Development of Real-Time Monitoring Clinical Decision Support System Utilizing Multimodal Data in the Emergency Department

Proposes the Potential of AI Algorithms for Early Prediction of Patient Deterioration

AITRICS, a medical AI, announced on the 23rd that its research paper on a multimodal deep learning model for emergency departments has been published in Scientific Reports, a journal under the prestigious Nature portfolio.

The study, titled "A Novel Deep Learning Algorithm for Real-Time Prediction of Clinical Deterioration in the Emergency Department for a Multimodal Clinical Decision Support System", builds upon research previously presented at MLHC 2023. It applies multimodal data processing techniques developed in earlier studies to real-world emergency department data, aiming to develop an AI-based Clinical Decision Support System(CDSS) model for use in emergency settings.

In collaboration with Professors Ji Hoon Kim, Arom Choi, and So Yeon Choi from the Digital Healthcare Innovation Center at Yonsei University Health System, the AITRICS research team conducted a retrospective study. The study analyzed 237,059 patient cases from Severance Hospital’s emergency department, using multimodal data, including vital signs, blood test results, and imaging data collected from electronic health records(EHR). The study evaluated the model's ability to predict acute deterioration events, such as in-hospital cardiac arrest, inotropic circulatory support, advanced airway management, and intensive care unit admission.

The results demonstrated that the emergency department deterioration prediction model developed by AITRICS showed high predictive accuracy by leveraging various unstructured data, including initial patient information. Moreover, the study found that incorporating continuous data collected from IoT devices further enhanced predictive accuracy compared to using static data alone.

Through this study, AITRICS highlighted the potential of AI algorithms in enabling early detection of patient deterioration. This achievement is expected to reduce uncertainty in clinical environments, enhance patient safety, and contribute to providing personalized medical care.

Sang-Cheol Han, a researcher at AITRICS, stated, “The deep learning model developed in this study achieved performance improvements by utilizing multimodal medical data and confirmed its potential as a universal clinical decision support system. In future research, we plan to optimize the prediction algorithm by expanding datasets and validating performance across multiple institutions.”

Professor Ji Hoon Kim from the Department of Emergency Medicine at Severance Hospital commented, “In emergency departments, gathering the necessary information to predict acute deterioration can be time-consuming, requiring medical staff to make quick and accurate decisions based on limited data. By utilizing deep learning technology, we have designed a novel algorithm applicable to various clinical applications. The prediction model developed through this study is expected to significantly aid medical staff by enabling real-time responses to changes in patient conditions.”