Our research lies at the intersection of signal processing, deep learning, and medical imaging hardware. We pursue end-to-end innovation: from imaging physics and algorithms to real-time deployment on portable, wearable, and robotic devices.
Analysis, measurement, and reliable deployment
Segmentation/detection-based automation of clinically relevant parameter measurement, and robust, trustworthy medical AI with calibration and uncertainty, controllability, and generalization for real clinical workflows.
Physics-guided reconstruction, inverse problems, quantitative imaging
Physics-informed neural networks that embed wave propagation, beamforming, and image-formation models into self-/unsupervised learning for high-fidelity ultrasound reconstruction, physics-guided inverse modeling for artifact reduction, and learning-based quantitative and perfusion ultrasound.
Active perception, sim-to-real, edge AI
Closed-loop robotic ultrasound guidance with active scanning policies, self-supervised adaptation across operators, devices, and patients, and data-/compute-efficient learning with hardware-aware design for edge devices.
