The First Affiliated Hospital of Zhengzhou University has deployed an integrated medical imaging digitalization and AI platform that has centrally received, managed and scheduled more than 350,000 imaging studies, lifting imaging examination volume by 20 percent and compressing full-dimension data governance from year-level to month-level cycles.
Before the platform, imaging AI at the hospital was largely limited to single-disease applications such as lung nodule and cardio-cerebrovascular screening, far from multi-disease clinical needs. Introducing many disease-specific AI engines at once created new bottlenecks — device port and IP exhaustion, PACS throughput pressure, network security exposure, and repeated large-file storage — while clinicians lacked a multimodal data foundation for AI research.
Co-declared with partner Hangzhou Deepwise Biolink Technology, the project was shared as a case study by the China Hospital Association Information Professional Committee in October 2024. Its solution won first prize at the 6th National Smart Healthcare Innovation Competition, and platform research — multi-dimensional functional CT for precise diagnosis of digestive system diseases — earned the 2023 Henan Provincial Science and Technology Progress Award first prize.
The platform runs a distributed computing architecture with Kubernetes containerized deployment, Hadoop for storage and computation, Spark and Presto as core engines, and HDFS and Hive for large-scale data, integrating NLP, computer vision and multi-source heterogeneous data processing. It comprises four modules: an intelligent imaging data management system for storage, transmission, computation, application and governance across campuses; eight imaging AI applications — lung nodule, coronary, target reconstruction, rib, abdominal, head-and-neck, lower-extremity vessels and bone density, four with Class III medical device certification; an intelligent multimodal disease-specific data system governed by NLP and computer vision; and a no-code research system with pre-built ML and deep learning models plus large-model-assisted automatic annotation.
Operationally, the platform unifies deployment, scheduling and real-time monitoring of AI across campuses, easing pressure on PACS, devices and the network. It has managed more than 350,000 imaging studies, cleaned and governed data for over 13,000 patients and 200,000 patient-visit records, opened nearly 60 research accounts supporting 20+ projects — cutting project cycles from 10-20 months to under 6 — and supported four published papers with three more under submission, across all campuses.
The hospital plans to extend the platform toward data-asset operations and deeper medical-engineering integration, while flagging two ongoing challenges: mastering the intersection of AI, big data, IT and clinical medicine including large-model techniques, and building teams that combine AI development capability with medical expertise.