Recently, at MICCAI 2025, the premier international conference in medical image computing and computer-assisted intervention, Shanghai Yinghe Medical Pulse Intelligent Technology Co., Ltd. ("Yinghe Medical Pulse"), an AI enterprise incubated by Sunshine Guojian Healthcare Data Technology Co., Ltd. (02522), emerged victorious among hundreds of global participants in the 3D Visual Language Multimodal Challenge. The company secured first place in the CT image automatic report generation task and third place in the 18-label abnormal classification task, demonstrating that Chinese teams now possess international leadership capabilities in this field.
**International Competition Excellence: Defeating Top Teams from Harvard and Stanford**
The MICCAI 2025 3D Visual Language Multimodal Challenge boasts an impressive lineup of international and authoritative organizers. Academic institutions include top universities such as the University of Zurich and its affiliated hospitals, Boston University, Harvard University, Johns Hopkins University, Imperial College London, and Shanghai Jiao Tong University, along with key national medical research institutions like the U.S. National Institutes of Health (NIH). From the industry perspective, leading computing platform provider NVIDIA provided support, offering top-tier research and data resources that ensured the challenge's broad impact in both academic value and clinical applications.
The 3D Visual Language Multimodal Challenge under MICCAI 2025 is regarded by industry professionals as a "litmus test" for measuring AI technology's clinical implementation capabilities, as it directly addresses clinical diagnostic pain points. This challenge attracted hundreds of participants from global leading institutions including Harvard Medical School, MIT, and Stanford University Hospital, covering four core clinical application scenarios including CT image automatic report generation and multi-label abnormal classification.
The CT image automatic report generation task required training multimodal models based on the CT-RATE dataset to automatically convert 3D chest CT scans into clinically accurate radiological reports. Yinghe Medical Pulse ranked first among 161 participants in this task. Their generated reports not only completely presented core information such as lesion location, size, and morphology but also provided diagnostic tendencies based on imaging features, highly aligning with clinical physicians' diagnostic logic and demonstrating outstanding performance in key scoring metrics, achieving an average BLEU (1-4) of 0.2791, 5 points higher than baseline standards.
The multi-label abnormal classification task required precise multi-label prediction for 18 types of chest abnormalities in the CT-RATE dataset, including challenging conditions such as pulmonary nodules, pneumonia, and aortic calcification. Yinghe Medical Pulse ranked among the top three out of 118 participants in this task, achieving an average AUC of 0.846 for 18 chest abnormalities, demonstrating strong generalization capabilities and providing efficient solutions for precise identification of complex lesions in clinical practice.
The competition results demonstrate that Yinghe Medical Pulse's technical capabilities in medical imaging AI development have reached international leading levels. Behind Yinghe Medical Pulse's technological breakthroughs and practical implementation achievements, this globally leading advantage was rapidly achieved through the long-term commitment to deep integration of "Industry + AI" by Sunshine Guojian and Yinghe Medical Pulse.
**Deep Integration of "Industry + AI" Accelerates Intelligent Medical Imaging Implementation**
As China's largest third-party medical imaging service platform, Sunshine Guojian not only provides Yinghe Medical Pulse with massive real-world imaging data, offering rich "materials" for model training and optimization, but also establishes a comprehensive verification environment "from R&D to implementation" through its nationwide imaging center network, allowing technology to be continuously refined in real clinical scenarios.
Simultaneously, Yinghe Medical Pulse's AI technological capabilities provide feedback to Sunshine Guojian's business development, forming a closed-loop ecosystem of "data-model-application-data." This "bidirectional empowerment" model creates synergistic advantages in medical imaging AI layout for both parties, becoming a typical representative of "data support + top-tier technology + extensive implementation" domestically, providing a reference practical path for the medical imaging AI industry's development.
Currently, Sunshine Guojian has built China's largest and fastest-growing medical imaging database, with a service network covering 115 imaging centers generating nearly 10 million new data entries annually, encompassing multimodal imaging including CT and MR. The company will accelerate expansion and quality improvement of its standardized datasets, strengthening the positive cycle of "more data - more accurate models - broader applications."
Sunshine Guojian's data advantage lies not only in "quantity" but also in "compliance and control." Through proprietary data governance architecture and privacy computing technology, it achieves "usable but invisible" data, addressing core concerns of medical institutions regarding data sharing.
On the other hand, Sunshine Guojian continues investing in AI application platform development, focusing on supporting iteration and commercial implementation of imaging foundation large models, providing technical reserves for subsequent multi-anatomical AI product development. Yinghe Medical Pulse, incubated by Sunshine Guojian, launched the world's first multimodal full-process medical imaging foundation large model "Yinghe Miya®," which represents its core technical barrier in seizing AI opportunities. This model is pre-trained with tens of millions of standardized data points and, through a "foundation pre-training + downstream fine-tuning" approach, reduces AI product development cycles by over 70%, achieving a transition from "point breakthroughs" to "full-process empowerment" in medical AI.
Additionally, Sunshine Guojian collaborates with Huawei, Alibaba, iFlytek, and others to jointly develop large model integrated machines and application-end data products, expanding application scenarios from imaging diagnosis to health management and medical insurance services, forming a complete closed loop of "data - model - product - clinical feedback."
This "comprehensive scenario coverage" layout aligns with the "AI + People's Welfare" goals mentioned in the State Council's "Opinions on Deeply Implementing 'AI+' Actions" and elevates "Yinghe Miya®" from a single technical tool to an industry ecosystem platform. By opening API interfaces to attract third-party developers and promoting "technological democratization," it responds to policy advocacy for "open-source ecosystem prosperity" while opening incremental markets for technology monetization.
From an industry trend perspective, imaging data, as one of the largest volume and highest clinical value categories in medical data, has become a core element driving medical AI innovation. Sunshine Guojian's practice proves that only by establishing a closed loop of "compliant data circulation - model iterative optimization - scenario commercial implementation" can the value of medical data be truly realized.
Previously, at the 2025 China International Fair for Trade in Services in September, Yinghe Medical Pulse, together with the National AI Application Pilot Base (Medical Field) and Beijing Tiantan Hospital affiliated with Capital Medical University, jointly released systematic R&D achievements of cranial medical imaging AI systems based on the "Hetu Plan" - the Cranial CT Core Definition Table and Cranial L1 Foundation Model. This plan represents a systematic framework for medical imaging artificial intelligence covering data standards, foundation models, and application ecosystems, committed to addressing long-standing pain points in the medical imaging AI field such as "fragmented product application scenarios, disorderly growth constraining implementation, and insufficient collaboration between medical and technological sectors," aiming to build China's autonomous and controllable smart healthcare infrastructure.
From winning the global MICCAI championship to demonstrating implementation achievements at the trade fair, Yinghe Medical Pulse's recent "dual breakthroughs" contribute to China's medical imaging AI technology transition from "following" to "leading."
Compared to international peers, Yinghe Medical Pulse's core advantage lies in consistently adhering to the philosophy of "technology originates from clinical practice and serves clinical practice." Leveraging Sunshine Guojian's rich real clinical resources, its AI models deeply integrate specific needs of Chinese medical scenarios from the early development stage, demonstrating not only excellent algorithmic performance in competitions but also superior applicability and reliability in practical applications. This development path closely combining clinical practice is key to Yinghe Medical Pulse's ability to both win international top competitions and achieve large-scale commercial implementation, providing a promising successful paradigm for China's autonomous innovation and industrial application of medical imaging AI.