NEXUS-AI is a non-invasive, single-time-point vital signs AI for early sepsis warning. Using only heart rate, oxygen saturation, body temperature, blood pressure, and respiratory rate, it provides alerts six hours before clinical diagnosis. It overcomes the reliance of current models on multimodal data from medical records. Trained on multicenter Taiwan data, it outperforms qSOFA, has been deployed at Taipei Veterans General Hospital, and has obtained QMS certification and Taiwan patents.

NEXUS-AI overcomes the reliance of current AI-based sepsis models on multimodal data by using non-invasive, single-time-point measurements. Requiring only five vital signs, it predicts six-hour sepsis risk. Trained on U.S. PhysioNet and four Taiwanese hospital datasets, it outperforms SIRS and qSOFA. SHAP analysis shows clinically meaningful features significantly associated with CRP, supporting real-time, interpretable clinical use.

This technology supports early sepsis detection, clinical deployment, and home-based care. NEXUS-AI uses a modular Docker architecture to connect with HIS and has been deployed at Taipei Veterans General Hospital for hospital-wide early warning. It can also integrate with wearables and mobile apps for long-term care and home care. Clinically, it can combine with POCT biomarker testing for early precision detection and is being commercialized with Chunghwa Telecom.