Test reports are essential for doctors to make accurate diagnoses, but they can be difficult to interpret. For patients, understanding the complex numbers and metrics in these reports can be confusing and anxiety-inducing. Even for doctors, mastering the interpretation of numerous metrics within their own field of expertise requires extensive training. But, with dozens or even hundreds of test reports to read every day, the task can be overwhelming.
Typical medical services at a hospital involve two steps: diagnosis and treatment, with testing being a crucial part of the diagnostic process.
-- Hu Zhenning,
Director and Deputy General Manager
of Runda Medical Technology
Runda Medical Technology is China's largest in vitro diagnostics (IVD) solution provider, offering a wide range of IVD products, technical support, and operations support for different types of medical labs. Its information systems are connected to over 2,000 hospitals.
The Huawei Cloud Pangu Medical Model is pre-trained on high-quality, unstructured data from 16 million academic journals, as well as knowledge graphs that incorporate over 1 million high-quality examples from clinical practice (diseases, medicines, operations, etc.). Through further supervised fine tuning (SFT), the model acts as an intelligent assistant in medical testing, diagnosis, and personal health management.
Powered by the Huawei Cloud Pangu Medical Model, the Runda AI Medical Assistant builds the world's largest medical testing knowledge graph.
-- Zhang Nan,
CTO of Runda Medical Technology
Trained on massive datasets, the Runda AI Medical Assistant can explain more than 4,500 test items and 2,800 diseases. In answering questions based on clinical report interpretation of real cases from 10 different departments, it achieved an overall accuracy of 87.74%, comparable to specialized human doctors. In clinical practice, the assistant considers patients' medical histories, symptoms, and other relevant factors, providing comprehensive evaluations and treatment suggestions, helping doctors and patients in choosing the best treatment options. It also helps in identifying rare diseases.
Take adrenal insufficiency in the elderly as an example. This disease is easy to miss in preliminary diagnosis as it is rather rare and its symptoms are quite similar to cardiovascular diseases and respiratory infections. Early detection is crucial for successful treatment. The Runda AI Medical Assistant has successfully helped doctors diagnose this rare disease by analyzing patients' sodium and potassium levels in their blood while ruling out the administration of adrenocorticosteroid drugs. In the three months since its deployment by the Second Affiliated Hospital of Sun Yat-sen University, this smart assistant identified 11 high-risk patients from 4,268 people. These patients were then transferred to the endocrinology department, where they were confirmed with adrenal insufficiency. All 11 patients have now recovered after treatment.