Functional Imaging at Ultra-High Field
Due to the higher signal-to-noise ratio, functional imaging can be performed at very high field strengths with a spatial resolution ≤ 1 mm. This makes it possible for the first time to investigate feed-forward and feedback processes as well as functional subunits, such as occular dominance columns, non-invasively in humans.
At the same time, however, fMRI at 9.4 T is associated with a number of difficulties. For example, the most common method for functional MRI (gradient-echo EPI) is susceptible to perturbations of the static magnetic field, which can lead to strong spatial distortions. In addition, there are limitations due to peripheral nerve stimulation and, for some sequences, SAR.
In my research, I modify imaging sequences to achieve the highest possible spatial and temporal resolution for fMRI at 9.4 T. This knowledge is the basis for a variety of projects and serves in the long term to enable basic neuroscience research on the sub-millimeter scale with MRI.
Blood-Oxygen Level Dependent (BOLD) Imaging
BOLD signal changes are commonly measured with gradient-echo EPI due to its sensitivity to T2*-changes. However, this imaging method is also particularly susceptible to spatial non-specific extravascular signals from large veins. As our research has shown, these signal not only overlay BOLD signals from the cortex at the surface, but can still have a strong influence even in deeper cortical layers.
At the same time, especially EPI images with very high spatial resolution often also exhibit strong spatial distortions due to the long readout train and the resulting low bandwidth in the phase-encoding direction. Therefore, distortion correction is usually necessary if functional data are to be co-registered with anatomical data for analysis. However, we observed that this correction can itself lead to spatial correlations, which make an interpretation of the functional signals in anatomical space impossible.
Thus, another part of my research deals with validation of alternative BOLD sequences with a focus on balanced steady state free precession (bSSFP) based methods. From signal simulations, bSSFP is expected to capture a much smaller signal change from large veins and therefore can capture BOLD signal change much more specifically than GE-EPI.
Measurement of Cerebral Blood Flow
Because BOLD signal changes are caused by parallel but non-synchronous processes such as changes in oxygen extraction fraction, blood volume, and blood flow, they are only very indirectly related to neuronal activation. Theoretically, therefore, more accurate statements about brain activity can be made by measuring local changes in cerebral blood flow (CBF) or blood volume (CBV). Both parameters are also interesting physiological parameters that are altered in a variety of diseases.
The most common method for measuring CBF is arterial spin labeling (ASL), which uses the blood itself as the endogenous contrast agent. However, this noninvasive method has two major drawbacks. First, the temporal resolution is quite poor and second, the SNR is very low because the contribution of the blood to the overall signal is very small.
I am trying to reduce the SAR burden on the sequence and increase the labeling efficiency by introducing labeling methods such as the so-called dual-coil ASL. In addition, inversion pulses have been developed for the so-called pulse ASL, which allow magnetization inversion throughout the head despite the inhomogeneous transmitted field at 9.4 T. It should also be noted that in functional measurements, especially in GE-EPI-based readouts, BOLD signal changes are usually superimposed on the ASL signal, which complicates the interpretation of the ASL signals. To minimize these effects, an ASL sequence with a fast bSSFP-based readout was also developed.
In the future, a focus will be on combining parallel transmission and ASL sequences. It is expected that already layer-selective B1+ shimming for the labeling and readout region will provide a major advance for both pulsed and pseudo-continuous ASL methods. In addition, segmented 3D EPI readouts will be optimized for use at 9.4 T.
High- and Very-high-resolution anatomical imaging
T1-weighted Imaging and Quantitative T1
For both functional and structural MRI studies, T1-weighted sequences are usually indispensable as they provide optimal contrast between CSF, gray matter and white matter. They are therefore mostly used for algorithms for automatic segmentation, which also allows the determination of additional parameters such as cortical thickness.
The inhomogeneous transmission field at 9.4 T requires an optimization of the inversion pulse and the readout train of both MPRAGE and MP2RAGE sequences. The latter, with a skillful choice of parameters, also allows quantification of the longitudinal relaxation time (T1) which is of increasing interest due to its close relationship with cortical myelination. However, since complete stability of MP2RAGE to variations of the transmitting field (B1+) at 9.4 T cannot be achieved, an automatic correction method was developed that allows to obtain correct T1 values using a B1+-map.
The developed standard protocols currently allow whole-brain imaging with 0.6 mm isotropic resolution in 11 min for MP2RAGE (for T1 quantification) and 8.5 min for MPRAGE. In addition, there are settings that enable 0.8 mm whole-brain imaging with MPRAGE in 3 min.
For very high resolution T1 measurements, a segmented MP2RAGE sequence which uses prospective motion correction was developed. This makes it possible to perform quantitative T1 measurements of the whole brain with 350 µm resolutions in less than one hour - an experiment duration that is not too long even for larger anatomical studies.
Susceptibility Weighted Imaging
While the stronger susceptibility effects are a challenge for functional imaging with GE-EPI at 9.4 T, they are a great advantage for imaging venous structures with a conventional gradient-echo sequence. Thus, it is possible to image even small ascending veins in the cortex or to automatically mask regions around larger veins for alternative analysis of functional data. If a further quantification of the susceptibility parameters or T2* is desirable, a multi-echo instead of low bandwidth single-echo readout is usually used. In the near future, we will also try to correct for the apparent increased venous vessel size at long echo times by using multi-echo data for venous segmentation as well. In addition to sequence optimization, image reconstruction plays a special role for such data in order to remove physiological B0-fluctuations as stably as possible.
2020 – present
Ph.D. in Neuroscience
Graduate School of
Neural & Behavioral Sciences and
MPI for biological
Dipl.-Ing. (FH) in Medical Engineering
University of Applied
2018 - present
|Reviewer for NeuroImage and Magnetic Resonance in Medicine (MRM)
2016 - present
Supervisor for undergraduate and graduate students during their bachelor, master and PhD projects