CRCFound
A self-supervised CT foundation model for diagnosis, staging, molecular prediction, and prognosis in colorectal cancer.
- 5137 unlabeled CT scans
- 8 downstream tasks
- Strong transfer under limited labels
Research Program
Research
Foundation models at the core of a broader translational AI stack
We build clinically grounded AI systems for colorectal cancer, with current flagship work centered on foundation models for imaging, multimodal prognosis, and cross-platform molecular stratification. Around that core, the lab continues to extend into explainable multi-omics modeling, surgical and longitudinal imaging intelligence, and the clinical data programs and research platforms that make translation possible.
Core Program
Foundation Models for Colorectal Cancer
Recent flagship systems span imaging pretraining, multimodal prognosis, and robust transcriptomic stratification. This is the current center of gravity in the lab's research portfolio.
Brim
An interpretable bridged multimodal fusion model that links histology and molecular data for pan-cancer prognosis.
- 12 cancer types
- Robust to missing modalities
- Histology-to-genomics bridging
PIANOS
A normalization-free single-sample classifier for robust colorectal cancer risk stratification across platforms.
- 24 cohorts / 5439 patients
- Single-sample and cross-platform
- Outperformed 105 published models
Supporting Direction
Translational Omics & Clinical Stratification
This line carries the lab's work on subtype discovery, multi-omics representation learning, and clinically interpretable molecular modeling.
DeepCC
A pathway-informed deep learning framework for robust cancer molecular subtype classification.
- Pathway-informed ANN
- Single-sample prediction
- Robust to missing data
TMO-Net
An explainable pretrained multi-omics model for transferable oncology prediction and pathway-level interpretation.
- Transferable multi-task learning
- Pathway-level interpretation
- Precision oncology workflows
Supporting Direction
Surgical & Imaging Intelligence
From sparse-annotation learning to longitudinal response prediction and 3D segmentation backbones, these projects support intervention-facing perception and decision pipelines.
SegMamba-V2
A general 3D medical image segmentation backbone for long-range volumetric modeling across organs and modalities.
- CRC-2000 benchmark
- Tri-orientated spatial Mamba
- 4 large-scale datasets
3D RP-Net
A longitudinal multi-task MRI model for treatment response prediction in rectal cancer.
- 2568 MRI scans
- AUC 0.95 / 0.92
- Segmentation plus response prediction
SOUSA
A unified weakly and semi-supervised segmentation framework that learns from sparse labels and unlabeled medical images.
- Teacher-student consistency
- Scribble supervision
- Multi-angle reconstruction loss
Program Layer
Clinical Programs & Data Infrastructure
Large clinical programs and cohort-building efforts provide the translational substrate behind the lab's algorithmic work.
ICGC-ARGO CRC
A large-scale colorectal cancer clinical genomics program that anchors translational AI and biomarker development in China.
- Large CRC cohort in China
- Multi-omics and pathology integration
- Platform for translational validation
MedAutoScience
An agent-first, human-auditable medical research platform for progressing disease-focused studies from versioned data assets to evidence packaging and submission-ready deliverables.
- Study progression and gating
- Auditable data asset management
- Evidence packaging and submission delivery