In the wave of artificial intelligence (AI), Taichung Veterans General Hospital (TVGH) began training medical staff at the AI Center of Tunghai University in 2017. This collaboration led to the development of the "Zoe Smart ICU System," which not only predicts patients' conditions but also selects the most urgent treatment targets and methods, reducing the workload of doctors and nurses. The system has even decreased the amount of medication used. These achievements are gradually being commercialized and extended to other hospitals, drawing international attention.
The system features 11 newly developed software applications that can be used for predicting acute kidney injury, classifying acute respiratory distress syndrome, predicting extubation success, managing pain, assessing nutritional risks, predicting bacteremia risks, and even estimating the risk of death 24 hours later. More than 20 papers have been published on these innovations, and the system has received over 10 patents in Taiwan and the United States. It also won a Silver Award at the National Biotechnology and Medical Quality Awards, ranking among the best in Asia.
Wu Chieh-Liang, the Deputy Director of TVGH and a thoracic physician, shared that after working in the ICU for over 20 years, he observed several challenges, such as the complexity and rapid changes in patients’ conditions. Constant, round-the-clock care and observation are necessary, and quick access to information and decision-making is critical.
However, the volume of data generated by 24-hour monitoring is vast, and it is crucial to quickly interpret and organize the data while being alerted to the most pressing issues, such as heart, respiratory, kidney, or infection problems, or even patient agitation. These are challenges the ICU faces daily. With TVGH’s extensive data resources and the rising trend of AI in 2017, Wu and four other clinical staff from different specialties were sent to Tunghai University’s AI center to learn the latest AI techniques.
He then set out to address common ICU issues using AI, collaborating across departments and receiving funding from the National Science and Technology Council. Working with Tunghai University, Advantech, and Huide Technology, the team utilized the hospital’s database to develop 11 predictive software applications.
Wu explained that AI has provided the ICU at TVGH with the equivalent of 11 capable assistants. As a result, average hospital stays have decreased, complications have been reduced, the accuracy of ventilator settings has improved, and the use of sedatives has decreased. Dietitians and nurses have reported that these innovations have been helpful, especially for new ICU staff. The data provided by the system is effective in patient care, although the clinical benefits are still undergoing evaluation.
Although AI has transformed the ICU, Wu emphasized that due to the increasing shortage of medical staff in ICUs, the development and application of these tools have become even more urgent. He believes that this technology should not only be used at TVGH but should also be adopted by other hospitals.
Looking ahead, Wu shared that TVGH is planning to integrate large language model technologies with the system. For example, asking the system, “How should this patient be cared for?” will provide immediate answers. This requires advanced equipment and technology and relies on strong government support.
Resource: 中榮加護病房AI化 能預測病情 減少用藥與住院天數
