Xie Lab develops artificial intelligence and machine learning methodologies for complex scientific and healthcare problems, with an emphasis on integrating data, physical knowledge, and domain expertise. Our current research directions include:
- AI Agents for Scientific Discovery and Healthcare. Developing domain-specialized AI agents that can reason over multimodal data, interact with computational models and scientific tools, and support complex workflows in biomedical research, simulation, and clinical decision support.
- Physics-Informed Machine Learning and Scientific AI. Designing learning frameworks that incorporate physical laws, mechanistic models, and domain constraints to improve interpretability, robustness, and generalizability, particularly for inverse problems and computational modeling.
- Cardiac Digital Twins and Computational Cardiology. Developing patient-specific cardiac digital twins for modeling cardiac electrophysiology, conduction systems, electromechanics, and heart function using point-cloud representations, medical imaging, physiological signals, and physics-informed learning.
- Biomedical Imaging and Multimodal AI. Developing deep learning methods for medical image analysis and integrating imaging with physiological signals, clinical data, and other modalities for disease characterization, risk assessment, and personalized healthcare.
- Physiological Signal and Time-Series Intelligence. Developing machine learning methods for ECG, blood glucose, and other longitudinal physiological signals, including representation learning, prediction, risk modeling, and extraction of clinically meaningful digital biomarkers.
Active, Adaptive, and Data-Efficient Learning. Developing active learning, sensor selection, uncertainty-aware learning, and other data-efficient AI approaches to determine what information should be collected, where measurements should be made, and how limited labeled or sensing resources can be used most effectively.
These methodological advances are applied across cardiovascular disease, diabetes, medical imaging, and other biomedical and engineering domains, with the broader goal of creating intelligent computational systems that connect data, physics, and human expertise to advance scientific discovery and human health.
Research interest tags: AI agents for science and healthcare, physics-informed machine learning, scientific AI cardiac digital twins, multimodal biomedical AI, medical image analysis, physiological signal processing, ECG and cardiovascular AI
