Research
Ultrasound Guidance for Transoral Robotic Surgery
Sep 2021 – present, Robotics and Control Laboratory and Prisman Lab, UBC
Head and neck-related cancers account for a large percentage of all cancers globally, and transoral robotic surgery (TORS) shows the potential to help preserve patient function after treatment. However, TORS is challenging because it requires surgeons to have profound knowledge of anatomy. We hypothesize that ultrasound guidance can improve treatment outcomes in head and neck cancers, and we aim at developing novel ultrasound technologies for robotic-assisted surgery. This work is summarized in my PhD thesis.
Feasibility of MRI-US registration in the oropharynx
SPIE Medical Imaging, 2023
Feasibility study for semi-automatic MRI-to-US registration to support system calibration for TORS.
Ultrasound-guided augmented reality system for TORS
IJCARS, 2023 · arXiv
System design and prototype of the first ultrasound-guided augmented reality system for TORS.
Robot-assisted ultrasound scanning for TORS
IJCARS, 2024
Exploring robotic manipulation of the ultrasound probe to perform intraoperative scans during TORS.
Point cloud registration using semantic information and biomechanical energy regularization
arXiv, 2025
A non-rigid registration framework incorporating semantic segmentation and biomechanical priors.
Clinical evaluation of transcervical ultrasound in TORS
Oral Oncology, 2024 · SPIE Medical Imaging 2026
Clinical evaluation of ultrasound integration in TORS with the Prisman Lab.
Tissue Tracking and Landmark Retrieval in Ultrasound
Jan 2023 – Jul 2025, Robotics and Control Laboratory, UBC
I am broadly interested in representation learning and self-supervised learning for ultrasound image analysis — specifically, how models can learn pixel- or frame-wise representations for ultrasound that transfer to downstream tasks like tracking and information retrieval.
A tracking-any-point model for ultrasound
ASMUS, 2024 · arXiv
A new tracking-any-point model for ultrasound that outperforms state-of-the-art optical flow.
Self-supervised representation learning for ultrasound view retrieval
IJCARS, 2025 · arXiv
Leverages intra-sweep temporal information to improve classical contrastive learning for view retrieval.
AI in Lung Ultrasound COVID-19 Diagnosis and Segmentation
Nov 2020 – Aug 2021, Biomedical Image Guidance Lab, Carnegie Mellon University
This project developed AI tools for ultrasound image analysis, including lung region segmentation algorithms used to evaluate how different lung regions affect AI accuracy in COVID-19 severity classification, and optical flow methods to improve semantic segmentation accuracy.
Lung region segmentation for COVID-19 severity classification in ultrasound
MICCAI LL-COVID19 Workshop, 2021
Lung segmentation and optical flow-based methods to improve AI accuracy in COVID-19 severity classification.
Robotic Needle Steering and Tracking
Sep 2019 – Aug 2021, Biomedical Image Guidance Lab, Carnegie Mellon University
My research focused on ultrasound-based needle tracking and steering, including robust tracking under bending or partial visibility, lateral-manipulation-based steering control, and recalibration between robot kinematics and ultrasound coordinates.
Dual-path bending needle tracking in ultrasound
ISBI, 2021
A bending needle tracking algorithm utilizing kinematic consistency in segmentation to improve accuracy.
Ultrasound-based needle tracking and lateral manipulation planning for common needle steering
M.S. Thesis, Carnegie Mellon University, 2021
Improved mechanical modeling of needle-tissue interaction and two replanning algorithms for lateral-manipulation-based needle steering, removing the need for manual insertion point selection.
Early Research at Peking University
2017 – 2019, The Robotics Research Group, Peking University
Before graduate school, I worked on sensor fusion for attitude estimation, human swimming locomotion analysis using IMUs, and locomotion mode recognition for lower-limb prostheses.
Analysis of human swimming locomotion using IMUs
IEEE TMRB, 2019
IMU-based joint angle measurement system compared against optical motion capture for swimming analysis.
Environment-aware locomotion mode recognition of lower limb prostheses
Advanced Robotics, 2018
A CNN-based locomotion mode recognition method using raw strain gauge signals.
Sensor fusion for attitude measurement based on quaternions and Kalman filter
Undergraduate Thesis, Peking University, 2019
Kalman filter-based fusion of nine-axis IMU data (accelerometer, gyroscope, magnetometer) for real-time attitude estimation.
Human hand motion primitives during haptic search and retrieval of buried objects
Biomechatronics Lab, UCLA, 2018
Modeled hand movement sequences during search-and-retrieval tasks using NLP-inspired motion primitive representations.