Events Calendar

Subspace Techniques for Parallel Magnetic Resonance Imaging

Speaker: Dr. Derya Dol Gungor, Ohio State University

Talk Title: Subspace Techniques for Parallel Magnetic Resonance Imaging

Series: Medical Imaging Seminar Series

Abstract: Parallel magnetic resonance imaging (pMRI) is an attempt to accelerate data acquisition by simultaneously collecting subsampled k-space data from multiple surface coils. The different sensitivity patterns for the various coils provide a spatial encoding and permit recovery from subsampled or otherwise aliased data. The smooth coil sensitivities in the image domain multiply with the single image representing the spin density of the excited slice. Via the Fourier transform, this can be written as a convolution of k-space representations of the coil sensitivities and the image. Since both the sensitivities and image are unknown in reality, this problem can be formulated as a blind multichannel deconvolution problem in the fully sampled case and this formulation allows us to use the established literature in signal processing to remedy the problems in parallel magnetic resonance imaging.

In this presentation, we particularly focus on subspace techniques to estimate both the coil sensitivities and the calibration kernels of the parallel imaging methods, which are conventionally extracted from a region of fully sampled low-pass calibration data. However, for high acceleration rates, the acquisition of the fully sampled calibration data becomes a limiting factor. Thus, we investigate extraction of coil sensitivities and calibration kernels from subsampled reference or ACS lines. We show that the subspace techniques can also be used for coil combination once the interpolated k-space data are obtained using coil-by-coil reconstruction techniques such as GRAPPA or SPIRiT. We demonstrate that the minimum mean square error (MMSE) criterion provides a non-iterative coil combination method that employs signal space vectors, and provides higher contrast images with less intensity inhomogeneity than well-known coil combination approaches such as square-root sum-of-squares (SoS) and adaptive coil combination. Finally, we show that subspace techniques can also be used in pre-processing to suppress noise by exploiting structure and low-rank property in matrices obtained from fully sampled and uniformly subsampled acquired data in parallel imaging.

Biography: Derya Gol Gungor received her B.S. degree in Electronics Engineering from Ankara University, Turkey in 2007, and Ph.D. degree in Electrical & Computer Engineering from the Ohio State University, USA in 2014. She spent a year in Bilkent University as a graduate research and teaching assistant. In 2013, she worked as a graduate research intern in Siemens Corporate Research, Princeton, NJ. Her general areas of interest are signal & image processing, and Magnetic Resonance Imaging. During her undergraduate, she was awarded with scholarships from Ankara University, Turkish Prime-ministry and Turkish Education Foundation (TEV).


Host: Professor Krishna Nayak

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248

Audiences: Everyone Is Invited

Contact: Talyia White