Fosun Pharma Accelerates Medical AI Deployment with Two High-Precision Products Launched Within a Year, Leveraging Clinical Feedback for Rapid Iteration

In recent years, computer image processing technology has advanced rapidly, reaching a level of maturity that meets the demands of clinical commercial applications. Consequently, an increasing number of AI-based medical imaging products are transitioning from laboratories to hospital departments, vying to become integral components of smart hospitals and serving clinicians in their practice.

However, even in the most common clinical application area of AI-assisted medical imaging diagnosis, the usability of medical AI remains very low. On one hand, among the thousands of diseases that can affect humans, medical AI is capable of providing auxiliary diagnosis for only a limited number of conditions. On the other hand, the professionalism and intelligence level of medical AI are far from meeting the qualifications of competent physicians, let alone those of the scarce top-tier specialists.

Medical AI, often portrayed as a high-tech product, may be viewed by some physicians as having limited utility. According to He Chuan, CEO of Xingmai Technology, only by accurately identifying market pain points can truly superior products be developed—a principle that is especially critical at this stage, when national standards for medical AI products remain undefined. “Iterative updates must be driven by continuous user feedback,” He Chuan told VCBeat (WeChat ID: VCBeat). The “users” he refers to are primarily hospitals.

Xingmai Technology is a medical AI project incubated within Fosun, dedicated to the innovation and R&D of artificial intelligence technologies in the broader healthcare sector. It enables early screening for various malignant diseases based on medical imaging and provides intelligent, precise diagnostic assistance services to diverse medical institutions.

He Chuan also serves as an Assistant to the President of Fosun International. She brings over eight years of experience in investment, financing, and industrial operations within high-tech enterprises. Centered on the core values of “Health, Happiness, and Prosperity,” Fosun Group has cultivated deep expertise in the healthcare industry for many years, accumulating extensive medical resources. He Chuan believes that by integrating Fosun’s abundant medical resources with medical AI, it is possible to develop truly high-quality products. Her vision has gained widespread recognition across Fosun Group, including from Fang Qu, who comes from a technical background. A veteran in the field of artificial intelligence, Fang Qu has been engaged in research and development in computer vision—such as facial recognition and image-based search—since 2007. When he joined Fosun in 2017, he served as a technical expert, providing big data and AI services to multiple sectors within the group, including finance, retail, and insurance. Fang Qu currently serves as the Chief Technology Officer (CTO) of Xingmai Technology.

Xingmai Technology has launched two products to date and established partnerships with more than 30 hospitals across China.

In February 2018, Xingmai Technology formally established an independent commercial entity. However, as early as February 2017, Xingmai Technology had already begun collaborating with Grade A tertiary hospitals to train algorithmic models. In April 2018, Xingmai Technology launched its first auxiliary diagnostic product for lung imaging, “Xingmai Ruiying–Lung.” In July 2018, Xingmai’s auxiliary diagnostic product for orthopedics, “Xingmai Ruiying–Orthopedics,” was officially released.

In September 2018, Xingmai Technology’s “Future Clinic” made its debut at the 2018 World Artificial Intelligence Conference, garnering widespread attention from the industry. He Chuan told VCBeat that the Future Clinic represents Fosun Group’s holistic vision for the health technology sector, with Xingmai Technology playing a pivotal role in this initiative.

According to Fang Qu, Xingmai Ruiying achieved first place in both categories—average recall rate for false-positive removal based on a designated dataset and on a nodule-screened dataset—at the LUNA Challenge, an internationally authoritative medical imaging competition held in September and November 2017. At that time, the average recall rate for pulmonary nodule detection by Xingmai Ruiying was 96.6%. “We made significant progress in 2018,” Fang told VCBeat. Although the new model had not yet been integrated into commercial products, its average recall rate was already very close to 100%, while the AUC (area under the curve) for benign judgment reached 94%. Another commercially launched product from Xingmai Ruiying, designed for assisted diagnosis of avascular necrosis of the femoral head, also demonstrated an AUC between 97% and 98%.

Fang Qu pointed out that the design philosophy of Xingmai Ruiying embodies a combination of breadth and depth.

In terms of breadth, current medical AI products on the market are far from covering all the signs required for medical laboratory testing. Therefore, Xingmai Ruiying has been continuously striving to expand its monitored anatomical regions and imaging modalities, extending its product line from radiology to clinical laboratory medicine, pathology, and ultrasonography.

In terms of depth, Xingmai Ruiying conducts targeted, in-depth analysis of various anatomical regions. Fang Qu stated that expanding medical AI into greater depth demands high technical capabilities from the team. To accomplish this, Xingmai Technology has assembled an R&D team comprising graduates from prestigious institutions such as Stanford University, the University of California, Berkeley, Peking University, and Tsinghua University.

Specifically, Xingmai Technology’s in-depth product optimization primarily focuses on iterative improvements for pulmonary and orthopedic conditions, as well as cardiovascular and neurological diseases slated for upcoming release. The company continuously enhances key performance metrics such as sensitivity and specificity, while optimizing product usability to align with hospital workflows. Xingmai Ruiying determines the scope of deep learning analysis based on the clinical realities of each disease, ensuring comprehensive coverage of all radiological signs associated with a given condition.

Taking the diagnosis of pulmonary diseases as an example, in addition to enhancing the detection of ground-glass nodules, Xingmai Ruiying also incorporates more than a dozen pulmonary conditions, such as bullae and emphysema, into its analysis. This approach aims to comprehensively expand the scope of computer-aided lung diagnosis and provide clinicians with structured diagnostic reports based on these findings.

In addition, Xingmai Technology has over 50 mature algorithmic directions in its pipeline and is expected to launch three new medical AI products this year. According to Fang Qu, the algorithmic areas covered by Xingmai Technology include auxiliary diagnosis and treatment for lung X-rays, breast cancer case detection, intelligent diagnosis of knee osteoarthritis, and auxiliary diagnosis of stroke.

In terms of commercial implementation, Xingmai Technology leverages the extensive high-quality medical institution resources of Fosun Group, thereby objectively facing relatively less pressure to deliver hospital coverage performance. Furthermore, the mature sales team established by the Group has accelerated the market expansion of its products.

He Chuan told VCBeat that Xingmai Technology places particular emphasis on user feedback during the implementation process, continuously iterating and updating its offerings—including algorithmic technologies, as well as the usability, user experience, and applicability of the products themselves. “Although many products, including Xingmai Ruiying, have already been launched in the market, there remains substantial room for enhancement and improvement in actual clinical practice.” Xingmai Technology regards users as partners, aiming to collaborate with partner hospitals during trial use and research to jointly achieve improvements, co-design product iteration goals, ensure the products address real-world problems, and thereby increase user stickiness.

“Xingmai Technology’s implementation strategy is to rapidly develop and launch a product, engage in deep integration and operational collaboration with hospitals to achieve a relatively ideal and optimized state, and then proceed with expansion,” He Chuan explained.

Furthermore, Xingmai Technology’s R&D team comprises experts from diverse fields such as deep learning, fluid dynamics, computer vision, and chip design, complemented by dedicated teams for medical affairs, product engineering, and commercialization. This organizational configuration and structure are rare among startups.

As a benchmark implementation of Fosun’s “Innovation-Driven” strategy, Xingmai is tasked with benchmarking against global high-tech giants. Looking ahead, Xingmai Technology envisions itself as a key pillar in the field of AI-assisted diagnosis and decision support, continuously extending its reach to grassroots levels and becoming an integral part of inclusive healthcare. He Chuan believes that future medical services in grassroots and home settings will rely on three core technologies: artificial intelligence, big data, and smart hardware. Currently, Xingmai Technology is focusing on artificial intelligence, which holds foundational significance.