Noise Assumptions in Complex-Valued SENSE MR Image Reconstruction

• Activity Number: 466
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Daniel B. Rowe*+ and Iain P. Bruce
• Companies: Marquette University and Marquette University
• Keywords: fMRI ; complex-valued ; phase ; MRI ; SENSE ; k-space

Abstract

In fMRI, brain images are not measured instantaneously and a volume of images can take two seconds to acquire at a low 64x64 resolution. Significant effort has been put forth on many fronts to decrease image acquisition time including parallel imaging. In parallel imaging, sub-sampled spatial frequency points are measured in parallel and combined to form a single image. Measurement time is decreased at the expense of increased image reconstruction difficulty and time. One significant parallel imaging technique known as SENSE utilizes a complex-valued regression coefficient estimation process with transposes replaced by conjugate transposes. However, in SENSE the noise structure is not properly modeled. This work properly models the noise structure for complex-valued least squares regression. Differences in estimated images between SENSE and our new estimation procedure are evaluated.