What is Futang·Baichuan
Futang·Baichuan is the world's first pediatric large model jointly released by the National Center for Children's Health, Beijing Children's Hospital Affiliated to Capital Medical University, Baichuan Intelligence, and Xiao'erfang Health. It covers a three-dimensional knowledge system of common and complex pediatric diseases, possesses strong pediatric clinical reasoning capabilities, and pioneers a pediatric "evidence-based model," capable of integrating the best medical evidence like a professional pediatrician to formulate scientific, personalized diagnosis and treatment plans for pediatric patients. The model is based on trillion-level token professional medical data, covering Chinese and English professional medical papers, medical guidelines, textbooks, and books, and integrates the clinical experience of over 300 renowned pediatric experts from Beijing Children's Hospital and decades of high-quality de-identified medical record data.
Based on the model, two applications have been launched: the "Futang·Baichuan" AI pediatrician primary care version and the expert version. The primary care version is designed for daily pediatric disease diagnosis and treatment scenarios, and its diagnostic capability has reached the level of an attending physician at Beijing Children's Hospital, helping to improve the pediatric diagnosis and treatment capabilities of primary care doctors. The expert version focuses on the diagnosis and treatment of complex and rare pediatric diseases. It "took up its post" at Beijing Children's Hospital in February 2025, and the consistency rate between its diagnosis and treatment plans and expert consultation results reaches 95%.
Main functions of Futang·Baichuan
- Three-dimensional knowledge system: Covers common and complex pediatric diseases. Integrates comprehensive medical knowledge such as Chinese and English professional medical papers, medical guidelines, textbooks, and books. Integrates the clinical experience of over 300 renowned pediatric experts from Beijing Children's Hospital and decades of high-quality de-identified medical record data.
- Pediatric clinical reasoning capability: Capable of integrating the best medical evidence like a professional pediatrician. Formulates scientific, personalized diagnosis and treatment plans for pediatric patients. Enhanced in evidence-based reasoning, memory, dialogue, and multimodality.
- Pediatric "evidence-based model": Pioneers a pediatric "evidence-based model." Provides well-founded answers through a "present facts and reason things out" approach. Self-built highly credible evidence-based medicine knowledge base, including over 40,000 Chinese and English guidelines and expert consensus documents, over 38 million scientific research papers, and nearly 170,000 drug instructions.
- Multi-round consultation and communication: When facing patients' parents, it can autonomously conduct multi-round consultations and communicate patiently. Through the condition inquiry function, it forms a preliminary diagnosis and gives examination recommendations.
Technical principles of Futang·Baichuan
- Artificial intelligence technology: The "Futang·Baichuan" pediatric large model uses the latest artificial intelligence technologies, natural language processing (NLP) and machine learning (ML). It enables the model to understand and process natural language data, and learn and extract useful information from it. The model also possesses strong pediatric clinical reasoning capabilities, capable of integrating the best medical evidence like a professional pediatrician.
- Data sources: The model's data sources are very extensive and authoritative, including:
- Chinese and English professional medical papers: Covers authoritative domestic and international pediatric clinical guidelines and scientific literature.
- Medical guidelines and textbooks: Includes standard treatment guidelines and textbooks in the field of pediatrics.
- Expert clinical experience: Integrates the clinical experience of over 300 renowned pediatric experts from Beijing Children's Hospital.
- High-quality medical record data: Uses decades of de-identified high-quality expert medical record data.
- Model training methods: The training of the "Futang·Baichuan" pediatric large model adopted the following methods:
- Training on massive professional medical data: A professional medical database with trillions of tokens was constructed.
- Integration of clinical experience: Experts' clinical experience was combined with model training.
- Multidimensional technical enhancement: The model was strengthened in five aspects: evidence-based medicine, reasoning, memory, dialogue, and multimodality.
- Self-built evidence-based medicine knowledge base: It includes more than 40,000 Chinese and English guidelines and expert consensus documents, more than 38 million scientific research papers, and 170,000 drug instructions.
Specific applications of Futang·Baichuan
- AI pediatrician primary care version：
- It is aimed at daily pediatric disease diagnosis and treatment scenarios.
- Taking the diagnosis of viral encephalitis in children as an example, the AI pediatrician can form a preliminary diagnosis and give examination recommendations.
- Its outpatient case diagnosis capability has reached the level of attending physicians at Beijing Children's Hospital.
- Pilot applications will be carried out in national regional medical centers, pediatric medical alliances, county-level tertiary hospitals, community hospitals, and other institutions.
- AI pediatrician expert version：
- It focuses on the diagnosis and treatment of difficult and rare pediatric diseases.
- It began working at Beijing Children's Hospital in February 2025.
- It has participated in more than 10 multidisciplinary consultations for difficult and rare diseases, pediatric grand rounds, and other activities.
- The consistency rate between its diagnosis and treatment plans and the consultation results of experts at Beijing Children's Hospital reached 95%.
Application scenarios of Futang·Baichuan
- Online consultation: Patients can consult the AI pediatrician through an online platform and obtain preliminary diagnoses and recommendations.
- Remote consultation: It supports remote multidisciplinary consultations, and experts can use AI assistance for diagnosis and treatment decisions.
- Medical education and training: It provides training and educational support for primary care doctors and medical students. Through simulated cases, it helps primary care doctors and medical students conduct clinical reasoning training. It provides the latest pediatric medical knowledge and guidelines to help doctors and medical students update their knowledge systems in a timely manner.
- Public health and disease prevention: It supports public health departments in disease surveillance and prevention work. Through big data analysis, it monitors the epidemic trends of pediatric diseases and provides prevention recommendations and health guidance for common childhood diseases.
- Scientific research and data analysis: It supports scientific research in the pediatric field and provides data support and analysis tools. It analyzes large amounts of pediatric case data to uncover potential disease patterns and treatment effects. It provides data support for researchers and assists in the implementation of scientific research projects.