CV
Yuxuan Yang · 杨宇轩
Ph.D. student · AI for Science · HIT Shenzhen / SLAI
Email ↗Google Scholar ↗GitHub ↗
Education
Harbin Institute of Technology, Shenzhen 
SLAI
Ph.D. student in Computer Science · Joint training
- Advisor at HIT Shenzhen: Prof. Yongbing Zhang (张永兵).
- Advisor at SLAI: Prof. Zhiguo Zhang (张治国).
Beijing Jiaotong University
M.Eng. in Electronic Information
Beijing Jiaotong University
B.Eng. in Information Engineering
Research Interests
AI for Science; foundation models; world models; multimodal learning. Current research settings include biological and biomedical systems.
Research & Industry Experience
United Imaging Healthcare
Algorithm Research Intern · Central Research Institute, LLM Lab
- Led work on image-grounded multimodal medical report generation.
- Contributed to head CT report generation, including image–text data preparation and instruction-tuning design.
Selected Publications
Full list on Scholar ↗Cancellous bone segmentation network in cone beam CT images for post-orthognathic assessment of condylar resorption
Dentomaxillofacial Radiology, 2026 · 55(3): 267–275
Paper DOI: 10.1093/dmfr/twaf081
PolarDETR: Enhancing Interpretability in Multi-modal Methods for Jawbone Lesion Detection in CBCT
MICCAI, 2025 · Medical Image Computing and Computer Assisted Intervention
Paper DOI: 10.1007/978-3-032-04927-8_48
Manuscripts Under Review
Decoupling Shape Priors from Boundary Details for Segmentation in Ambiguous Medical Images
DSB-Net decouples shape-prior learning from boundary-detail refinement for segmentation in ambiguous medical images.
Selected Projects
Whole-Slide Image Encoding
Submitted to ISBIResearch project
Visual representation learning and image encoding for whole-slide pathology images. The manuscript is under review.
Spatial Evidence
In progressResearch project
Ongoing research on learning from spatial molecular measurements.
Virtual Cells & Cell Foundation Models
ExplorationResearch direction
Exploring models of cellular states and responses to perturbations, including participation in a team-based virtual-cell modeling competition.
DSB-Net: Shape–Boundary Decoupling
ICASSP · Under reviewResearch manuscript · Co-first author
Decouples shape-prior learning from boundary-detail refinement for segmentation in ambiguous medical images.
Technical Skills
- Programming & frameworks
- Python, PyTorch; deep-learning model implementation and experimental pipelines.
- Multimodal learning
- Vision–language models, instruction tuning, image grounding, and image–text dataset construction.
- Computer vision
- Visual representation learning, object detection, 2D/3D medical image segmentation, and whole-slide image modeling.
- Additional interests
- Music and generative AI; audio-structure analysis and digital audio production.
Honors & Awards
- First-class Scholarship, Beijing Jiaotong University.
- First-class Scholarship, Beijing Jiaotong University.