Projects

I have leveraged ideas from signal processing, optimization theory and machine learning to solve problems that arise in various Magnetic Resonance Imaging (MRI) applications. Specifically, I have focused on developing novel algorithms to accelerate parameter mapping of the brain, correct artifacts in Echo Planar MRI, and speed up the acquisition of Cardiac MRI.

Structured Low rank matrix recovery methods for Multi-dimensional MRI

Obtaining Magnetic Resonance (MR) images with good spatial, temporal resolution and large slice coverage simultaneously is very difficult due to the slow acquisition nature of the MR modality. One of the ways to speed up the acquisition process is to undersample the Fourier data and regularize the recovery using sparsity, smoothness and low rank priors. Recently, structured (Toeplitz/Hankel) low rank matrix regularizers have emerged as powerful alternatives for classical regularizers. Using different signal models and under few assumptions, I have been able to derive a structured matrix prior in the Fourier domain; this prior is used to solve a few problems in MRI. Click the links below to know more about the algorithms and applications involving structured matrix priors.

Accelerating dynamic MRI using self expressiveness prior

Inspired by the subspace clustering algorithm in machine learning, we propose an algorithm to recover a series of MR images from heavily undersampled Fourier measurements. Click the following link for more information.