Magnetic Resonance Imaging: Multidimensional non-Gaussian Diffusion-weighted MRI - Image quality analysis
MetadataVis full innførsel
- Institutt for fysikk 
To investigate the viability of two different nuclear magnetic resonance imaging systems for acquisition of highly diffusion weighted, multidimensional human brain data, images were acquired on both, using two different diffusion sensitizing gradient setups. The sequence used was a diffusion weighted spin echo sequence with an echo planar read out. The diffusion gradients tested were monopolar and twice-refocused bipolar. These were used to acquire data on a Siemens Prisma and Skyra scanner. The data acquired with monopolar gradients on the Prisma scanner suffered from eddy current induced geometric distortions, while the bipolar scan on the Prisma and both Skyra acquisitions displayed lower signal to noise ratios. The difference between bipolar and monopolar signal intensity was noticeably greater on the Prisma system, due to an unexpectedly large drop when utilizing bipolar gradients. Background noise level were lower overall on the Prisma system, but particularly on the images acquired without diffusion weighting, due to the Skyra system experiencing a significant increase in background noise levels when not using diffusion gradients. Eddy current correction was unsuccessful for the Prisma monopolar sequence when using affine transformation with 12 degrees of freedom. Removal of non-brain matter proved difficult, due to the software having problems differentiating between background noise and brain matter in images with low signal to noise ratio. Of the scanner and gradient setup combinations tested, the Prisma scanner, using a bipolar gradient setup appears to be the superior choice, despite having comparable signal to noise ratio with the Skyra systems.