Cracking the Profitability Code of Medical AI: An Exclusive Interview with Professor Feng Xiaoyuan
Recently, almost everyone has been talking about ChatGPT.
This generative AI chat application surpassed 1 million users within just five days of its launch. In merely two months, it achieved over 100 million monthly active users. ChatGPT has not only fully demonstrated humanity’s enthusiasm for artificial intelligence but also triggered a surge in limit-up gains for AI-related stocks, propelling the AI industry—once fraught with sighs of disappointment—back into the spotlight. The shadows of profitability challenges and the predicaments of implementation hurdles have been left far behind in the past.
The advent of ChatGPT has naturally sparked excitement and enthusiasm within the industry. However, in the more serious realm of medical AI, challenges such as regulatory review and approval, data silos, algorithmic limitations, and product homogenization remain critical issues for companies to address. As collaborations between medical AI enterprises and hospitals continue to advance, we must ask: How can a mainstream value system and viable commercial pathway be established to truly overcome profitability challenges?
Recently, Shenzhi Technology invited Professor Feng Xiaoyuan, a renowned authority in Chinese medical imaging, to serve as Chairman of its Expert Committee. In this role, Professor Feng will advance the integration of precision diagnosis and treatment capabilities into the company’s high-tech products, continue to expand its product pipeline, provide value-added services such as discipline development and educational training to primary healthcare institutions, and offer strategic guidance for the company’s future growth.
Seizing this opportunity, VCBeat engaged in an in-depth discussion with Professor Feng Xiaoyuan regarding the pain points of AI in medical imaging. The dialogue aimed to decipher the code to overcoming the profitability dilemma of AI by analyzing clues left from years of development in the field, summarizing the findings into five key points. This serves both as a review of the industry’s past trajectory and an outlook on the future development of AI in medical imaging.
Former Vice President of Fudan University
Dean of Shanghai Medical College
Chairman of the Expert Committee on Intelligent Medical Imaging, Shenzhi Technology
Professor Feng Xiaoyuan
Artificial intelligence technologies face various limitations in real-world clinical applications. Enterprises must understand the entire application workflow and address practical problems through a clinically oriented approach tailored to specific scenarios, rather than deploying single models or algorithms directly into clinical practice.
In Professor Feng’s view, as a technology, AI needs to empower a specific system or discipline and create new value, thereby accelerating the development of that discipline or industry. Therefore, in medical applications, AI must also adhere to the rules of the medical context and create new value while aligning with the characteristics of clinical medicine.
If AI merely provides a marginal boost to physicians’ workflow efficiency or automates tasks that can already be performed manually, it will struggle to stand out in a crowded market. Many challenges remain unresolved in clinical medicine. AI companies should adopt a clinically oriented approach, focusing on technologies that create new value and deliver novel breakthroughs in clinical practice, thereby improving the standard of care and reducing patient costs. Such R&D efforts will be sustainable and represent the key to addressing the homogenization of AI products.
Using this as a clue to clarify the rules for applying artificial intelligence technology and define clear endpoints is fundamental to the application of AI in healthcare scenarios. Therefore, future AI will undoubtedly empower healthcare by deeply understanding clinical contexts and creating new value, rather than focusing solely on breakthroughs in AI technology itself.
As previously mentioned, the work currently performed by artificial intelligence is largely foundational and auxiliary, remaining within the scope of physicians’ capabilities. Furthermore, its diagnostic proficiency, developed through big data training, remains unsatisfactory. In other words, while this technology has improved physicians’ work efficiency and reduced their workload to some extent, it has yet to generate substantial new value for clinical practice.
How to Address This Issue? Professor Feng Offers Four Approaches:
First, AI enables the diagnosis and treatment of diseases that were previously difficult or impossible to manage, such as by revealing diagnostic findings invisible to the naked eye;
Secondly, for disease diagnosis and treatment processes that require extensive data analysis and consume significant time and energy from physicians, the integration of AI technology can streamline these workflows. This liberates doctors from repetitive, screening-based, and rule-driven tasks, ultimately enhancing diagnostic and therapeutic efficiency and success rates;
Furthermore, by leveraging AI to enhance the efficiency of diagnosis and treatment for existing diseases, as well as to expand diagnostic and therapeutic methods and scenarios, we can ensure that many individuals who lack the opportunity or time to visit hospitals are able to access treatment and achieve better clinical outcomes.
Finally, AI empowers the physician’s mind. In an era of information overload, the human brain cannot store, analyze, transmit, synthesize, and compare large volumes of data within a short timeframe. The role of artificial intelligence is to process massive amounts of information or data, uncover inherent patterns, and perform precise quantitative analyses, including details that may be overlooked during diagnosis. Ultimately, AI handles pattern recognition and logical tasks, enabling physicians to return to the patient’s bedside, where they apply human wisdom to address innovation and humanistic aspects, thereby further reshaping and refining diagnostic criteria and workflows.
Profitability has become a key consideration for leading AI companies. Although the approval rate of AI medical imaging products has continued to rise since the first Class III medical device certificate was awarded in this sector in 2020, with related products from various enterprises being approved one after another, achieving large-scale commercialization still requires clearing the hurdles of pricing approval and health insurance reimbursement inclusion. Moreover, there is no consensus on charging patients for AI-assisted diagnostic imaging systems, and the business model remains unclear. There is still a long way to go for AI products to achieve large-scale commercialization.
Professor Feng Xiaoyuan believes that the processes for market access, pricing approval, and medical insurance reimbursement are extremely lengthy, involving numerous factors; thus, they cannot be determined by technology alone. Meanwhile, obtaining regulatory clearance only demonstrates the safety and efficacy of the technology, without having a fully direct correlation with whether the final product can achieve commercial success in the market. Currently, the most critical questions AI companies need to answer are “who will use it” and “who will pay for it,” and they must learn to integrate considerations of these issues into their product design and research and development efforts.
For instance, during product design, companies must clearly identify what customers truly need, how to deliver tangible benefits to them, and ultimately ensure that someone is willing to pay. These customers may include physicians; when physicians benefit, the direct outcome is that patients also gain advantages. If patients experience significantly reduced diagnosis and treatment times or improved accuracy and precision due to physicians’ use of AI products, they will naturally be willing to pay, thereby completing the commercial loop for AI. On the other hand, from a long-term perspective, if hospitals can enhance diagnostic and therapeutic efficiency and reduce labor costs through AI assistance, this will contribute to increased hospital revenue and strengthen their willingness to adopt and pay for such solutions.
From this perspective, the top priority for any AI company is to give its products a clear positioning. Companies must first identify who their users are, what value they can deliver to them, and whether this value is sufficient to make customers willing to pay. When the problems that technology can solve align with the needs and interests of target users, the value of the technology becomes evident, making it easier to break through profitability challenges.
Primary care has always been a key application scenario for AI in medical imaging. Relevant data indicate that 70% of healthcare institutions at the county level have demand for AI products. However, due to constraints such as insufficient funding and the remote location of grassroots areas, the deployment of related products at the primary care level faces challenges, leaving unmet needs. In the implementation process, common practical pain points include “unaffordable equipment, unintelligible images, and inability to perform treatments.” Owing to issues related to cost and operation, existing diagnostic devices are hardly able to penetrate down to the primary care level.
Professor Feng stated that the key to implementing artificial intelligence technology at the primary care level lies in enhancing two core capabilities. First, the professional competence of physicians themselves. Second, the capacity for diagnosing and treating diseases at the primary care level, improved by remotely allocating additional medical resources. This also represents the strategic focus of AI enterprises.
Specifically, companies can simultaneously enhance physicians’ capabilities in disease management and diagnosis through remote education and training, while leveraging artificial intelligence platforms as intermediaries to rationally allocate telemedicine resources. By providing various resource supports to primary healthcare institutions via the internet and data processing platforms, these companies enable patients to receive services at primary hospitals that are equivalent to those offered by Grade A tertiary hospitals. For instance, although primary healthcare facilities may possess appropriate equipment, such equipment often remains idle due to a lack of trained personnel. Through intelligent platforms, however, primary hospitals can connect with specialists from large hospitals or access multidisciplinary diagnostic and therapeutic resources, thereby ensuring full utilization of equipment and effectively addressing clinical challenges. In this way, the value of technology is truly realized.
“The driving force behind our deep commitment to intelligent hardware and software is the continuous reduction of product costs. Our ultimate goal is to optimize solutions and enhance the operational efficiency of primary healthcare services, thereby facilitating adoption in highly fragmented primary care settings and improving the economic viability of our products,” said Feng Xiaoyuan.
Furthermore, Professor Feng discussed the importance of miniaturization, portability, and mobility in medical imaging equipment, as well as the potential for AI applications in this field.
Imaging equipment such as CT, MRI, and ultrasound systems is bulky, expensive, and requires specialized expertise to operate. As a result, primary healthcare institutions often cannot afford the associated costs, leaving the diagnostic and treatment needs of patients at this level unmet. Furthermore, even in well-equipped hospitals, the immobility of large-scale imaging devices poses significant limitations during emergency medical situations or when bedside diagnostics are required, potentially introducing substantial risks to patient care.
Therefore, miniaturized, portable, and mobile devices can not only expand the application scenarios of medical imaging products but also effectively address critical pain points in hospitals. Meanwhile, integrating AI technology further empowers these devices by, for example, reducing physicians’ image interpretation time, enhancing image clarity, and improving diagnostic accuracy.
On the other hand, the development of telemedicine technologies and the promotion of tiered diagnosis and treatment policies have further amplified the demand for AI technologies among primary healthcare institutions. AI technology, combined with miniaturized, portable, and mobile devices, will break down hospital walls, enabling imaging equipment to be applied in broader scenarios. From the perspective of information acquisition, this represents a significant step forward.
It is important to note that computing power remains the core element. The black-and-white images displayed by medical imaging devices are composed of pixels. To approximate and access the essence of lesions behind these pixels—namely, their biological attributes—it is necessary to enhance the computational power of AI technologies. Without sufficient computing power, all other efforts would be meaningless.
It is understood that, through the collaboration between Professor Feng Xiaoyuan and Shenzhi Technology, the company has extended its core computing power technology from ultrasound equipment to other imaging devices. Currently, Shenzhi Technology is developing miniaturized, portable, and mobile MR equipment, having created a magnetic resonance imager with the world’s lowest field strength of 0.05 Tesla. This innovation addresses the long-standing requirement for shielded rooms in magnetic resonance imaging over the past five to six decades, as well as the challenge of deploying high-performance premium imaging equipment at primary care levels.
Since AlphaGo defeated world Go champion Lee Sedol in 2016, several years have passed for artificial intelligence. However, after years of development, the field has fallen into a period of stagnation. Fortunately, this year, people's enthusiasm for AI has been reignited by a generative AI chat application called ChatGPT, much like it was seven years ago when AlphaGo beat Lee Sedol. For businesses, ChatGPT has also lived up to expectations by launching its ChatGPT Plus paid subscription plan, rapidly initiating commercial monetization and offering AI companies a glimpse of profitability.
The renowned consulting firm Gartner once asserted that any industry or enterprise, provided it has specific application scenarios, accumulated data, and computational power, can implement artificial intelligence applications. We have reason to believe that, with concerted efforts from various stakeholders, a profitable future for medical AI may be just around the corner.