guest ::
| Home > Publications Database > Advances in Whole-Brain Diffusion MRI at High and Ultra-High Field |
| Home > Publications Database > Advances in Whole-Brain Diffusion MRI at High and Ultra-High Field |
| Habil / Postdoctoral Thesis (Non-german Habil) | DZNE-2026-00680 |
2024
This record in other databases:
Please use a persistent id in citations: urn:nbn:de:hbz:061-20250305-125158-2
Abstract: Magnetic Resonance Imaging (MRI) is an essential tool in both clinical diagnostics and research, offering a wide range of contrast mechanisms for imaging anatomical structures and assessing tissue properties. One key MRI application is diffusion-weighted imaging (DWI), which sensitizes the MRI signal to the diffusive motion of water molecules, enabling the investigation of tissue microstructure. This is especially valuable for imaging the human brain, as DWI can reveal white matter pathways by measuring diffusion anisotropy. DWI provides a non-invasive method to indirectly assess structures on the micrometer scale, despite MRI’s typical millimeter-scale spatial resolution. This thesis focuses on the design of fast sequences for whole-brain DWI, which are combined with techniques for correcting field inhomogeneity effects at high (3T) and ultra-high (7T) field strengths. In the first part, a framework for open-source sequence development and image reconstruction is presented. Different open-source tools for sequence design, image reconstruction and image analysis were combined to form a workflow that allows rapid prototyping of MR sequences and reconstructions, can be shared among other researchers, and is independent from proprietary software. The workflow was validated by acquiring and reconstructing data from different MRI scanners of different vendors. Example applications included a 3D gradient echo (GRE) sequence with controlled aliasing (CAIPIRINHA) acceleration, mapping of static deviations from the main magnetic field B0 and non- Cartesian imaging with spiral k-space trajectories. Both Cartesian and non-Cartesian image reconstruction algorithms were integrated into the workflow using open-source image reconstruction toolboxes. The reconstruction algorithms included optional k-space trajectory correction using the gradient impulse response function and correction for magnetic field inhomogeneity. The workflow was additionally integrated into the MRI simulation framework JEMRIS to allow direct comparison of experimental results to simulation results. The proposed pipeline was extensively used for sequence design and image reconstruction in the subsequent chapters. The second part describes the improvement of non-invasive axon radius estimation in the human white matter using a multiband spiral sequence on a 3T scanner with a highperformance gradient system. Magnetic field monitoring was used to reduce artifacts stemming from dynamic field inhomogeneities, mainly caused by eddy currents during the course of the DWI sequence. The actual encoding fields, measured up to the third spatial order, and static magnetic field inhomogeneities, measured using a dual-echo GRE sequence, were both incorporated into the image reconstruction algorithm. The spiral sequence was compared to the current gold standard multiband echo-planar imaging (EPI) sequence with image-based field corrections. An established diffusion-weighted imaging protocol, which showed reproducible results in a previous study, was used to estimate axon radii in white matter voxels. Strong diffusion weighting was applied to suppress all extra-axonal signals, enabling the use of a single-compartment model for the intra-axonal space. The axon radii were then estimated from the orientationally averaged DWI signal. A test–retest study was performed to assess the repeatability of axon radius estimation with EPI and spiral sequences. The higher signal-to-noise ratio (SNR) provided by the spiral readout led to reduced test–retest variability of axon radius estimates. Incorporating the actual encoding fields in the image reconstruction algorithm effectively reduced artifacts related to eddy currents, which are caused by strong diffusion gradients. However, a significant bias was detected in the test–retest measurements of some subjects for both EPI and spiral sequences, indicating a potential issue with the repeatability of axon radius estimation. In the third part a new variant of the 3D dual refocusing echo acquisition mode (3DREAM) sequence is developed at a 7T scanner. The 3DREAM sequence allows for mapping of the B1 field, which is a prerequisite for parallel transmit (pTx) techniques. The new variant uses a 3D stack-of-spirals instead of a Cartesian readout scheme, with the aim of reducing blurring and increasing the effective resolution in B1 maps. The blurring is caused by the fast decay of the stimulated echo signal, which is prepared only once at the beginning of the sequence and decays with each following excitation. Using a spiral readout allows for more efficient k-space sampling, thus reducing the number of excitations. The spiral 3DREAM sequence was compared to the Cartesian 3DREAM sequence and three other established B1 mapping methods in phantom and in-vivo measurements. Both 3DREAM sequences showed excellent agreement with the three other methods, while their acquisition time was significantly lower. Blurring and ventricular contrast in B1 maps were reduced for the spiral 3DREAM compared to Cartesian 3DREAM, with the reduction becoming more significant at higher resolutions. In the last part of this thesis, whole-brain DWI is performed on an ultra-high field (7T) scanner. At ultra-high fields the inhomogeneity of the B1 field leads to severe signal dropouts in lower brain areas such as the cerebellum. An eight transmit channel RF coil and specifically designed pTx pulses were used to mitigate these inhomogeneities. Multiband pTx pulses were integrated into EPI and spiral DWI sequences. Static and dynamic field inhomogeneities were addressed by using field monitoring and static B0 field mapping. The performance of the pTx pulses was compared to standard circularly polarized pulses. EPI and spiral images reconstructed with field monitoring data were compared to EPI data with image-based field inhomogeneity correction. It was found that deteriorating effects of B1 inhomogeneities were largely resolved by using pTx pulses. Increased signal in the cerebellum improved diffusion tensor fitting and resulted in more accurate fractional anisotropy and mean diffusivity maps. A remaining challenge is the low bandwidth of the pTx pulses, which results in bended slices in regions with large B0 inhomogeneity. In accordance to the results of part two, the SNR for spiral DWI was increased compared to EPI DWI. However, at a field strength of 7T, static B0 inhomogeneity posed a more severe problem in image reconstruction compared to 3T, as artifacts such as geometric distortions and blurring could not be fully resolved.
; 
|
The record appears in these collections: |
PDF
PDF (PDFA)