BACKGROUND:
Physicians often face challenges including fragmented medical information sources, the burden of processing unstructured data, and a mismatch between AI tools and clinical needs, necessitating intelligent systems for direct decision support.

OBJECTIVE:
To develop and optimize a city-level intelligent medical decision-making platform for real-world clinical, enhancing the usability, relevance, and safety of AI in clinical workflows to improve comprehensive disease management and healthcare accessibility.

METHODS:
The platform utilizes natural language processing and semantic understanding technologies for accurate intent recognition of clinical queries. It employs an autonomous model scheduling mechanism for on-demand invocation of both general and specialized large language models. The medical city brain deeply integrates intelligent recognition and parsing capabilities for medical imaging and clinical documents, enabling automatic extraction and comprehensive analysis of unstructured data from imaging reports, test results, and progress notes. Through application management and session management modules, it facilitates unified deployment, operational monitoring of clinical AI applications, and full traceability of doctor-patient interactions. Its knowledge base merges local high-quality medical databases with authoritative medical knowledge graphs, covering the entire spectrum of disease prevention, control, diagnosis, and treatment, and supports personalized treatment planning.

RESULTS:
The platform reduced clinician time spent on information retrieval by 50%, allowing them to focus on core clinical judgment. It significantly enhanced AI-assisted decision-making across specialties by providing evidence-based, context-aware support. The system achieved automatic multimodal data integration to construct precise patient datasets, establishing a robust decision support framework that aids in diagnosis and personalized planning. This reduces treatment burdens, and serves citizens through more accessible and efficient healthcare.

CONCLUSIONS:
The platform demonstrates significant value in enhancing clinical decision support, promoting the practical implementation of medical AI, and optimizing workflows. It aids in improving early diagnosis rates, reducing treatment burdens, and offers a replicable model of smart healthcare for primary care institutions.