Rui Tian

Current project

Spread spectrum MRI with local B₀ coil array

My scientific path

My research interests emcompasses the physics and mathematics of magnetic resonance imaging (MRI), employing versatile skillsets (e.g., spin physics, pulse sequence programming, image reconstruction, mathematical signal processing) to explore new MRI methodology and contribute to imaging science.

My journey in MRI started with some undergraduate research experience in RF coils. During master study, I mainly worked on super-resolution MRI with phaseless encoding, a technique which borrows the essential elements of super-resolution fluoresence microscopy, particularly the structured illumination miscroscopy, to MRI. Specifically, a sinusoidal tagging modulation of the longitudinal magnetization is applied prior to signal acquisition, so that several k-space bands are simultaneously excited (i.e., similar to inducing aliasing in k-space). Therefore, multiple shots of low-resolution scans can be acquired, with high spatial frequencies mixed in the low-resolution EPI acquisition window and later resolved by phase shifts of the tagging pattern. In this way, a super-resolution image can be reconstructed from several low-resolution magnitude images, eliminating the notorious shot-to-shot phase variations which can lead to ghost artifacts and require additional 2D navigators or post-processing to remove them but with certain limitations.

To gain further access to magnetic resonance in microscopic realm (e.g., nanometer), during the last 6 months of my master degree, I digressed to quantum sensing using nitrogen vacancy center, programming EPR pulse sequence to probe 5nm nanodiamonds, on a confocal microscopy platform.

Currently, my main project is about nonlinear B₀ fields in MRI. With a recently proposed concept called local B₀ coil array, I am particularly interested in, how the independently driven B₀ coil elements can be optimally manipulated in spatial-temporal domain, to accelerate MRI scans in its ultimate limit. Towards this goal, it is crucial to rethink many established concepts across all aspects of MRI. While this ambitious project will inevitably confront those familiar challenges in MRI such as image reconstruction or field calibration, the conventional k-space formalism, serving as the mathematical foundation for all MRI methods, becomes disrupted in the presence of nonlinear gradient encoding.

So far, fortunately, by employing a reproducing kernel Hilbert Space formalism, the k-space sampling domain has been brought back to nonlinear gradient encoding, which again, allows a quantitative analysis of the signal sampling and pulse sequence design as conventional MRI. The image reconstruction technique has also been substantially improved, thanks to the advances of mathematical signal processing and parallel imaging in the past decades so that we can learn from their wisdoms. For future study, the nonlinear B₀ fields will be investigated in more detailed scenarios for image acceleration, under the theoretical framework and engineering pipeline substantially improved in the last several years. In turn, the lessions learned from nonlinear gradient encodings will benefit methodology using linear gradients only, and therefore, pushing forward not only the local B₀ array technology, but also a better and more general understanding of how MRI images can be formed.