BEIJING – With home-grown artificial intelligence (AI) medical devices under priority review, mainland China is quickly putting together a regulatory framework to more rapidly tap into the power of AI to develop devices and drugs.

China has long realized the value of AI to supercharge the development of the country and multiple industries, so it's moving fast to apply the emerging technology to health care.

The Chinese government has included AI in its national development strategy. In 2017, the country officially released the Plan for the Development of New Generation Artificial Intelligence. It outlines a three-step strategy aimed at helping China achieve one of the world's most advanced levels of AI by 2020 and become the global center for AI technologies by 2030.

The ambitious plan affects all sectors, including health care. It proposes promoting AI as "a new model and a new form" and includes plans to develop surgical robotics, wearable and biocompatible monitoring systems, AI imaging and more. It also mentions the use of AI in drug development and drug regulation.

"Judging from the current policies, the government will promote AI technologies mainly in diagnosis (imaging for early screening), treatment (surgical robotics and telemedicine), hospital management (smart hospitals), health care services (health management), public health as well as new drug development," Zheng Liang, professor of China Institute for Science and Technology Policy at Tsinghua University, told BioWorld.

"As far as I know, the Chinese government has yet to introduce any AI-specific policies or guiding opinions in the health care sector. With that said, a three-year plan has made AI medical imaging a key development area," he added.

That plan was put forward five months after the AI development blueprint of 2017. It considers the use of AI in medical imaging and the need to "standarize and regulate data collection for medical imaging" and "support the R&D of AI imaging for brain, lungs, eyes, bones, cardiovascular diseases."

As Liang pointed out, medical AI is only incorporated in broader policies, such as the Internet Plus Healthcare strategy announced in May 2018 that promotes online diagnostic services using AI-aided programs.

"Overall, the Chinese government is highly supportive of, even actively pushing, the application of AI technologies in the health care space," Caroline Meinhardt, a Beijing-based health care consultant at Apco Worldwide, told BioWorld. "The general hope is that advanced medical AI can help tackle some of the pressing health care challenges to deal with resource shortages, high misdiagnosis rates and inaccuracies in early disease prediction.

"Over the past year or so, we've increasingly seen health care regulations and plans by the National Health Commission and other industry regulators explicitly encourage the adoption of medical AI in hospitals," Meinhardt added.

Regulations in the pipeline

Health care regulators responded quickly by introducing this year a regulatory framework for AI-powered products for market approval.

In June, the Technical Review Guideline on AI-Aided Software was released. It included a list of criteria for registration with a focus on the algorithms and data used. The aim of the guideline is to help manufacturers develop and register AI software.

The use of AI-aided software as a medical device is completely new to the China industry, Grace Fu, CEO of consulting firm China Med Device, told BioWorld MedTech.

"It was after April 2018 when the U.S FDA approved the first AI-based device [IDx-DR by IDx LLC] to detect certain diabetic retinopathy, that AI software – especially diagnostics based on image recognition – has drawn attention of the Chinese public," said Fu. (See BioWorld MedTech, April 12, 2018.)

The guideline requires regulators to look at the intended use of software and any core functions to ensure safety and effectiveness. It puts emphasis on how the software controls data quality and enables generalization that allows an algorithm to be effective across a range of inputs and applications, as well as what clinical risks may arise.

Meanwhile, clinical institutions must meet the requirements for collecting, pre-processing and labeling data and building datasets. For example, data masking and cleansing are required.

For imported software, developers must provide ethnic and epidemiological differences of the data. If they cannot ensure the safety and effectiveness of software, they must conduct clinical trials in the country to prove so.

As for algorithm design, regulators will look at algorithm selection, training and performance assessment as well as cybersecurity to ensure its interpretability.

"I'm sure we'll see a steady stream of policies in the months to follow that will tackle different aspects of medical AI ranging from registration to clinical trials," said Apco's Meinhardt. "But naturally it will take some time for China to build a robust regulatory framework for these emerging technologies."

"For now, regulators are taking a relatively conservative approach, especially in their classification of any devices that incorporate AI software with explicit diagnosis functions as Class III devices," she added.

The Medical Device Classification Catalog released in August 2017 was the first to include AI diagnostics. Algorithm-driven diagnostic products that do not directly provide diagnosis, only advice, are classified as Class II devices. Those that enable automatic identification of lesion site and give clear diagnosis are classified as Class III devices due to the higher risks they pose.

In March an AI Medical Device Standardization Unit was set up and given responsibility for the standardization of terminology, classification, the quality management of datasets, basic common technology, quality management systems, product evaluation processes and special evaluation methods for AI medical devices.

Approvals to come

China has yet to approve any AI-backed Class III medical devices that are limited to hospital use and not commercialized for wider application. This points to a bottleneck in the development of AI in the medical sector.

Given the increasing importance of AI to health care, there is a need to speed up approvals of AI-based devices and help developers turn a profit. Currently, the first batch of AI-powered medical devices, which are mostly diagnostic products, are going through the priority review channel created for innovative devices.

In June, Shenzhen Sibright Co. Ltd.'s AI-enabled diabetic retinopathy screening software was granted priority review. The software collects fundus images and uses deep learning technology to screen these images automatically to help doctors diagnose diabetic retinopathy more efficiently.

Quite similar to the FDA-approved IDx-DR, Sibright's software is widely expected to be the first AI-based medical product approved in China.

Another AI-based device under the priority pathway is Beijing Curacloud Technology Co. Ltd.'s Deepvessel FFR platform. It uses deep learning technology to calculate the fractional flow reserve value out of computed tomography angiography images in five minutes, allowing efficient assessment of the functional significance of coronary stenosis.

The platform has already obtained a CE mark in Europe and it is under regulatory review as a Class III device in China.

"We expect the marketing approval to come this September," Deyin Wang, Curacloud's human resources vice president, told BioWorld.

While China moves closer to greenlighting its first homegrown AI-based medical device, the country still has to polish its regulatory framework for medical AI. So far, much of the focus is limited to diagnostic devices.

"In the future, there will be more specific policies for areas including health care based on clearer regulations on some common and fundamental issues regarding AI development such as data ownership, data privacy, cybersecurity and ethics," Tsinghua's Liang said. "The government will promote AI development with moderate regulation."