2026/06/04 12:14
Huang Chongyao (right), director of the Applied Medicine Department of the Department of Digital Medicine of Mackay Memorial Hospital, and Li Guangshen (left), professor and senior consultant physician of the Department of Biotechnology and Medical Sciences, jointly announced the results of smart medical and generative AI applications at the GTC Taipei 2026 venue. (Provided by Mackay Hospital)
[Reporter Luo Bi/Taipei Report] Generative AI has developed rapidly and has gradually entered the medical field. However, the focus of the medical community is no longer just whether AI can answer questions, but whether the answers are correct and credible. Mackay Memorial Hospital recently announced at the GTC Taipei 2026 conference hosted by NVIDIA that the "EBM Guardian (Empirical Medicine Assistance)" system developed by the hospital uses AI to combine evidence-based medicine with a multi-layer verification mechanism to help doctors quickly search for literature and organize research evidence, while reducing the risk of erroneous information generated by AI.
Zhang Wenhan, chief director of Mackay Memorial Hospital, said that clinicians face a large amount of medical information every day and must make judgments in a short time based on the patient's condition. However, the complete evidence-based medicine process, from literature search, data screening to analysis and compilation, often takes a lot of time. The purpose of developing EBM Guardian is to use AI to help doctors grasp medical evidence more quickly, reduce repeated search and review time, and allow the medical team to devote more attention to patient care.
Huang Chongyao, director of the Department of Applied Medicine, Department of Digital Medicine, Mackay Memorial Hospital, who has obtained Huida's "Generative AI Large Language Model Professional Architect (NCP-GENL Certified)" certification, pointed out that although generative AI has powerful language generation capabilities, it may also produce so-called "hallucination", producing content that seems reasonable but is actually wrong or unfounded. If medical decision-making relies solely on AI answers and lacks empirical data and verification mechanisms, it may lead to incorrect judgments.
He said that EBM Guardian does not simply allow AI to generate answers, but establishes a clinical auxiliary process that is based on medical evidence, traceable to sources, and can be verified. The system can help doctors break down clinical problems into search structures such as patient conditions, treatment methods, comparison items, and expected results, actively search medical databases such as PubMed, and organize important research and literature highlights.
Huang Chongyao said that in the past, it might take hours or even days to complete a set of empirical medical analysis processes, but clinical sites often have to make judgments within minutes. AI can help doctors grasp important evidence faster. In addition, the system also introduces a multi-layer verification and AI audit mechanism. After the first set of AI generates results, it will be verified by another set of AI to reduce the risk of quoting wrong information or making wrong inferences.
He also pointed out that the system mainly operates in the hospital environment, preventing sensitive medical data from being directly transmitted to the external cloud, and taking into account information security while promoting AI applications.
In addition, Li Guangshen, professor and senior consultant physician at the Department of Biotechnology and Medical Sciences of Mackay, also shared the application and development of AI in the fields of biomedical research, cell research and precision medicine. He hopes that through cross-field cooperation, AI technology can be actually introduced into clinical practice to help solve pain points at the medical site.
Mackay Hospital stated that in the future, AI will not only be a clerical tool, but will also have the opportunity to become an important supporting role in clinical care, improving medical efficiency and care quality.
