Research

My research focuses on AI for Healthcare (AI4Healthcare), with representation learning and 3D perception as the main algorithmic foundations. I am interested in learning robust, transferable representations from medical data and developing spatially aware models that can support clinically meaningful analysis and decision making.

Current Themes

AI4Healthcare

I develop learning-based methods for healthcare problems, especially tasks involving medical images and clinically relevant data. My goal is to connect algorithmic innovation with reliable evaluation and practical medical value.

Representation Learning

I study how models can learn robust and transferable representations from limited, heterogeneous, or weakly annotated medical data. My recent work in computational pathology, PET synthesis, and ultrasound provides the application foundation for this direction.

3D Perception Algorithms

I am interested in algorithms that reason about volumetric structure and spatial relationships in medical data. This direction includes 3D visual perception, geometry-aware representation learning, and efficient modeling for complex spatial inputs.

Representative Topics

• Computational pathology and multiple instance learning
• PET image synthesis with learned priors
• Ultrasound and photoacoustic-ultrasound fusion imaging
• Robust and transferable representation learning
• 3D visual perception and volumetric medical data understanding
• Reliable AI systems for healthcare applications

Selected Publications

• MoMIL: Multi-order enhanced multiple instance learning for computational pathology
• VQPET: Leveraging Vector-quantized Codebook Prior for PET Image Synthesis
• Research Progress and Clinical Translation of Photoacoustic-ultrasound Fusion Imaging in Breast Cancer Diagnosis and Therapy
• 多模态基础模型与超声影像的交叉应用
• Mamba2MIL: State Space Duality Based Multiple Instance Learning for Computational Pathology

Methods and Tools

• Python, PyTorch, scikit-learn, OpenCV
• Representation learning, 3D perception, and medical image analysis pipelines
• Research workflows for Jekyll/GitHub Pages, LaTeX, and academic writing

I am most motivated by research that combines strong algorithmic foundations with clear healthcare relevance.