Guanhua Wang

I am a computational scientist engineer at Q Bio, Inc. I received my Ph.D. from Dept. Biomedical Engineering @ UMich, advised by Prof. Douglas Noll and Prof. Jeffrey Fessler in 2023 (thesis).

My main research interest includes fMRI/MRI imaging, inverse problems, and machine learning. Now in the industry, I work more on hardware imperfection correction (using computational approaches) and large-scale solvers. It is exciting to learn how to build a real product!

Fast (always accessible and affordable) and accurate medical imaging is what excites me every day.

Education

University of Michigan, Ann Arbor, MI 2019 - 2023

Research Assistant, fMRI Laboratory

Tsinghua University, Beijing, China 2015 - 2019

B. Eng

Experience

Q Bio, Inc., San Carlos, CA June 2023 – Present

Computational Science Engineer

Correction of system imperfections by combining hardware-based and computational methods.
Development of efficient large-scale inverse problem solvers.

Q Bio, Inc., San Carlos, CA June 2021 – Sept. 2021

Radiomics Intern

Developed an unprecedentedly fast, model-based imaging methods for longitudinal health tracking.

Stanford University, Palo Alto, CA July 2018 – Sept. 2018

Research Assistant, Dept. Electrical Engineering

Chinese Undergraduate Visiting Research (UGVR) Program

Research Projects

The spark of my phd research was ignited during a summer internship at Stanford:

as a computational imaging modality, can we build a differentiable digital twin of the MRI system, and optimize the imaging system as a whole, via gradient methods, instead of the component-by-component (consider the fact that pulse design and reconstruction are rarely jointly optimized) and GSGD (gradient descent by graduate students) manner.

Gradient Optimization of MRI Sampling Strategy (BJORK and SNOPY)

3D: https://arxiv.org/abs/2209.11030
2D: https://arxiv.org/abs/2101.11369
Theory: https://arxiv.org/abs/2111.02912
Code: https://github.com/guanhuaw/SNOPY

Combine Blind Learning with Supervised Learning for Medical Image Reconstruction (BLIPS)

Code: https://github.com/JeffFessler/BLIPSrecon

Differentiable MRI: MIRTorch, A Medical Image Reconstruction/Optimization Toolbox

Supports linear operators, proximal optimization and unrolled network. Powered by PyTorch.
Available here: https://github.com/guanhuaw/MIRTorch. Waiting for your advice!

Projects before I came to Michigan

Data-driven Synthetized Multi-contrast MR

Adviser: Prof. Greg Zaharchuk, Prof. John Pauly & Dr. Enhao Gong

June 2018- Oct. 2018 | Stanford University | Research Assistant

Generate multi-contrast neuroimaging using the deep network, with much fewer artifacts than traditional model-fitting algorithms like SyMRI.
Proposed and evaluated a novel network (Multi-Task GAN) that can generate multiple contrasts using one single set of weights.
Worked closely with radiologists to evaluate the robustness and applicability of the proposed model.

Fast 3D Temperature Mapping

Adviser: Prof. Kui Ying

Nov. 2017- Present | Tsinghua University | Research Assistant

Modified 3D golden-angle-ratio stack-of-stars radial sequence to MR temperature mapping.
Accomplished fast 3D reconstruction algorithm.

Non-cartesian MRI Reconstruction

Adviser: Prof. Karen Ying & Prof. Huijun Chen

Dec. 2016 – July 2017 | Tsinghua University | Research Assistant

Proposed a method to reconstruct hepatic DCE imaging, which could achieve higher spatiotemporal resolution than traditional L+S method.

Selected Publications

G. Wang, J. F. Nielsen, J. A. Fessler, and D. C. Noll. Stochastic Optimization of 3D Non-Cartesian Sampling Trajectory (SNOPY). 2023. Magn Reson Med.

G. Wang and J. A. Fessler. Efficient approximation of Jacobian matrices involving a non-uniform fast Fourier transform (NUFFT). 2023. IEEE Trans. on Comput. Imag.

G. Wang, T. Luo, J. F. Nielsen, D. C. Noll, and J. A. Fessler. B-spline parameterized joint optimization of reconstruction and k-space trajectories (BJORK) for accelerated 2D MRI. 2022. IEEE Trans. on Med. Imag.

A. Lahiri*, G. Wang*, S. Ravishankar, and J. A. Fessler. Blind Primed Supervised (BLIPS) Learning for MR Image Reconstruction. 2021. IEEE Trans. on Med. Imag.

G. Wang, E.Gong, S. Banerjee, D. Martin, E. Tong, J. Choi, H. Chen, M. Wintermark, J.M. Pauly, and G. Zaharchuk. Synthesize High-quality Multi-contrast Magnetic Resonance Imaging from Multi-echo Acquisition Using Multi-task Deep Generative Model. 2020. IEEE Trans. on Med. Imag.

G. Wang, E. Gong, S. Banerjee, J. Pauly, and G. Zaharchuk. Accelerated MRI Reconstruction with Dual-Domain Generative Adversarial Network. 2019. In Proc. Int. Work. Mach. Learn. Med. Image Recon (MLMIR).