AIDE (AI-Driven Emergency Care)

 

Category: *Precision Health & Smart Medical

Exhibitor: TAIPEI MEDICAL UNIVERSITY

Booth No: N416

 

Characteristic

 

Led by Professor Yung-Chun Chang from the Graduate Institute of Data Science at Taipei Medical University, our team brings together experts from emergency physicians from Shuang-Ho Hospital, AI engineers, and data scientists to form an interdisciplinary collaboration. Using nearly 300,000 real emergency department (ED) electronic medical records, we developed AIDE (AI-Driven Emergency Care), an intelligent ED system designed to meet real clinical needs. AIDE is the first decision support system tailored to the linguistic and clinical context of Taiwan’s emergency care, offering three major competitive advantages.
First, in terms of clinical integration and innovation, AIDE breaks through the limitations of traditional single-task prediction models. It enables multi-task prediction at the triage stage by integrating patients’ vital signs and symptom descriptions, including critical case identification, medical resource allocation, admission needs assessment, and return visit risk prediction. This effectively addresses challenges posed by limited ED resources and time constraints. Several core modules have been developed and published in five JCR Q1 international journals, covering applications such as patient disposition prediction (e.g., ICU admission risk and 72-hour unplanned return visits), medical interventions, and CT examination recommendations. The predictive models demonstrate strong performance, with AUROC scores ranging from 0.8 to 0.9, proving their accuracy, stability, and potential for clinical decision support.
Second, in terms of technological innovation, AIDE features highly original language processing and multimodal data fusion technologies. It is trained and optimized using a large volume of real clinical records, enabling contextual understanding specific to Taiwan’s medical environment. The system integrates unstructured chief complaints, structured examination data, and real-time information from wearable devices to build accurate, real-time predictive models—overcoming the limitations of current medical AI systems that typically rely on a single data type. One of the most notable breakthroughs is our self-developed CNTR (Clinical Narrative Text Representation), which significantly enhances language models’ semantic stability and understanding of short, noisy medical texts. This makes the models more aligned with clinical language and physicians’ actual needs, while also showing promise for multilingual and cross-institutional applications—laying a strong foundation for localized and scalable medical language processing.
Third, AIDE offers end-to-end support across the entire patient journey, from admission to discharge. It can initiate risk assessments immediately upon patient arrival, quickly identifying high-risk cases and providing recommendations for moderate- to low-risk patients, thereby reducing unnecessary testing, improving operational efficiency, shortening wait times, and lowering the rate of patients leave without being seen (LWBS). For rare but high-risk conditions such as stroke and sepsis, where data imbalance is a common challenge, the system is specifically designed to enhance identification sensitivity and model robustness under extreme sample distributions.
By integrating clinical decision support, natural language question-answering, and automated summary generation, AIDE goes beyond a single-point technology to become a system-level innovation that brings comprehensive intelligence to emergency care. It has already begun prospective clinical validation at Shuang-Ho Hospital, demonstrating its feasibility and scalability for real-time clinical deployment. In addition, the CT recommendation model has been successfully validated at Shin Kong Wu Ho-Su Memorial Hospital, achieving an AUROC of 0.87, confirming its cross-institutional reliability and applicability—laying the groundwork for nationwide deployment.
AIDE also holds significant market potential and strong technical protection, with core technologies patented in Taiwan (114103684) and the U.S. (19/037,346). The system can be extended to other high-acuity, outpatient, and inpatient settings, forming a modular framework for scalable promotion and commercialization, licensing, and international export. More than a single AI model, AIDE is an integrated innovation platform combining original language technology, clinical process insight, and system design. It is one of the few AI systems in Taiwan capable of frontline ED deployment with full workflow support. AIDE demonstrates high feasibility, originality, and clinical impact, solidifying our team’s leadership in medical AI and showcasing a significant competitive edge in this domain.

 

Other Products

 

• 

  multi-parameter physiological monitoring device

• 

  MTAM (Microtube Array Membrane)

• 

  Annulus Fibrosus Allograft

• 

  SYNOMICS® Gut Microbiome Analysis

• 

  NeuroSwift Pro-Medical Decision Support System for Brain Diseases

• 

  EG BioMed Multi-Cancer Blood Testing Platform

• 

  Armed-T cell therapy

• 

  Tumor microenvironment chip, TME-Chip

• 

  Smart Innovation Platform for Preclinical Drug Development

 

 

 

Back