FEMH AI Research Team
Advancing Medical Artificial Intelligence at Far Eastern Memorial Hospital
🔬 Explore Our Research
Advancing Medical Artificial Intelligence at Far Eastern Memorial Hospital
🔬 Explore Our ResearchMeet the dedicated researchers driving innovation in medical AI
A leading medical institution committed to excellence in healthcare and research
Far Eastern Memorial Hospital has been serving the community for over three decades, providing comprehensive medical care with state-of-the-art facilities and cutting-edge technology. Our commitment to patient-centered care has made us one of Taiwan's most trusted healthcare institutions.
Our hospital is at the forefront of medical research. We collaborate with leading universities and research institutions to advance medical knowledge and improve patient outcomes.
The FEMH AI Research Team represents our commitment to integrating artificial intelligence into clinical practice. We focus on developing AI solutions that enhance diagnostic accuracy, streamline clinical workflows, and ultimately improve patient care quality.
Our latest contributions to the field of medical artificial intelligence
A novel framework that structures LLM reasoning to align with real-life clinician workflows using SOAP methodology. Our two-stage approach significantly outperforms baseline methods in both assessment accuracy and treatment plan quality.
A comprehensive multimodal framework for EHR data fusion that combines masked lab-test modeling with large language models to improve clinical prediction tasks and enhance healthcare decision-making processes.
An innovative mobile and web platform that leverages large language multimodal models for chronic disease risk prediction, providing accessible and accurate health assessment tools for both healthcare providers and patients.
A reinforcement learning framework that combines preference-driven reasoning refinement with task-adaptive templates aligned to clinical protocols. RGPO introduces groupwise ranking optimization based on the Bradley-Terry model with KL-divergence regularization, and shows consistent gains on PubMedQA, MedQA-USMLE, and real-world validation at FEMH, with a 2B-parameter model outperforming larger 7B-20B baselines.
A fully on-premise multi-agent system for one-year type 2 diabetes mellitus (T2DM) risk screening and guideline-grounded report generation from EHRs, combining calibrated risk prediction, deterministic clinical signal extraction, Reciprocal Rank Fusion over ADA guidelines, and a hybrid rule-based / LLM-entailment verification layer for auditable, privacy-preserving clinical decision support.
