AITRICS to Publish a Paper on Early Prediction of Septic Shock in Sensors

2022.10.18

cover

Professor of Severance Hospital including Ji hoon Kim and Arom Choi conducting joint research Septic Shock early predictions through Accumulated Model [October 18, 2022] AITRICS (CEO Kwang joon Kim, Jin-gyu Yoo), a company specializing in artificial intelligence (AI) technology announced on the 18th that a paper of AITRICS's research team was published in Sensors, a prominent journal in the field of sensor and signal processing. This paper is 'Advantage of Vital Sign Monitoring Using a Wireless Wearable Device for Predicting Septic Shock in Febrile Patients in the Emergency Department: A Machine Learning-Based Analysis which is jointly studied by Severance Hospital Professor of Critical medicine Emergency medical services systems, Ji hoon Kim and A Rom Choi, and Kwan hyung Lee and Hee jung hyun of AITRICS research team. In this study, Research team developed and validated a machine learning model for early prediction of septic shock using wireless wearable devices and collected data from 468 patients who visited the emergency room with fever symptoms from July 2020 to June 2021.To this end, the research team compared the Accumulated Model, which learned the data collected from the patient's arrival time, with the Fragmented Model, which learned the data recorded at the closest point in the prediction time, and compared the results of learning the manual data recorded by the medical team at 1 hour intervals and the data collected from the wireless wearable device. As a result of the study, the area under ROC Curve (AUROC) of the data collected from the time of admission of the patient was 0.861, which is superior to the prediction accuracy of the manual data recorded by the medical team every hour (0.853). In addition, septic shock was predicted at least 5 hours and 30 minutes when data collected from the time of admission to the patient were used. As a result of the study, the area under ROC Curve (AUROC) of the data collected from the time of admission of the patient was 0.861, which is superior to the prediction accuracy of the manual data recorded by the medical staff every hour (0.853).In addition, septic shock was predicted at least 5 hours and 30 minutes when data collected from the time of admission to the patient were used. Based on these results, the research team confirmed that continuous vital signs monitoring using wearable devices can shorten the time to recognize potential deterioration in the patient's condition. Ji hoon Kim, Severance Hospital Professor of Critical medicine Emergency medical services systems, said, "This study is significant in that it can develop a machine learning model that predicts septic shock early by learning bio-signals collected in real time through IoT equipment." and "In addition, the use of IoT equipment suggests that it can reduce the burden on medical staff at busy sites where many patients need to be monitored at the same time and improve model performance by collecting real-time detailed information," he said. AITRICS researcher Hee-Jeong Hyun, who participated in the study, said, “There is a limitation in that wireless device data is collected at dense intervals in real time and noise caused by patient movement can occur frequently. In order to use this effectively, there were many technical considerations from data preprocessing. As a result, it is an honor to be recognized for its biosignal research capabilities in that it was able to effectively and early predict septic shock in emergency situations. We will do our best to provide better medical services to patients.” Meanwhile, AITRICS is a startup specializing in artificial intelligence technology composed of the best machine learning R&D personnel, and its technology is recognized worldwide with remarkable excellent results every year at various international artificial intelligence conferences such as NeurIPS and ICML.