China's First Batch of AI Medical Devices Enters Class III Certification Review, Opening the Door to Commercialization

The first batch of Class III certificates for domestically developed medical AI products in China has entered the approval stage.

Recently, Ji Xin, CEO of Beijing Shanggong Medical Technology Co., Ltd. (SHANGGONG), a medical AI company, told Badian Health News that its artificial intelligence-based automated screening product for diabetic retinopathy (hereinafter referred to as “DR”) has passed the testing conducted by the National Institutes for Food and Drug Control (hereinafter referred to as “NIFDC”) and entered the clinical trial phase. This means that the long-awaited Class IIIMedical DevicesCertificate approval process, officially commenced.

Similar to SHANGGONG, products from multiple companies have recently passed testing and entered the approval process.

These are the first domestically developed Chinese medical AI products to enter the approval process.

Current medical AI remains in a phase of high investment. Zhong Xin, founder of Tuma Shenwei, previously told Ba Dian Jian Wen that the cost of developing a medical AI product is at least several million yuan, with a development cycle of 6 to 12 months.

However, in the face of this emerging innovation, relevant authorities require time to establish standards and detailed regulations for qualification accreditation, making it difficult to determine product pricing and fee structures.

What has given the industry hope is that relevant government authorities have been actively promoting the approval process for related products. In December 2018, the National Medical Products Administration (NMPA) held a special public-interest training session in Beijing, introducing the approval procedures and key points for Class III AI medical devices, as well as recommendations for clinical trial requirements. Subsequently, on February 1 of this year, the Center for Medical Device Evaluation of the NMPA released the “Key Points for the Review of Medical Device Software Using Deep Learning-Assisted Decision-Making (Draft for Comments)” and solicited public feedback. This signals that the review standards for Class III AI medical devices are close to implementation, and the policy bottlenecks hindering industrial development are expected to be broken.

It takes nearly 40 months to obtain approval for a Class III medical device certificate.

“Currently, our paying users account for less than 10% of the total user base, and AI products in many hospitals are still in a ‘trial’ phase,” Ji Xin told Ba Dian Jian Wen. He noted that medical AI companies currently on the market all face the same challenge: it is difficult to charge for their products.

According to the new version of the "Medical Device Classification Catalog" released in August 2017, if AI diagnostic software provides diagnostic recommendations through algorithms and only has auxiliary diagnostic functions without directly giving diagnostic conclusions, it should be declared as a Class II medical device; if it automatically identifies lesion sites and provides clear diagnostic prompts, its risk level is relatively higher and it must be managed as a Class III medical device.

Class II and Class III medical devices differ in terms of registration and market access. According to relevant regulations, obtaining a Class III medical device certificate requires clinical trials. In contrast, Class II medical devices are subject to a clinical trial exemption list. Furthermore, Class II medical devices only need to be registered with the provincial drug administration, whereas Class III devices require product registration with the National Medical Products Administration (NMPA).

Due to the aforementioned factors, none of the more than 140 medical AI companies in China have yet obtained a Class III medical device registration certificate.

Compared with 2018, significant breakthroughs have been achieved in the registration of Class III medical AI devices.

Generally, the commercialization of Class III medical devices involves three major steps prior to market launch: testing by the National Institutes for Food and Drug Control (NIFDC), registration and approval by the drug regulatory authorities, and pricing determination by the National Healthcare Security Administration. These three steps typically take nearly four years to complete.

In the first half of 2018, the National Institutes for Food and Drug Control (NIFDC) was still constructing standard databases for diabetic retinopathy and pulmonary nodules. A year later, some products have passed NIFDC testing, securing a critical stepping stone toward registration and approval by drug regulatory authorities.

Ji Xin told “Ba Dian Jian Wen” (8 o’clock Health News), “As long as the product meets standards, the software system is sound, and the database verification passes, the entire testing process can be completed in approximately three months.” He stated that testing by the National Institutes for Food and Drug Control (NIFDC) is the fastest among the three major steps; subsequent registration approval theoretically requires 36 months, while pricing approval by the National Healthcare Security Administration takes roughly six months.

It is worth noting that the government has offered many preferential policies for the approval and declaration of innovative products, such as establishing a “green channel” for submissions. Ji Xin stated, “Although it is uncertain how much time the green channel can save in the approval and registration process, it is certainly a positive signal.” It is understood that the non-invasive CT-FFR DeepVessel FFR product developed by Keya Medical has entered the “Innovative Medical Device Green Channel,” becoming the first Class III artificial intelligence medical imaging product to enter this green channel.

There is a market both within and outside hospitals.

Commercialization is by no means an overnight achievement. Ji Xin told Health News at Eight that although it was challenging to achieve commercial implementation of medical AI in the early stages, market education must be conducted in advance, and products should undergo trial operations in healthcare institutions to facilitate integration with these facilities. “During our promotional efforts over the first three years, we carried out extensive education programs for physicians and nurses to raise hospitals’ awareness of this disease management task,” he said.

From Ji Xin’s perspective, the three years of market education have yielded tangible results. He revealed that SHANGGONG’s product, “Huiyan Tangwang” (Smart Eye for Diabetic Retinopathy), has been deployed in over 400 hospitals across 28 provinces in China, including more than 200 tertiary hospitals, all of which have introduced fundus cameras into their endocrinology departments. Currently, the product is used by 1,000 to 2,000 patients per day on average.

According to market observations by Ba Dian Jian Wen, the market prospects for medical AI products are in fact extremely broad, with current applications primarily focused on two major sectors: in-hospital settings and out-of-hospital entities such as health examination centers and health service institutions.

Taking AI-based fundus screening as an example, the medical AI company Airdoc has entered not only the in-hospital market but also the out-of-hospital market, which is primarily dominated by health checkup institutions and optical retail chains.

Airdoc told Ba Dian Jian Wen that its medical AI product for fundus screening is currently collaborating with iKang Healthcare Group and Baodao Optical chain stores. Incorporating Airdoc’s AI-based retinal health risk assessment product into iKang’s health checkup packages helps users detect health risks at an early stage. In the collaboration with Baodao Optical, considering that myopia can easily lead to retinal complications and seriously affect individuals’ visual health, Airdoc integrates its AI solutions to help users rule out retinal impacts on vision, thereby enabling more accurate eyeglass prescriptions.

It is reported that Meinian Onehealth is applying AI technology to lung cancer screening, gastric cancer screening, ultrasound imaging of the breast and thyroid, and diabetic retinopathy screening in ophthalmology.

In November 2018, Zhang Ligang, Chairman and CEO of iKang Guobin, also stated at the press conference that AI can provide physicians with auxiliary diagnosis and data analysis, thereby improving their work efficiency, while also addressing the shortage of medical specialists in areas with weak healthcare infrastructure. Taking ophthalmology as an example, Mr. Zhang noted that when ophthalmologists lack sufficient clinical experience, there is a risk of misdiagnosis and missed diagnoses. The newly introduced Airdoc retinal image analysis system can help physicians rapidly identify fundus lesions and screen for ocular diseases.

Ji Xin told Baodian Jianwen that SHANGGONG has already launched pilot commercial collaborations with New China Life Insurance and selected health checkup institutions. Following the success of these pilots, large-scale promotion and expansion will be undertaken. In Ji Xin’s view, despite the many advantages of the out-of-hospital market, the more challenging in-hospital market must not be abandoned. He noted, “The sustainability of the out-of-hospital market is relatively weak. For instance, health checkup institutions, driven by profit motives, may be affected by unhealthy competition, which could undermine the sustained profitability of their products.”

Furthermore, Ji Xin insists that from the perspective of public credibility, patients have a higher level of trust in hospitals and doctors compared to health checkup institutions. Therefore, products developed within hospitals carry greater authority and are more conducive to the long-term, stable development of the product.

Source: HealthInsight Author: Zheng Qi