In the respiratory department clinic of Haihe Hospital in Tianjin, Deputy Chief Physician Wang Herong is treating a patient who has been coughing for two weeks. After she entered the words "cough, fever" on the computer, an intelligent analysis panel popped up on the right side of the screen, showing that the patient's CT report from 3 months ago was automatically retrieved and labeled as mild pulmonary fibrosis lesion in the lower lobe of the right lung; Recently, inflammatory indicators such as elevated C-reactive protein and abnormal white blood cell count have been highlighted in the blood routine; The initial draft of the structured medical record lists differential diagnoses such as "community-acquired pneumonia" and "acute exacerbation of chronic obstructive pulmonary disease" according to priority, and includes examination recommendations and medication references.
Wang Herong said that in the past, when accessing images, it was necessary to remember the 3-level operation path of the PACS system, manually retrieve medical history by logging into the EMR system, and switch to the LIS system to verify drug sensitivity results before prescribing; Now, AI will actively push information based on diagnosis and treatment scenarios, saving doctors at least 5 minutes.
Behind these 5 minutes saved, lies the powerful intelligence of the AI native hospital system created by the Tianhe solution for AI native hospitals.
Kang Bo, the program leader and general manager of Tianjin Zhilin Tianhe Technology Co., Ltd., introduced that when doctors input keywords such as "cough" and "fever", the natural language processing engine of the AI native hospital system can parse the clinical intentions behind the keywords, just like "understanding" the unspoken needs of doctors; Multi modal data fusion technology synchronizes and connects scattered data such as imaging, testing, and medical records, actively aggregates them to form a complete patient information chain, and ends the dilemma of "doctors looking for data".
In traditional medical scenarios, data is a static "file" that needs to be manually queried, but driven by the "AI brain", data will dynamically flow with the diagnosis and treatment scenario. Kang Bo gave an example that in the intelligent assisted diagnosis and treatment scenario based on AI native hospital systems, when doctors consider adjusting antibiotics, the system will automatically associate three pieces of information: the patient's sputum culture drug sensitivity results, past drug allergy history, and hospital antibiotic management standards. After verification, medication recommendations will be generated to solve the problem of fragmented and lack of collaboration in the intelligent process from the source.
In addition, edge computing power is responsible for millisecond level feedback after keyword input, making the smart panel "on call"; Local computing power is responsible for supporting deep fusion and inference of multimodal data, ensuring accurate diagnostic recommendations.
According to data provided by Haihe Hospital in Tianjin, three months after the implementation of intelligent assisted diagnosis and treatment scenarios, the daily average number of outpatient doctors' visits has significantly increased, and the completeness of medical records has greatly improved.
During the midday doctor fatigue period, AI real-time reminders significantly reduce the omission rate of key medical histories and significantly lower the risk of missed diagnoses. "Yang Wanjie, the director of Haihe Hospital in Tianjin, said.
