Research Projects & Code
Research Projects
A collection of my research projects, code repositories, and open-source contributions focused on digital pathology, representation learning, and histopathology image analysis.
MSRL
Histopathology-Genomics Multi-modal Structural Representation Learning
A framework for learning from paired whole-slide images (WSIs) and genomics data with missing-modality inference. Features graph structure learning, multi-modal fusion, and dynamic graph construction during fine-tuning.
PAMA
Position-Aware Masked Autoencoder for WSI Pre-training
Pan-cancer histopathology WSI pre-training method using position-aware masked autoencoder. Incorporates spatial information for improved representation learning on large-scale histopathology images.
Lifelong WSI Retrieval
Continual Learning for Histopathology Image Retrieval
Framework for lifelong/continual learning in WSI retrieval tasks. Features distance consistency rehearsal and coreset selection to mitigate catastrophic forgetting while learning from new data streams.
Digital Pathology Cross-modal Retrieval
Digital Pathology Vision-Language Cross-modal Retrieval
Vision-language representation learning combining histopathology images and medical text. Enables fine-grained cross-modal retrieval and improved zero-shot diagnosis capabilities.
Research Topics
My research focuses on several key areas in digital pathology and multi-modal biomedical data analysis:
- Representation Learning: Self-supervised learning methods for gigapixel histopathology images
- Multi-Modal Fusion: Integrating whole-slide images with genomics, clinical data, and text
- Graph Neural Networks: Structure learning and dynamic graph construction for patient stratification
- Foundation Models: Leveraging pre-trained models for histopathology feature extraction
- Continual Learning: Lifelong learning systems for evolving medical datasets
- Precision Oncology: Applications to cancer diagnosis, prognosis, and treatment prediction
Getting Started
Most projects include detailed README files with installation instructions, usage examples, and citation information. For questions or collaborations, feel free to reach out via email.