Chinese medical AI company DIAGENS is claiming a paradigm shift in clinical artificial intelligence after hospital teams using its foundation model platform won first and second prizes in China's National Healthcare Security AI Image Recognition Contest. The results suggest that large-scale pre-trained models may enable clinicians to develop specialized diagnostic AI tools without programming expertise—a development the company characterizes as a "ChatGPT moment" for medical imaging.

The contest, organized by China's National Healthcare Security Administration, drew 1,332 participating teams nationwide. First prize went to a CT diagnostic model for kidney cancer developed by the Second Affiliated Hospital of Guangxi University of Chinese Medicine, while second prize was awarded to a breast cancer X-ray diagnostic model from the First Affiliated Hospital of Guangxi Medical University. Both winning teams built their models using DIAGENS' iMedImage® platform.

The Foundation Model Approach to Medical Imaging

The iMedImage® platform contains 104 billion parameters and was pre-trained on 80 million clinical imaging studies, according to DIAGENS. The foundation model architecture supports multi-modal compatibility across CT, X-ray, MRI, ultrasound, karyotyping, and cytomorphology—covering 26 clinical specialties and over 4,000 disease scenarios.

This "one foundation for thousands of models" approach mirrors the strategy that propelled large language models into mainstream use, where a single pre-trained system enables diverse downstream applications. DIAGENS' iMedLoop™ platform provides what the company describes as an integrated closed-loop development pipeline for medical imaging AI research and production acceleration.

The company reports deploying 158 model projects across 99 hospitals, with 65 classified as Grade III Class A facilities—China's highest hospital tier. These deployments span 43 anatomical sites and 61 disease categories.

Implications for Clinical AI Development

The contest results highlight a potential shift in how medical AI tools are developed and deployed. Rather than requiring dedicated AI engineering teams or partnerships with technology vendors for each diagnostic application, foundation model platforms may enable clinical departments to create institution-specific tools tailored to local disease patterns and imaging protocols.

For rare diseases and underserved specialties where commercial incentives for standalone AI products remain limited, this democratization of development tools could accelerate availability of AI-assisted diagnostics. However, questions about model transparency, validation rigor, and regulatory pathways for clinician-developed AI tools remain unresolved.

DIAGENS, formally known as Hangzhou Diagens Biotechnology Co., Ltd., was founded in 2016 and is listed on the Hong Kong stock exchange. The company claims to be the world's first publicly listed company focused specifically on medical imaging foundation models.

The platform-as-a-service model represents a potentially scalable commercialization pathway for medical imaging AI, shifting from bespoke model development toward enabling rapid deployment across multiple disease categories and imaging modalities. Whether this approach can maintain clinical validation standards while accelerating development timelines will likely determine its broader adoption in healthcare systems worldwide.

Sources

  1. Chinese medical AI model wins top state science award as sectoral "ChatGPT moment" arrives — biotech-365.com