Post-Pandemic Medical Imaging AI: Unlocking the Vast Potential of Grassroots Markets and Global Expansion

During the prevention and control of the COVID-19 pandemic, AI applications have been widely adopted, ranging from self-assessment tools for residents to AI-assisted medical imaging screening and big data monitoring by Centers for Disease Control and Prevention. The presence of AI is ubiquitous. AI companies, along with their technologies and products, have also been rigorously tested and refined during the pandemic.

What new development trends have emerged in the AI sector in the wake of the pandemic? With the severe epidemic situation overseas, can Chinese AI companies seize the opportunity to expand into international markets? Conversely, what impact will the overseas epidemic have on these enterprises?

In light of this, VCBeat recently hosted the “Imaging Special” session of its 2020 Tech-Enabled Epidemic Prevention and Control Online Forum. Focusing on imaging trends and the value of AI in the post-epidemic era, the panel featured moderator Wang Haijiao, Executive Partner at GTJA Investment; Zhan Yiqiang, Chief Operating Officer of United Imaging Intelligence; Yuan Jianhua, Director of the Clinical Application Center at MinFound Medical Systems; and Chen Kuan, Founder and CEO of Infervision. Each shared their insights, delivering an engaging audiovisual experience for the audience. VCBeat has compiled their key viewpoints below.

Although medical AI is developing rapidly in China, only a small fraction of companies have successfully established viable business models. This is particularly true for the startups that have emerged in large numbers in recent years, for whom the prospects for concrete implementation of their business models remain unclear.

For startups, “survival” is the top priority. Achieving commercialization remains a widespread challenge. To continue capturing a share of this market, companies must achieve breakthroughs and drive innovation.

Currently, United Imaging Intelligence has generated revenue by providing software or integrated hardware-software solutions. Zhan Yiqiang compares the company’s business model to an amphibian, empowering both physicians and medical equipment by “walking on two legs.” He noted that there is a misconception in the market regarding the implementation of AI: namely, that empowering equipment must be tied to hardware, while empowering physicians must take the form of software. “This is not the case. When companies truly bring their products to market, only through ‘integration of software and hardware’ can they achieve a ‘1+1>2’ effect.”

In the view of Chen Kuan, CEO of Infervision, as an innovative medical technology, the implementation of medical AI requires undergoing a high-barrier validation cycle to verify its effectiveness and safety. Once a product demonstrates an irreplaceable role in clinical trials, its commercial value will emerge, and its business model will naturally follow.

Currently, AI-assisted imaging diagnostics cannot yet be priced. This is an issue that medical AI companies will inevitably have to face in the future. So, how can artificial intelligence enterprises “survive” when the revenue model remains unclear?

Wang Haijiao, Executive Partner at GTJA Investment, stated that establishing independent reimbursement for imaging AI will drive the long-term development of the AI industry, although this will require a prolonged transitional period. Currently, medical institutions are the payers for imaging AI, procuring AI systems to enhance operational efficiency and reduce costs. “Even without independent reimbursement, the value of AI remains undeniable. In the long run, the key to imaging AI becoming a separately billable service lies in ensuring its value is quantifiable.”

Tertiary hospitals are the primary battleground for medical imaging AI. Overcrowded large hospitals leave doctors stretched thin. The advent of AI can alleviate physicians’ workloads, freeing up their time to address more complex clinical challenges and enabling them to refocus on the core practice of medicine.

Take the COVID-19 pandemic as an example. The demand for AI in large hospitals is primarily manifested in the need for physicians to process vast amounts of patient data and information within a short timeframe. Failure to do so promptly could result in 10–20% of COVID-19 patients infecting others while awaiting results. AI solutions can rapidly identify potential COVID-19 cases, alleviate physicians’ workload, and shorten the cycle for report interpretation. The value of AI was further amplified during the pandemic. In fact, this demand was already substantial prior to the outbreak.

Yuan Jianhua, Director of the Clinical Application Center at Mingfeng Medical, believes that AI currently serves primarily as an auxiliary tool, playing a significant role in a limited number of medical diagnostic processes. “It will take a long journey before AI can truly replace human physicians.” Furthermore, AI should not be confined to large hospitals; primary care institutions also need the support of AI.

Wang Haijiao expressed agreement with this viewpoint. Both AI software and system equipment hold significant potential for application at the primary care level.

For primary healthcare institutions, insufficient medical resources mean that primary care physicians lack adequate experience in interpreting medical images due to limited exposure to and learning from a wide variety of cases. AI can assist these physicians in making diagnoses.

“Both large hospitals and primary care institutions need AI, albeit with different demands. As technology advances, AI will demonstrate greater social and commercial value in primary care settings,” said Chen Kuan.

Zhan Yiqiang agrees with the view that there is a rigid demand for AI in both large hospitals and primary care institutions. He added that new medical technologies often trickle down from higher-tier to lower-tier facilities. AI technology is refined, piloted, and matured in large hospitals; once it gains recognition, it is then deployed to primary care institutions. This development path differs from that of AI technologies in other industries.

“With economic development, our interactions with domestic and international clients have revealed that an increasing number of high-end medical devices are being deployed at the primary care level. This trend has significantly increased the volume of imaging studies reviewed by primary care physicians, placing greater demands on their diagnostic interpretation capabilities. This presents a significant opportunity for the development of medical AI.”

According to publicly available statistics as of the end of 2018, a total of 428 million inspection and testing reports were issued in 2018, averaging 1.17 million reports per day. In the training of algorithmic models, large volumes of high-quality, diverse medical images and precisely annotated expert data are required; such data are most abundant, comprehensive, and authoritative in Grade A tertiary hospitals.

How Can Data Be Utilized Effectively? This Involves Issues of Patient Privacy Protection and Data Interoperability.

Zhan Yiqiang stated that “data” is a sensitive term. With the advancement of AI technology, China has placed increasing emphasis on data security. Ensuring data security requires safeguards on two fronts: institutional and technical.

On the one hand, enacting corresponding laws and regulations can facilitate a more rational resolution of issues pertaining to the application of medical data. On the other hand, AI companies can collaborate with hospitals to develop cybersecurity technologies, ensuring that data is protected when physicians upload it to AI servers. Furthermore, at the algorithmic level, technologies such as “federated learning” offer the possibility of achieving physical isolation of training data.

Compared with overseas countries, China’s data protection measures are more standardized and stringent. Regarding data security management, Chen Kuan provided a detailed elaboration on the differing management styles between China and the West.

The difference between the two lies in the fact that Western countries place greater emphasis on procedural compliance, whereas China adopts a result-oriented approach. “From a purely behavioral perspective, what enterprises need is not the data itself, but the general patterns underlying the data. The final AI products produced are also independent of the specific data information,” said Chen Kuan. “In this context, China’s regulatory model is better suited to the development of the AI sector. As such systems continue to be refined, and leveraging its large population base, China is poised to develop a data protection framework that is more conducive to the advancement of artificial intelligence.”

In response, Wang Haijiao added that overseas countries place greater emphasis on procedural compliance in data management—for instance, whether relevant personnel are informed and have granted authorization when data is used. Wang Haijiao believes that, in the future, China’s data regulatory authorities may integrate Chinese and Western management approaches by establishing a big data regulatory body, thereby providing enterprises with a clear legal framework for all stages from data collection to utilization.

The current epidemic outbreak presents both challenges and opportunities for the healthcare industry. Not only is data regulation stringent, but the approval process for bringing medical AI products to market is also highly rigorous. What are the three panelists’ views on the review and approval of AI products discussed in the forum?

Regulatory approval is a critical step for medical technologies to enter the market, even shaping the entire industry’s landscape. For innovative technologies, particularly those with the potential to exert significant influence on the industry, high standards are essential. Chen Kuan further noted that although this imposes considerable pressure on AI companies in the short term, it is worthwhile from the perspectives of practitioners, patients, hospitals, and physicians.

Zhan Yiqiang shared his insights on the regulatory authorities’ changes to product registration.

On March 5, 2020, the Center for Medical Device Evaluation of the National Medical Products Administration issued the “Key Points for the Review of CT Imaging-Assisted Triage and Assessment Software for Pneumonia (Trial).” “The regulatory authorities have introduced review guidelines specifically targeting COVID-19 pneumonia, which is a remarkably swift response.”

In terms of content, Zhan Yiqiang highlighted two key points. First, the "Review Guidelines" designate triage as an intended use for AI software. Previously, AI products in China were primarily defined as auxiliary diagnostic tools. Triage refers to the process by which AI determines whether a patient may have a specific disease and notifies physicians to adjust the priority of image interpretation based on disease severity. AI-based triage is particularly meaningful for highly contagious or urgent conditions. The second noteworthy point is that the "Review Guidelines" emphasize the role of AI in quantitative disease analysis. This is precisely an area where AI holds a distinct advantage over humans. It is believed that the new elements introduced in these "Review Guidelines" will provide broader insights for the development of future AI products.

The performance of stock markets in Europe and the United States demonstrates that the COVID-19 pandemic has inflicted a significant shock on the global economy. Amid this tidal wave, the healthcare industry has emerged as one of the few survivors not swept away.

During the pandemic, the “Sky Eye” CT scanner, integrated with United Imaging Intelligence’s uVision smart camera technology, became a powerful tool in the fight against COVID-19 by enabling fully automated, contactless scanning and thereby reducing the risk of infection for technologists. This represents a typical case of AI-empowered medical equipment.

In empowering physicians, United Imaging’s first “AI+CT” intelligent auxiliary analysis system for novel coronavirus pneumonia was officially completed on February 8. This software can detect subtle lesions, segment lesions using deep learning algorithms, and automatically generate reports for physicians, thereby “improving efficiency while also helping physicians enhance diagnostic accuracy.” Subsequently, the software was deployed in multiple frontline hospitals. For primary healthcare institutions, the application leverages United Imaging Cloud for cloud-based deployment, delivering the models to the frontline without delay.

From product definition, data collection, and annotation to the development of the first model prototype and its practical application, the entire process took only one week. In Zhan Yiqiang’s view, they accomplished a task that seemed impossible. “This was a pleasant surprise for me.” Since its establishment in 2017, United Imaging Intelligence has focused on technological accumulation and the refinement of its R&D model, dedicating itself to developing core algorithm engines and building a professional annotation team. This is precisely why United Imaging Intelligence possesses such rapid responsiveness.

"The impact of the pandemic on AI companies has objectively expanded the influence of AI, 'particularly by helping the vast grassroots market recognize the utility and benefits of AI. This may prompt many physicians to explore more applications of AI. However, this is not reflected in sales volume or revenue metrics.'"

Early detection, early diagnosis, and early treatment are the fundamental principles of prevention and control. In response to China’s epidemic prevention needs, Minfound Medical has rapidly launched the Ark CT (a comprehensive mobile cabin-based emergency CT solution). This system integrates a CT scanner, an independent operator room, an independent scanning room, air conditioning, and other components, enabling immediate operation upon power connection. The physical separation between the operator room and the scanning room facilitates a seamless workflow of scanning–disinfection–scanning, ensuring one-patient-one-disinfection protocols. This design prevents cross-infection between healthcare workers and patients, interrupts transmission routes, and thereby safeguards early CT diagnosis in the prevention and control of infectious diseases.

Currently, China’s epidemic has entered its final phase, with the focus of prevention and control shifting to preventing imported infections. Outside China, the number of COVID-19 cases is rising sharply. In this context, AI companies are also expanding overseas, seizing the opportunity to enter international markets.

Wang Haijiao stated that while overseas countries currently favor nucleic acid testing, China emphasizes the combined use of nucleic acid testing and imaging techniques.

Building on its strong foundation in the Japanese market, Infervision partnered with Doctor Net, Japan’s largest tele-diagnosis company, to promptly deploy Infervision’s “Intelligent AI-Assisted Screening and Epidemic Monitoring System for Pneumonia.” In Japan, where CT scanner availability is high but radiologists are in short supply, this AI system for COVID-19—an “expert” that rapidly emerged during the pandemic—has established a safety net to reduce missed diagnoses and enable precise diagnosis, forming the first line of defense in epidemic screening through chest CT imaging.

In response to the ongoing spread of the COVID-19 pandemic, the University Hospital Campus Bio-Medico in Rome, Italy, promptly sought assistance from Infervision’s European division, hoping to rapidly deploy Infervision’s “Intelligent Auxiliary Screening and Epidemic Monitoring System for Pneumonia” within its facilities. Sun Yipeng, head of Infervision’s European division, led his team in urgently mobilizing products and personnel. They drove from Frankfurt to Rome, risking their lives to bring China’s anti-epidemic experience and technology to Italy, the European country hardest hit by the outbreak.

Chen Kang candidly stated that the destiny of Chinese medical AI enterprises lies in expanding overseas; even if they are “acquired” by leading foreign companies, they must position themselves to compete on a global scale. He believes that the competitors for Chinese medical AI firms should come from overseas markets.

In fact, the United States, Israel, and India are all formidable competitors. The United States holds an advantage in its highly streamlined commercialization pathways. Its healthcare sector is massive in scale, and society maintains an encouraging stance toward the implementation and reimbursement of innovative technologies. In countries like Israel, government-led initiatives help enterprises access scientific research data and ensure product accuracy, thereby enabling them to expand globally and capture a share of the international market.

In the future, the competitive landscape of the industry will inevitably involve numerous factors, including talent and technology, institutional standards and regulations, and government guidance.