In Taiwan, there are 287 deaths per 100,000 sepsis patients and the mortality rate is up to 29.2%. Since sepsis is also called the “silent killer” for its uncertain signs and symptoms as well as its diverse potential causes of pathology, the differential diagnosis of sepsis is difficult. Superintendent Der-Yang Cho, China Medical University Hospital, has focused on the issue of growing threats due to super bacteria and teamed up with the following departments: the Department of Pharmacy, Department of Laboratory Medicine, Department of Medical Research, Department of Infectious Disease, the Information Center, the AI Center for Medical Diagnosis and Big Data Center, etc., to integrate the AI power of the AI Innovation Center and develop the cross-department AI-assisted resistant bacterial identification and treatment strategy system. The system has won cross-territory recognition and awards for excellence, including the 2021 Bio-driven City award.  

Sepsis is an infectious syndrome with a high mortality rate. Newly published data in a famous medical journal, The Lancet, which was conducted by a research institute in the US and showed updated statistical data from 196 countries worldwide, revealed that one out of five deaths are due to sepsis. According to research results from Taipei Municipal Wanfang Hospital, there were 643 newly developed sepsis cases among every 100,000 people in Taiwan. Superintendent Der-Yang Cho mentioned that, traditionally, physicians would base their approach on their personal experience when prescribing empirical antibiotics for hospitalized patients who manifested the infection so there is a certain proportion of patients who are not correctly diagnosed at the beginning of the infection, resulting in the golden treatment time being missed. 

Dr. Chin-Chi Kuo, Chief of the Big Data Center, said that as the condition of patients in emergency rooms are changing quickly, it’s critical to apply big data analysis to differentiate bacterial infections. How to leverage big data and use the right antibiotics to reduce the risk of sepsis for patients is one of the functions of the AI-assisted resistant bacterial identification and treatment strategy system platform. Such a cross-department platform can combine antibiotic selection advice, drug resistance predictions and sepsis prevention approaches. Three major functions have been introduced in all medical care systems of CMUH. In the future, it’s expected to be promoted to other hospitals to improve the quality of overall medical care for patients suffering from sepsis.

Dr. Edward Hsu, Chief of the AI Center for Medical Diagnosis, mentioned that by applying advanced learning techniques to develop the AI-assisted medical diagnosis system for sepsis, physicians can identify sepsis through the system and elevate the patient survival rate. The AI Center for Medical Diagnosis addressed the system in the research article published on CIBB 2021. The system is designed to predict the probability of sepsis occurring in patients by comparing gender, age and physiological indicators, as well as blood sample data over three days of hospitalization from big data references established in the past. If the system sounds, the probability of sepsis complications is high, and the medical care team may proactively enhance the care provided and physiological data monitoring. In emergencies, the medical care team can provide antibiotics and fluid infusion treatment on time. The system can also provide treatment advice based on the damage level of the organs. According to the preliminary trials, the accuracy of the CMUH-trained AI prediction rate is near 90%.     

Superintendent Der-Yang Cho said that the current clinical process of medical tests needs 3-5 days to provide results for an antibiotics treatment assessment after the bacterial strain differentiation and antimicrobial susceptibility testing has been implemented. Superintendent Der-Yang Cho and the AI Innovation Center team led by Chief Jia-Xin Yu have initiated the project to speed up the whole testing process and consolidate it into one machine which will be able to manage tasks that previously were executed with two machines and to improve the antibiotics application for reasonable use. Also, the “AI fast predicting system for resistant bacterial strains” is referenced in the big data of the bacterial strain differentiation and resistance reports in the past and has been linked to mass spectrometer information to identify the bacteria species and predict if they possess resistant protein. The whole process would be shortened to one hour to allow physicians and patients to ensure the right drugs are used to reduce medical costs and mortality rates.   

CMUH has collaborated with Ever Fortune. AI CO., Ltd. to prepare to enter the markets aggressively, including to get US FDA and TFDA license approvals. Dr. Chia-Hui Chou from the Infection Department, CMUH, said that the system has combined wearable devices for physiological monitoring and smart wards. It also possesses the potential to develop remote medical care in the future. The system can help small hospitals and hospitals in remote areas with manpower insufficiency issues to elevate the efficiency of examinations and decrease the mortality rate of sepsis patients. It also helps alleviate long waits of patients in crowded emergency rooms.