Borui Jia
贾博睿

Third-year B.Eng. Student  ·  Tongji University, Department of Computer Science

I work on graph neural networks and deep learning for medical image analysis and brain disease detection, advised by Prof. Yufei Chen (陈宇飞). My current interests sit at the intersection of fMRI-based brain network modeling, graph-structured learning, and medical image classification. I am attending ICML 2026 in Seoul (Jul 5–11) to present our accepted paper.

BJ
ICML
2026
2026

ADHD Disease Detection Based on Short- and Long-Term Brain Function Encoding and Memory Graph Network

Dongxun Jiang (蒋东珣), Borui Jia, et al.

We propose SLT-BFGN, a dual-channel graph network that jointly encodes short-term dynamic brain connectivity (sliding-window partial correlation graphs) and long-term static functional connectivity. A memory-augmented graph module captures cross-window temporal dependencies, achieving state-of-the-art ADHD detection on ADHD-200.

CAS Institute for Brain and Cognitive Sciences  ·  脑智卓越中心 Apr 2026 – Present

Research Intern  ·  Shanghai

Zebrafish Whole-Brain Image Registration

Compared VoxelMorph (deep learning) vs. ANTs SyN (traditional) for registering 77 zebrafish whole-brain volumes (May 2026). VoxelMorph at 2× downsampling with trilinear flow upsampling achieves mean NCC 0.656, outperforming ANTs SyN.

VoxelMorph ANTs SyN 3D Registration Neuroscience

Mouse Digital Brain Model (Ongoing)

Building a multimodal predictive model for a mouse digital brain with wet-lab validation. Guaranteed ≥ 2nd authorship on resulting publication.

Multimodal Neuroscience In Progress

Spinal Vertebra Fracture Classification

Three-class (H/N/O) CT-based vertebra fracture detector, in collaboration with Shanghai Tenth People's Hospital. Built a full DICOM→NIfTI slice extraction→ROI cropping pipeline. DenseNet169 backbone with case-level stratified cross-validation; Grad-CAM analysis of hard examples to diagnose activation drift outside vertebra masks.

Medical Imaging DenseNet169 Grad-CAM CT PyTorch

Zebrafish Whole-Brain Image Registration

Compared VoxelMorph (deep learning) vs. ANTs SyN (traditional) for registering 77 zebrafish whole-brain volumes. VoxelMorph at 2× downsampling with trilinear flow upsampling outperforms ANTs SyN (mean NCC 0.656), a publishable finding for the zebrafish neuroscience field.

VoxelMorph ANTs SyN 3D Registration Neuroscience

fMRI Brain Disease Detection Survey

Comprehensive review of graph-network paradigms for fMRI-based brain disease detection, covering GNN, Hypergraph NN, Graph Transformer, and VLM approaches. Targeting Science China Information Sciences.

Survey GNN fMRI Brain Networks
Tongji University  ·  同济大学 Sep 2023 – Jun 2027

B.Eng. in Computer Science and Technology

Advisor: Prof. Yufei Chen (陈宇飞)  ·  Guaranteed postgraduate admission (保研)

GPA 4.1 / 5.0
  • ICML 2026  — Paper accepted, Second Author
  • MCM 2026 Outstanding Winner (O Prize) & INFORMS Award  — Team Leader
  • Tongji University Academic Scholarship (Third Prize)