Image source @ Visual China
Wen 丨 vb arterial network
Finding the right word to describe 2021 for medical AI is not simple.
The distant IBM Watson search for the owner of the gold was a slightly gloomy beginning, but before the sigh was over, Koya Medical handed over its prospectus. In the months that followed, as reviews and approvals, financing mergers and acquisitions, and policy support came and went, AI seemed to be increasingly recognized by hospitals, investors, and governments.
Imaging and informatization are the two fastest tracks in AI. In 2021, the disclosure of prospectuses, bidding and contract transactions gave us key data to judge the future development of AI, with images moving from "0" to "1" and informatization from "1" to "10".
Now it is a new year, the intersection of new and old, arterial network found nearly ten industry people to communicate, is to review 2021, but also to look forward to 2022.
The era of burning money is over, and medical AI looks at revenue
On January 15, 2021, Yidu Technology rang the bell on the Hong Kong Stock Exchange, followed by Zero Kr Technology, and submitted a prospectus to the NASDAQ in June. The successive rush of the two leading enterprises marks the leap of a new generation of medical IT companies from the primary market to the secondary market, and also tells several models of medical AI to commercialization.
Coincidentally, image AI also ushered in a wave of listing in 2021, Keya Medical, Eagle Pupil Technology, Speculative Medical, and Shukun Technology have submitted prospectuses, and several times the growth of revenue in 2020 has become the norm for image AI.
Corresponding to this is the weakness of the primary market, eggshell research institute data show that in 2022, the total number of financing events in the primary market of image AI is 28, and the number of informatization AI markets is 25, and the number of financing is relatively small, and most of the funds also flow to enterprises above the C round.
From a data level, the IPO is the well-deserved "key" to medical AI in 2021, but in the final sense, can the IPO really bear this responsibility?
In the past few years, in the research of medical AI by the Eggshell Research Institute, we have paid special attention to investment and financing data. After all, since the rise of medical artificial intelligence in 2015, medical artificial intelligence companies have always been running on the road of product research and development and hospital development, and most of the enterprises have less than ten million revenues. At this stage, "financing" is a more intuitive dimension to describe the popularity of the AI industry, which can measure the speed of production lines and hospital cooperation of AI companies to a certain extent. But today, "profitability" has become the key to the head AI consideration problem.
What determines profitability? In the short term, it is product quality and real demand Product quality and real demand.
"The essence of the lack of commercialization is the lack of product quality of enterprises, and doctors do not recognize this thing." A practitioner who has been in the field of AI for many years told Arterial Network, "Some products are not of good quality, and some products cannot meet clinical needs." ”
"The screening of demand has screened out a large number of enterprises. Up to now, the remaining enterprises are almost all in accordance with the clinical needs of doctors to carry out sequential research and development. After all, the AI here is 'medical + AI', 'not AI + medical'. ”
Several AI companies that submitted prospectuses clearly met the requirements. Since the registration approval ran through, the annual revenue of Eagle Pupil Technology in 2020 has increased by 50% compared with last year; it is estimated that the annual growth of medical care has increased by 318%; and Shukun Technology has even doubled 32 times, and the half-year revenue has exceeded 50 million. Although other companies did not have clear data disclosures, they all said that there was a substantial increase.
However, much of the current profitability of medical AI comes from the successful commercialization of accumulated products over the past few years. Dividends will always fade one day, and at that time, what development will medical AI companies rely on?
The Long-TermIsm of Medical AI: A Breakthrough Exploration of Closed Loop, Cross-Border and "Stagnation"
Solution integrity is key to building the long-term reach of healthcare AI and expanding market reach. Judging from the AI products currently approved by NMPA, AI can only be solved in one or a few specific problems, but for imaging doctors, chronic lung resistance, hydropulmonary obstruction, fractures outside the lungs, heart-related diseases, and all diseases seen by imaging doctors when reading the film need to be reported. Therefore, AI with a single point function may not necessarily reduce the workload of doctors, but may be counterproductive. In other words, if AI companies want to survive in the long run, they must start a comprehensive upgrade from point to point.
Most of the head companies began to build their own AI closed loop 2-3 years ago.
To give a few examples, starting from the FFR with strong potential demand for AI, Keya Medical has gradually expanded a single product to the whole process of coronary heart disease diagnosis and treatment, and has forced balloons and other pipelines that are biased towards cardiac treatment, forming a closed loop of business in heart disease diagnosis and treatment hospitals. Recognized by tertiary hospitals and complete sets of technologies, Keya Medical may next exert efforts in the cardiovascular diagnosis and treatment line, precipitating the upper layer of technology downwards, and maximizing its contact with more customers.
Zhiyuan Huitu, Eagle Pupil Technology, and Voxel Technology have all carried out in-depth layout in ophthalmology, trying to shift from ophthalmology to diabetes management. Ghrepower Huitu and Voxel Technology have made fruitful attempts overseas, and Ghrepower Huitu has cooperated with the Optometry Center to equip its fundus equipment with multi-disease fundus imaging assisted diagnosis software. With the support of AI, patients can gain insight into eye health through a fundus photography, which can be used to supplement myopia examinations or create independent eye examinations to help residents find eye lesions in a more timely and effective manner and intervene in eye diseases as soon as possible.
The closed loop of Huiyi Huiying and Shenrui Medical is based on the collaborative closed loop of medical IT infrastructure and medical technology and clinical departments, emphasizing the closed loop of data and application flow, and betting on the process digitization of future clinical applications. Huiyi Huiying's product system includes a big data cloud platform, a data middle platform, and AI-assisted diagnostic tools such as aorta, bone, and breast, and the entire closed loop has begun to run. Shenrui is very strong in assisted diagnosis, winning a number of three-class certificates, and a large number of medical AI-related papers have landed in authoritative journals. After the acquisition of Yitu Medical in July last year, the relevant layout of Shenrui Medical in medical IT has been very mature.
In addition to the above cases, enterprises such as Zero Kr Technology, Keya Medical, and Medical Quasi Intelligence have also built their own closed loops. In this way, in the past, the way to label enterprises with "AI + image", "AI + informatization" and "AI + health management" has become somewhat outdated. Most of the companies at the head have the capabilities of the above multiple tracks, and the competition for medical AI is converging at the same time as it is also shifting to a new medical track.
There are pros and cons to this trend. On the one hand, the greater market choice of AI means more efficient and clinical solutions, but on the other hand, most of the current "surfaces" originate from the "points" of the past, but now, we rarely see companies involved in new "points".
Sun Yuhui, CEO of Zhiyuan Huitu, explained this in an interview: "The research and development of every medical AI product must go through the process from data collection, algorithm research and development, clinical verification, and registration and declaration. Between 2019 and 2020, the products accumulated in the medical AI industry in the past were launched in a concentrated manner and successively obtained certificates, so it gave people the feeling of new products.
However, from the perspective of the company's R & D strategy, the products developed at the beginning are relatively high-quality in terms of technical accessibility, and the follow-up new product research and development is more difficult and requires more time, which is a normal phenomenon. In addition, there are not many segments that are still blank in the field of medical AI, which is also one of the reasons for this phenomenon. ”
Some business leaders believe that the lack of new "points" is only temporary. Sales are the most important thing for years to come. "In the past few years, tens of billions of funds have been invested in medical AI, and the seeds sown in the past have finally ushered in the harvest season with the approval of approval, and the priority is naturally to take advantage of the autumn wind to catch the harvest." Therefore, at this stage, how much the IPO has raised and how much the market value is not so important, the diversity of products is not so important, and what is important is to eat the market share quickly with the three types of certificates and the 'national examination' of the tertiary hospital. After all, with profitability, the company can continue to develop, it is possible to open up new points. ”
Imagine the next five years
Between 2020 and 2021, medical AI has achieved "from beginner to proficient" in approval and approval. In 24 months, NMPA issued a total of 23 three-class certificates, and medical AI software in various parts passed the approval and review, and even cdSS with NLP as the core, Senyi Intelligent also won the second-class certificate in October 2021.
However, the passage of registration access only means that AI can enter the medical market and sell outside the hospital system. In order to implant AI closed-loop solutions into hospitals and find more users to pay for them, medical AI companies must pass through the next price access and medical insurance access.
Three admissions facing medical AI
Keya Medical took the lead in opening the application for price access after its CT-FFR deep pulse score passed the NMPA, tried to provide medical AI-related services directly to patients through the hospital, and won the deep pulse score in Hebei, Anhui, Shandong, Jiangsu and other provinces and cities within one year, but from the perspective of the whole industry, the price access of medical AI is still in its infancy, and there are less than 10 approved products in the country.
More far-reaching health insurance access is the key to achieving profitability of medical AI companies, and it is also a future that is more difficult for them to reach.
Looking overseas, the United States is a country with a relatively successful commercialization of medical AI, and many AI products have passed medical insurance access. In large part, the cost of equipment and diagnosis for radiological examinations in the United States is separate, and labor costs are particularly high. In contrast, in China, the cost of radiological examination is very low, will not be calculated separately, and most of it is equipment costs. This is why at this stage, the United States AI can be independently charged, independent reimbursement of the reason, in contrast, there are still certain examples in China.
Example of a partially insured AI product in the United States (Source: Medicare.gov)
Therefore, although revenue is important, it may take years for medical AI to complete the two approvals and realize every idea in the medical AI production line.
In this regard, Sun Yuhui, CEO of Zhiyuan Huitu, said: "The state very much encourages the application of artificial intelligence in the medical industry, and we believe that the next 1-2 years is the peak period of price access, but on medical insurance access, it also needs the support and promotion of the state at the policy level." ”
HuiYi Huiying also expressed a similar view: "We have been trying to promote price access, but there are several prerequisites for this matter. The first is that enterprises need to have a certificate; the second is that the number and time of applying for prices in each province each year are different; the third is that many provinces still need patients to buy at their own expense after applying for prices, and they have to run for at least 1 to 2 years to convert to medical insurance charges. So to accomplish these three things, companies need a 4 to 5 years of time. ”
Fortunately, the promotion of medical AI by policies has been deepening. From "Healthy China 2025", single-disease quality control, to DRG, "high-quality development of public hospitals, the status of medical AI in hospitals continues to increase, and even affect the rating of smart hospitals in public hospitals, performance appraisal of tertiary public hospitals."
Individual companies such as Huiyi Huiying and Shukun Technology are expected to turn losses into profits in the past two years, but it may take five years for the medical AI industry to become generally profitable.
Compared with the last five years, there are now plans, incomes, and support from hospitals, governments, and investors. The commercial exploration of AI has reached the halfway point, from being alone to passing side by side, the dawn has been slightly enlightened.