Our research focuses on three core directions: intelligent neural navigation, image-enabled brain-computer interfaces, and machine learning for smart healthcare. We combine rigorous methodology with close clinical collaboration to build systems that work in the real world.
针对扩散磁共振成像（dMRI）领域传统沿白质束量化技术高度依赖纤维束追踪、存在假阳性流线与交叉纤维精度不足等固有缺陷、扫描要求高难以适配临床与大样本研究，本研究提出了无需纤维束追踪的端到端深度学习框架 AutoATQ，通过 U-Mamba 网络从参数化球均值图像中实现 72 条人脑白质束的精准分割，再基于 PointNet 点云网络直接从白质束掩码预测束中心线并完成沿束微结构量化，无需传统纤维束追踪即可实现可靠的白质束测量分析。
We design explainable, privacy-preserving machine learning systems for clinical decision support—from early detection of neurological disorders to treatment response prediction in oncology.
We build closed-loop brain-computer interface systems that decode visual cortex activity to reconstruct perceived imagery and enable communication for patients with severe motor impairments.
We develop AI-driven navigation frameworks that fuse intraoperative imaging with preoperative MRI/CT to guide neurosurgical instruments in real time, achieving sub-millimeter targeting accuracy.