自挖来大总管”“火力全开！Google进军健康领域！”34个月前，国内媒体对于Google Health成立的兴奋之辞，历历可数。

"The commercialization of Google medical AI is accelerating! Jeff Dean personally dug up the main manager" "fire is on! Google enters the health field!" Thirty four months ago, the excitement of the domestic media about the establishment of Google Health was numerous.

纳入“阿尔法狗”的研发公司Deep Mind的健康部门、成立印度首个AI实验室、在医学顶刊上发表论文……作为科技公司在医疗领域创业的代表产物，Google Health曾给许多医疗AI创业者以信心。此外，苹果、微软、Facebook等国际科技巨头也在几年间陆续布局医疗AI，侧重点各有不同。

Included in the health department of deep mind, the R & D company of "Alpha dog", established India's first AI laboratory, and published papers in the top medical journal... As a representative product of technology companies in the medical field, Google Health has given confidence to many medical AI entrepreneurs. In addition, apple, Microsoft, Facebook and other international technology giants have also successively deployed medical AI in recent years, with different emphasis.

转折似乎出现在2020年。谷歌在JAMA发文，称其AI糖尿病视网膜病变筛查产品准确率达到90%以上，欲在泰国落地。此后却意外发现当地拍摄质量不高、信号差，诊断准确率与时长都远低于预期。

The turning point seems to be in 2020. Google posted an article in JAMA, saying that the accuracy rate of its AI diabetes retinopathy screening product has reached more than 90%, and it intends to land in Thailand. After that, it was unexpectedly discovered that the local shooting quality was not high and the signal was poor, and the diagnostic accuracy and duration were far lower than expected.

“这往往是一些大IT公司容易犯的毛玻一个解决方案怎么能够适应不同的场景？在这个例子中，公司是否只是提供了一个帮助设备联网的小硬件？这足够了吗？怎么选取不同的网络运营商？数据标记质量是不是得分级？要提供一个医疗产品，而不只是一个算法。”专注于眼科辅助诊疗的AI公司致远慧图创始人、CEO孙宇辉告诉“医学界智库”。

"This is often a mistake that some large IT companies are prone to make. How can a solution adapt to different scenarios? In this example, does the company only provide a small hardware to help connect devices? Is this enough? How to select different network operators? Does the quality of data tags have to be graded? It is necessary to provide a medical product, not just an algorithm." Sun Yuhui, founder and CEO of Zhiyuan huitu, an AI company specializing in ophthalmic auxiliary diagnosis and treatment, told "medical think tank".

孙宇辉曾在英特尔任职15年，在渠道创新方面经验丰富，“我们要做的远远超越了要发一篇很好的论文，而是要去了解实际的应用场景，把算法和实践结合去解决具体问题。就像一辆漂亮的跑车放到乡村的土路上，可能跑都跑不起来，这不是车本身的问题，是你没有按实际的路况去设计你的发动机、底盘等等。”

Sun Yuhui has worked at Intel for 15 years and has rich experience in channel innovation, "What we need to do is far beyond issuing a good paper. Instead, we need to understand the actual application scenarios and combine algorithms with practice to solve specific problems. It's like a beautiful sports car placed on a dirt road in the country. It may not be able to run. It's not the problem of the car itself. It's your failure to design your engine and chassis according to the actual road conditions."

产品难以落地，商业化自然也无从谈起。如今，Watson Health、Google Health的结局似乎在某种程度上表明医疗AI“大厂时代”将暂告一段落，行业进入冷静期。同样地，今年8月，AI行业独角兽依图科技被传出售其医疗业务，商汤科技在IPO招股书上未将医疗业务列出，而这二者都被誉为国内的“AI四小龙”。

Products are difficult to land, and commercialization is naturally impossible. Now, the outcomes of Watson health and Google health seem to indicate to some extent that the "big factory era" of medical AI will come to an end and the industry will enter a cooling off period. Similarly, in August this year, the Unicorn Yitu technology in the AI industry was rumored to sell its medical business. Shangtang technology did not list its medical business in the IPO prospectus, and both of them were known as the "four little dragons of AI" in China.

多位医疗AI的从业者向“医学界智库”表示，科技公司跨行来做医疗，考验着他们对行业的理解力。“人工智能的本质就是让机器去复制人的经验，产生价值。企业本身不了解这个行业，经验没法复制，或者靠自己的想象去复制，这怎么行呢？”医疗人工智能公司左手医生CEO张超对“医学界智库”说。

A number of medical AI practitioners told the "medical think tank" that the cross-border medical work of technology companies tested their understanding of the industry. "The essence of artificial intelligence is to let machines copy people's experience and generate value. The enterprise itself does not know the industry, and experience cannot be copied, or it can be copied by its own imagination. How can this work?" Zhang Chao, CEO of the left-hand doctor of the medical artificial intelligence company, told the "think tank of the medical industry".

吕晨认为科技公司对于医疗行业存在“学习曲线”，从招募医学团队开始，到发展过程中的沟通与理解，都需要成本。何况，两个行业发展的逻辑与节奏大相径庭，“在互联网行业中，三年就应该出很大成绩了，当母公司发现在医疗领域里投入这么久都没有商业化，就会对行业看衰，甚至认为医疗AI是‘伪需求’，资源就会抽回去。”

Lu Chen believes that technology companies have a "learning curve" for the medical industry. From the recruitment of medical teams to the communication and understanding in the development process, costs are required. Moreover, the logic and rhythm of the development of the two industries are quite different. "In the Internet industry, great achievements should be made in three years. When the parent company finds that it has not commercialized its investment in the medical field for so long, it will look down on the industry and even think that medical AI is a 'false demand', and resources will be withdrawn."

“很多大厂很难躬身入局。”张超直言，“医疗本身是一个需要长线投入的行业，大厂可能在乎的是你每个小时能带给我多少价值，每一行代码能带给我多少收益。回过头来，你做一些底层价值驱动的事情，短期内没有回报，就比较难进行下去了。”

"It's very difficult for many big factories to get involved." Zhang Chao bluntly said, "the medical industry itself is an industry that needs long-term investment. What big companies may care about is how much value you can bring to me every hour and how much profit each line of code can bring to me. Looking back, if you do something driven by low-level value, it will be difficult to carry out it if there is no return in the short term."

回想三四年前，科技巨头们带着“颠覆医疗”的雄心进入行业，IBM还把“人类将会被取代”的口号叫得响亮，如今都成为它们“不懂医疗”的证据。“实际是外行，夸大了AI的作用，忽略了医疗的本质。”吕晨表示。

Three or four years ago, technology giants entered the industry with the ambition of "subverting medical care". IBM also loudly shouted the slogan of "human beings will be replaced". Now it has become evidence that they do not understand medical care. "In fact, they are laymen, exaggerating the role of AI and ignoring the essence of medical care." Lu Chen said.

要数据就不能收费？

Can't you charge for data?

产品研发怎么维系？

How to maintain product research and development?

回到国内，近年在资本助力下，医疗行业出现医准智能、深睿医疗、森亿智能、汇医慧影等独角兽。2020年9月至今，医疗AI公司医渡云、科亚医疗、鹰瞳Airdoc、推想医疗、数坤科技等陆续递交招股书。至此，国内医疗AI行至下半常

Back home, in recent years, with the help of capital, the medical industry has seen unicorns such as medical quasi intelligence, Shenrui medical, senyi intelligence and Huiyi Huiying. From September 2020 to now, medical AI companies such as Yidu cloud, Keya medical, Yingtong airdoc, putative medical, and Sukun technology have successively submitted prospectuses. So far, the domestic medical AI has reached the second half of normal

新纽科技是众多曾幻想用AI技术“颠覆”医疗行业的科技公司之一。不过，这家IT服务商在金融行业深耕20余年，在进入医疗行业前多方引入专家分析、做长线调研，最终得出了更务实的结论：利用AI做事之前，首先要解决医疗行业信息系统建设竖井化的问题，将割裂的“数据孤岛”缝合起来，建立关联。

Xinxin technology is one of many technology companies that once dreamed of using AI technology to "subvert" the medical industry. However, this it service provider has been deeply engaged in the financial industry for more than 20 years. Before entering the medical industry, it introduced many experts to analyze and conduct long-term research, and finally came to a more pragmatic conclusion: before using AI to do things, it is first necessary to solve the problem of building a shaft in the information system of the medical industry, sew up the separated "data islands" and establish connections.

“人工智能技术本质是基于概率的推论，关键在于数据的积累。它的突破是革命性的，因为计算机开始从人为制定的规则走出，到海量数据中自己寻找规则。”杭实资管研究员朱之颜如此解释AI与数据的关系。简单来说，没有大量的、高质量的数据，对AI应用研发来说就是无米之炊。

"The essence of artificial intelligence technology is inference based on probability, and the key lies in the accumulation of data. Its breakthrough is revolutionary, because computers begin to go out of the rules made by people and find rules themselves in massive data." Zhu Zhiyan, a researcher of Hang Shi asset management, explained the relationship between AI and data in this way. In short, without a large amount of high-quality data, AI application research and development is like making bricks without straw.

新纽科技副总裁丁耀欣表示，希望打造一个医疗大数据的开放平台，供医疗AI的同行所用，改善行业的发展条件。CT等高端医疗设备及软件系统自主研发公司开影医疗软件技术总监张也提示，无论是针对人体哪个部位的诊疗，在每一个方向上，医疗AI都需要独立做大量的前端工作，处理大量的相关数据，因此，企业在致力于研发应用时，还应在基础数据研究上下大工夫。

Ding Yaoxin, vice president of Xinxin technology, said that he hoped to create an open platform for medical big data for medical AI peers and improve the development conditions of the industry. Zhang, the technical director of Kaiying medical software, an independent R & D company of CT and other high-end medical equipment and software systems, also pointed out that medical AI needs to do a lot of front-end work independently in every direction, and process a large amount of relevant data. Therefore, when enterprises are committed to R & D and application, they should also make great efforts in basic data research.

“获取各种数据源、标定数据源，然后分类统计，在此基础上做临床应用的挖掘，这样效果会更好一些，否则一些企业烧了十几亿元，最后产出还是与投入不成比例。”张说。

"Obtain various data sources, calibrate data sources, and then classify statistics. On this basis, do clinical application mining. This will achieve better results. Otherwise, some enterprises will burn more than 1 billion yuan, and the final output will still be out of proportion to the input." Zhang said.

事实上，得不到数据支撑、研发进程缓慢，被视作是谷歌和IBM的相关业务最终被出售或解体的直接原因之一在信息孤岛化的基础上，国外对医疗数据的严格监管使它们只能拿到很少的数据，据丁耀欣观察，谷歌最大的一次数据采集，患者量也就在1万个左右，而在中国这个数字可能会是百万甚至千万级。

In fact, lack of data support and slow R & D process are regarded as one of the direct reasons why Google and IBM's related businesses were eventually sold or disintegrated. On the basis of information isolation, the strict supervision of medical data abroad makes them only get a small amount of data. According to Ding Yaoxin's observation, Google's largest data collection, the number of patients is about 10000, In China, the number may be one million or even tens of millions.

在保证数据安全、个人隐私的前提下，尽管中国的医疗数据比起国外已算相当开放，但其集成与挖掘历来是一件难事。丁耀欣对“医学界智库”表示，2010年以前，其他行业的信息系统建设基本都实现了平台化，唯独医疗行业直到如今，其影像处理、病例处理、计费系统等几个主要业务体系仍是割裂的。

On the premise of ensuring data security and personal privacy, although China's medical data is quite open compared with foreign countries, its integration and mining has always been a difficult task. Ding Yaoxin told the "medical think tank" that before 2010, the construction of information systems in other industries basically realized the platform, but until now, the medical industry still has several major business systems such as image processing, case processing and billing system.

这在很大程度上制约着AI在医疗中的应用与发展，且问题并不只存在于国内。2018年，IBM被曝其AI系统Watson Health因未处理好禁忌症数据的关联而给出多个错误的治疗意见，严重时可致患者死亡。

To a great extent, this restricts the application and development of AI in medical treatment, and the problem does not only exist in China. In 2018, IBM was exposed that its AI system Watson health failed to handle the association of contraindication data and gave several wrong treatment opinions, which could cause patient death in serious cases.

为了解决数据量与质的问题，北京清华长庚医院神经内科主任武剑向“医学界智库”透露，该院在2014年建立之初就有大数据中心，已设定好清华大学社区、天通苑社区模型，而后者即医院所处之地是亚洲最大的社区，临床数据量不会少，质量也比较高。这给神经内科与各方合作研发AI应用设备创立了非常好的条件。

In order to solve the problem of data quantity and quality, Wu Jian, director of the Department of Neurology of Beijing Tsinghua Changgeng hospital, disclosed to the "think tank of medical circles" that the hospital had a big data center at the beginning of its establishment in 2014, and had set up models of Tsinghua University community and Tiantongyuan community. The latter, that is, the hospital is located in the largest community in Asia, and the amount of clinical data is not small and the quality is relatively high. This has created very good conditions for the Department of neurology to cooperate with all parties to develop AI application equipment.

为了与这样的医院达成合作、拿到临床数据，大部分公司会在应用系统落地初期免费给医院相关科室使用，导致在医疗AI公司“井喷”的几年间，一些医院会有数套同类型的AI系统。谁敢踏出商业化的第一步？谁能率先结束医疗AI的免费时代？这考验着其产品的成熟度和创业者的决断力。

In order to reach cooperation with such hospitals and obtain clinical data, most companies will provide the application system to relevant departments of the hospital for free at the initial stage of implementation, resulting in several sets of AI systems of the same type in some hospitals during the "blowout" of medical AI companies. Who dares to take the first step of commercialization? Who can take the lead in ending the free era of medical AI? This tests the maturity of its products and the determination of entrepreneurs.

左手医生的大规模商业化自2020年开始，以医院用户为主。“觉得不收费的话，这个事情也不能往下走”，张超表示，唯有商业化，才能给到行业一个筛选机制和进步的动力，否则同类产品只能在医院里无序竞争。

The large-scale commercialization of left-handed doctors began in 2020, mainly by hospital users. "If there is no charge, this matter can not go down," Zhang Chao said. Only commercialization can give the industry a screening mechanism and a driving force for progress. Otherwise, similar products can only compete disorderly in hospitals.

而商业化的前提，是做出足够好的产品。丁耀欣指出，研发与盈利，二者实际上应该放在一起思考：怎样设计出能解决实际问题的产品？医院愿意为什么东西付费？当研发与产业模式的构建形成闭环，就能摆脱单纯依靠资本运作去维系产品研发的模式。否则在资本的豢养下，产品研发闭门造车，最终可能会离市场需求越来越远。

The premise of commercialization is to make enough good products. Ding Yaoxin pointed out that R & D and profitability should actually be considered together: how to design products that can solve practical problems? Why is the hospital willing to pay for things? When R & D and the construction of industrial model form a closed loop, we can get rid of the mode of relying solely on capital operation to maintain product R & D. Otherwise, with the support of capital, product research and development will be done behind closed doors, and may eventually be farther and farther away from the market demand.

“不要拿着锤子找钉子。”有不愿具名的创业者表示，“很多公司只专注于技术层面，例如数据量、算法、算力的提高，但医疗AI在临床上的核心价值，在于解决和改善现状的能力，比如提高服务可及性、落实指南和临床路径，提高通量、降低漏诊、误诊率等。”

"Don't look for nails with a hammer." An entrepreneur who did not want to be named said, "many companies only focus on the technical level, such as the improvement of data volume, algorithm and computing power. However, the core value of medical AI in clinic lies in the ability to solve and improve the current situation, such as improving service accessibility, implementing guidelines and clinical paths, improving throughput, reducing missed diagnosis and misdiagnosis rate."

混战、竞合，医疗AI商业化

Melee, Competition and Cooperation, Commercialization of Medical AI

的前途是什么？

What is the future of?

“经过近10年发展，绝大部分基于计算机视觉与自然语义处理的赛道已经趋于成熟，处于商业化探索阶段，但尚无一家企业通过医疗人工智能销售获得盈利，因而暂时没有企业进入规模商业化阶段。”动脉网日前发布的《2021医疗AI报告》提到，“这一阶段中，未收获器审中心审评审批认证的医疗人工智能企业忙于临床试验；已获得认证的企业专注于将产品进入区域医疗价格目录，并加速其市场占有；无需审评审批的医疗IT类企业则聚焦于产品销售与推广。”

"After nearly 10 years of development, most of the racetracks based on computer vision and natural semantic processing have become mature and are in the stage of commercialization exploration. However, no enterprise has made profits through medical artificial intelligence sales, so no enterprise has entered the stage of scale commercialization for the time being." According to the 2021 medical AI report released by arterial network a few days ago, "in this stage, the medical artificial intelligence enterprises that have not been reviewed and approved by the harvester review center are busy with clinical trials; the enterprises that have been certified focus on entering the regional medical price catalog and accelerating their market share; the medical IT enterprises that do not need review and approval focus on product sales and promotion."

互联网大厂和新兴的医疗AI公司们不是“唯二”的竞争主体，赛道上不乏其他身影。中山大学肿瘤防治中心自主研发的上消化道肿瘤人工智能系统在2018年启动；武剑表示，北京清华长庚医院神经内科与清华大学的电子工程系、计算机与智能研究所等单位合作，共同研发一些AI辅助诊疗系统，目前有产品一期、二期临床试验做得不错，顺利完成三期后即可获批上市。

Internet giants and emerging medical AI companies are not the only two competitors, and there are many other figures on the track. The artificial intelligence system for upper gastrointestinal cancer independently developed by the cancer prevention and treatment center of Sun Yat sen University was launched in 2018; Wu Jian said that the Department of Neurology of Beijing Tsinghua Changgeng hospital cooperated with the Department of electronic engineering and the Institute of computer and intelligence of Tsinghua University to jointly develop some AI assisted diagnosis and treatment systems. At present, the phase I and phase II clinical trials of the products have done well, and the products can be approved for listing after the successful completion of phase III.

医院和高校有学科优势、有更多官方机构的背书，是否会和医疗AI公司们形成竞争关系？“各有分工，相互配合，医疗机构和研究机构会去推进一些基础研究，企业负责把这些创新的技术应用到产品上，实现科研成果的转化。”孙宇辉说。薛皓则认为，政府行为会更偏科研、更聚焦，横向推广的工作还是得企业来完成。

Will hospitals and universities form a competitive relationship with medical AI companies because they have disciplinary advantages and have more endorsement from official institutions? "Each has its own division of labor and cooperates with each other. Medical institutions and research institutions will promote some basic research. Enterprises are responsible for applying these innovative technologies to products and realizing the transformation of scientific research results." Sun Yuhui said. Xue Hao believes that the government's behavior will be more scientific research and more focused, and the horizontal promotion work still needs to be completed by enterprises.

据动脉网统计，截至2021年8月16日，总计19款医疗人工智能器械获得国家药监局批准的医疗器械三类证。临床试验的审核标准已见雏形，但在武剑看来，无论是科研机构还是企业去研发医疗AI系统，都还缺乏一套健全的管理制度和流程，技术先行，行业还在野蛮生长阶段，“后续我们的管理体系跟上，就能反推技术的发展。”

According to the statistics of arterial network, as of August 16, 2021, a total of 19 medical artificial intelligence devices have obtained class III medical device certificates approved by the State Food and drug administration. The audit standards for clinical trials have taken shape, but in Wu Jian's view, no matter whether scientific research institutions or enterprises are developing medical AI systems, they still lack a set of sound management systems and processes. Technology is first, and the industry is still in the stage of savage growth. "If our management system keeps up with it, we can reverse the development of technology."

一些地方走在了前面。在安徽、江苏等一些地方，糖网AI筛查已经进入了收费目录，。“市场化脉络已经初步形成”，孙宇辉说。

Some places went ahead. In Anhui, Jiangsu and other places, sugar network AI screening has entered the charging directory,. "The context of marketization has taken initial shape," said Sun Yuhui.

比较之下，投资人薛皓不这么乐观。他感觉到，无论是患者付费还是医院买单，几条医疗AI的商业模式都让人不太满意，“患者信任度不高、付费意愿不强；跟大型医疗器械商合作，把设备卖给医院然后分成，医疗AI公司话语权又弱。”

In contrast, investor Xue Hao is not so optimistic. He felt that the business models of several medical AI are not very satisfactory, whether it is the patient's payment or the hospital's payment. "The patient's trust is not high and the willingness to pay is not strong; the medical AI company has a weak voice when it cooperates with large medical device manufacturers to sell the equipment to the hospital and divide it into shares."

薛皓认为，目前市场上对医疗AI公司估值过高，存在泡沫，“一家公司单独发展的路径我是不太看好，想不出来特别好的一个商业模式。”而医疗AI企业有创新技术，大型设备商有硬件能力和稳定的进院渠道，后者将前者收购可能是比较好的结局。

Xue Hao believes that the current market valuation of medical AI companies is too high and there is a foam. "I am not optimistic about the path of a company's independent development, and I can't think of a particularly good business model." While medical AI enterprises have innovative technologies, large equipment manufacturers have hardware capabilities and stable access to hospitals. The latter may have a better outcome by acquiring the former.

实际上，近年西门子医疗、飞利浦医疗、联影医疗、东软医疗等传统医疗器械公司都建立了AI医疗部门，大型设备商与医疗AI公司的合作已不鲜见。比如GE医疗在中国推出爱迪生数字医疗智能平台，与医准智能、数坤科技、安德医智等医疗AI公司签署战略合作备忘录，共同开发基于该平台的数字医疗应用。

In fact, in recent years, traditional medical device companies such as Siemens Medical, Philips Medical, Lianying medical and Neusoft medical have established AI medical departments, and cooperation between large equipment manufacturers and medical AI companies is not uncommon. For example, GE Medical launched the Edison Digital Medical Intelligence Platform in China, and signed a strategic cooperation memorandum with medical AI companies such as medical quasi intelligence, digital Kun technology and ande medical intelligence to jointly develop digital medical applications based on the platform.

医院、高校以及政府机构等也乐于和医疗AI公司合作。日前，中国医师协会相关专业委员会、中国科学院大学附属肿瘤医院联合医准智能举办了一场超声诊断人机体验赛，旨在以体验促进对人工智能的了解与发展；2018年，北京天坛医院与安德医智合作，共同成立神经疾病人工智能研究中心CHAIN，并于次年发布BioMind“天泽”脑血管病诊疗辅助决策系统。

Hospitals, universities and government agencies are also willing to cooperate with medical AI companies. A few days ago, the relevant professional committees of the Chinese Medical Association and the Cancer Hospital Affiliated to the University of Chinese Academy of Sciences jointly held a human-machine experience game for ultrasonic diagnosis with the aim of promoting the understanding and development of artificial intelligence through experience; In 2018, Beijing Tiantan Hospital cooperated with ande medical intelligence to jointly establish the neural disease artificial intelligence research center chain, and released the biomind "Tianze" cerebrovascular disease diagnosis and treatment assistant decision system in the following year.

另外，2021年4月，在上海市静安区和新疆巴楚县卫健委的牵头下，一批鹰瞳Airdoc的眼底照相机和软件在当地实现应用，辅助当地村医更好地掌握和使用眼底诊疗技术。

In addition, in April 2021, under the leadership of the Health Committee of Jing'an District of Shanghai and Bachu County of Xinjiang, a batch of fundus cameras and software of Yingtong airdoc were applied locally to assist local village doctors to better master and use fundus diagnosis and treatment technology.

“越是基层的医院，对医疗AI辅助诊断工具应用的需求其实是更高的。”吕晨表示，而这几乎已成为行业共识，许多公司从三甲医院起步，如今在二三线城市的基层医疗机构已有所布局。

"The more grass-roots hospitals, the higher the demand for medical AI assisted diagnostic tools." Lu Chen said that this has almost become the consensus of the industry. Many companies started from the third-class hospitals, and now the basic medical institutions in the second and third tier cities have made a layout.

无论医疗AI商业化的前途如何，支付方是谁，独立发展或是将被收购，正如武剑所说，医疗数字化是大趋势，而AI是其中不可或缺的角色。“这最终是一个机器语言如何融入社会、被人们所信任的问题。”张说。

No matter what the future of medical AI commercialization is, who the payer is, whether it develops independently or will be acquired, just as Wu Jian said, medical digitization is the general trend, and AI is an indispensable role in it. "This is ultimately a question of how machine language can be integrated into society and trusted by people." Zhang said.

参考资料：

References:

[1]“医疗+AI”行业研究_详细解读_最新资讯_热点事件_36氪 https://36kr.com/p/1189593572034818

[1] "Medical + AI" Industry Research _ Detailed Interpretation _ Latest Information _ Hot Events _ 36 Krypton https://36kr.com/p/1189593572034818

[2]《2021医疗AI报告》发布，行至IPO的医疗AI，商业化走到哪一步？_详细解读_最新资讯_热点事件_36氪https://36kr.com/p/1416072801834630?channel=wechat

[2] "2021 Medical AI Report" is released. Where is the commercialization of medical AI going to IPO? _ Detailed Interpretation _ Latest Information _ Hot Events _ 36 Krypton https://36kr.com/p/1416072801834630? Channel=WeChat

来源：医学界智库

Source: Medical Think Tank

责编：汪航

Editor: Wang Hang

校对：臧恒佳

Proofreading: Zang Hengjia

制版：舒茜

Plate-making: Shu Qian

赞助本站

Sponsor this site