In the “Research Report on China’s medical AI market in 2019” released by yiou think tank, it mainly analyzes the development level and market application of AI under five application scenarios. So, what kind of test does the enterprise need to go through in the process of commercialization? It should be noted that at present, the overall ecology of medical AI can be described by “hierarchy”. This paper will discuss AI enterprises, which are middle-level technology providers.
From going out of the laboratory to commercialization, enterprises are not only faced with the change of the scene. In addition to polishing products and landing promotion, it is also necessary to cross a higher entry threshold, review and approval requirements, hit hard demand, get trust, generate flow, achieve profits, and form a stable and sustainable business model. In general, the commercialization process includes six links: data acquisition, technical verification, registration and approval, market access, pricing and business model verification. Registration examination and approval is a problem that enterprises pay more attention to at this stage, and there are many bright spots in this work in China.
First of all, China learns from the advanced international experience, and refers to the practice of the United States in the aspects of rule making, process setting and indicator setting. In July 2017, FDA released the digital health innovation action plan to reduce the registration and approval threshold, aiming at accelerating the landing process of medical AI products. By the end of 2018, FDA had approved 17 medical AI products (or services). Especially after the appearance of IDX Dr, the first medical AI diagnostic product approved by FDA, its registration and approval process has become a typical learning reference case of medical AI product registration and approval in China.
Secondly, we should encourage innovation on the premise of safety. The basic concept of innovative medical device review and approval in China is to ensure the safety and effectiveness of medical devices, encourage the research and innovation of medical devices, set up a green channel, and promote the promotion and application of new technologies of medical devices. The concept of the United States is to encourage enterprises to innovate, simplify regulatory procedures, adopt scientific evaluation methods, and speed up the provision of novel, safe and effective medical device products to patients, with minimum workload and critical path as the basic principles. In contrast, China strictly controls the “safety” gate, tightens the entry caliber of enterprises, encourages innovation, and supports the application of high-quality products as soon as possible.
At present, China has published the classification catalogue, approval process and key documents, and carried out the public welfare training of registration and approval, and the related work is steadily advancing. This process takes a long time and is very strict. However, enterprises generally believe that the approval process of the three types of certificates is proceeding at a normal speed, and the state’s response to the registration approval is relatively rapid.
Cross the threshold of registration and approval, only to get a “ticket.”. The challenge of commercialization of medical AI is more than that. Industry standards need to be improved, black box problem of medical scene, unknown bargaining power, high algorithm risk, high complexity of stakeholders will be the test in the process of commercialization. In the medical scenario, there is a strong demand for data desensitization and high requirements for data operation compliance. The temporary absence of standards will undoubtedly affect the application and promotion in this field, so we need to establish data standards. In addition, the technical methods and test indicators of medical artificial intelligence products also need to be introduced.
In most industries, it is not necessary to understand how the system works. However, in the medical field, product users (doctors and other professionals) need to understand the mechanism of diagnosis or decision-making recommendations of the products currently used, so as to confirm whether the algorithm is really reasonable or “crooked”, and need to ensure that it will not lead to major medical accidents because of its defects.
At present, medical artificial intelligence products are not included in hospital fees, so it is difficult for enterprises to realize commercial operation. However, at present, enterprises in various segments have begun to seize the market, and absolute market share holders in some segments have emerged. When the market access mechanism is complete in the future, the first mover advantage will help enterprises to have differentiated bargaining power.
The surging wave of AI is much faster than the process of policy, regulation and access. Therefore, before medical AI becomes the norm, there will inevitably be a period of “value exploration”. With the development of products from imagination to R & D, then to clinical, embedded in today’s medical system, and commercialization, the situation faced by medical AI is becoming more and more complex, and more and more from the lofty fantasy to reality. In this process, technology, system, standard and standard should be in place; For enterprises, the idea of “flying horse and flying horse” will gradually become true in technical verification and commercial verification. Only those enterprises with a lot of accumulation and excellent level can survive.