Computational imaging · Inverse problems · Machine learning
Research Fellow, Medical Physics
Memorial Sloan Kettering Cancer Center
New York, USA
I develop physics-informed learning methods for fast and reliable image reconstruction, with applications in MRI, industrial CT, and image-guided radiation therapy.
I am a Research Fellow in Medical Physics at Memorial Sloan Kettering Cancer Center. My research focuses on inverse problems in computational imaging: how to recover useful images from incomplete measurements while preserving consistency with the underlying physics.
I previously worked as a Postdoctoral Research Associate at Oak Ridge National Laboratory on high-resolution X-ray CT reconstruction for additively manufactured parts. I received my PhD in Electrical and Computer Engineering from the University of Iowa, advised by Prof. Mathews Jacob.
My current work includes real-time 3D MRI and respiratory motion estimation for MRI-guided radiotherapy. Longer-term interests include generative priors, uncertainty quantification, and foundation models for scientific imaging.
I develop reconstruction methods that combine acquisition physics with learned image priors. Work on O-MoDL, C-MoDL, and Deep-SLR learns structured relationships in Fourier space for accelerated, calibrationless parallel MRI. Joint reconstruction and segmentation connects image recovery to downstream analysis.
My work on monotone operator learning (MOL) studies fixed-point uniqueness, convergence, and stability under explicit model constraints, with a deep-equilibrium formulation that reduces training memory. ENSURE addresses reconstruction learning without clean, fully sampled training images.
Ada-MoDL adapts a single MRI network to different contrasts, field strengths, and acceleration factors. In industrial cone-beam CT, I investigate memory-efficient iterative reconstruction, automatic regularization, and artifact reduction priors for sparse-view imaging and nondestructive evaluation.
Current research
Respiratory motion complicates imaging and treatment of moving abdominal anatomy. I investigate an offline–online strategy that learns a respiratory motion dictionary, then uses incoming projections to estimate a 3D motion state. A learning-based extension, 3DLift, aims to reconstruct motion outside the dictionary’s observed range.
Ongoing research presented in my research talks.
Journal of Nondestructive Evaluation, 44(4), 147, 2025.
Combines a multi-slice artifact reduction prior with data consistency for sparse-view industrial XCT, preserving fine structures and improving defect detection.
Magnetic resonance in medicine, 90(5), 2033-2051, 2023.
Conditions an unrolled MRI reconstruction network on acquisition settings, adapting feature scaling and regularization across contrasts, field strengths, and acceleration factors.
IEEE Transactions on Computational Imaging, 9, 260-275, 2023.
Studies fixed-point uniqueness, convergence, and robustness under strict monotone constraints, with a deep-equilibrium formulation that reduces training memory. The paper also evaluates an approximate constraint with empirical rather than strict guarantees.
IEEE Transactions on Medical Imaging, 42(4), 1133-1144, 2023.
Estimates reconstruction risk across randomized measurement operators to train image reconstruction networks without clean, fully sampled targets.
2023 journal issue; first published online in 2022.
IEEE Transactions on Medical Imaging, 39(12), 4186-4197, 2020.
Learns Fourier-domain annihilation relations for calibrationless parallel MRI, combining structured low-rank modeling with neural priors.
Browse journal articles, conference papers, abstracts & patents →
I enjoy running trails and half marathons, and occasionally recording songs.