MR signal formation in cortical vessel networks
Brain imaging at 9.4 T or 14.1 T opens the possibility to resolve structures far below the thickness of 2-3 mm of the human neurocortex. In most areas the cortical sheet is divided into six cortical layers with a generic pattern of long-range feed-forward connections to higher-order brain regions and feedback connections to lower-order cortical and subcortical regions. Some cortical regions, are organized as cylinder-like cortical columns perpendicular to the cortical surface of cell bands with similar somatotopic response properties.
At ultra-high fields, anatomical images with an isotropic resolution below 200 mm have been obtained that clearly resolve subcortical structures such as the line of Gennari or subcortical units within the brain stem. Functional BOLD images acquired with gradient echo or spin echo methods, in principle, may offer the same spatial resolution as anatomical images. Therefore, at very high fields the BOLD response from single cortical layers or other mesoscopic substructures might be resolved. However, while the spatial point-spread-function in structural images is basically given by a sinc function with a width given by the inverse of the sampled k-space coverage, in functional BOLD imaging the neurovascular point-spread-function defines the resolution of a point-like neuronal event that is blurred by vascular-related changes related to tissue oxygenation, blood flow and blood volume. For a more detailed understanding of these effects we applied physiological and statistical methods to simulate and measure MR signal formation within a neurovascular vessel network for different types of acquisition sequences.
Functional MRI using balanced SSFP
Since about 20 years EPI has been the working horse for functional MRI. With its intrinsically high T2* sensitivity and signal stability it currently seems to be the best choice for BOLD fMRI. Disadvantages such as image distortions and signal drop-outs have been largely improved even at very high field strength using point-spread function corrections and acceleration techniques. However, severe problems such as co-registration to anatomical data especially at very high resolution as well as the high and somewhat unspecific and large vessel-sensitive T2* contrast in addition to a certain blurring in phase-encode direction remain. In this project we investigate the feasibility of high-resolution functional imaging of the human brain using pass band balanced SSFP at 9.4T. Furthermore, we propose a multi-line bSSFP sequence optimized for short TR and fast imaging and analyze its acquisition efficiency by looking at the ADC duty cycle, the temporal and thermal signal to noise (tSNR and SNR0), as well as the observed BOLD signal changes in the bSSFP passband upon a visual task for echo trains ranging from one up to seven echoes. We have demonstrated the feasibility of high-resolution bSSFP at 9.4T to detect functional activation using high parallel imaging acceleration and minimized repetition times. Activation patterns and signal changes were very stable and reproducible across subjects within the visual cortex, and comparable to reported values of SE-EPI at 7T and 9.4T. The acquired functional maps are without any spatial distortions, allowing for a precise registration to anatomical images. Using segmentation of the intra-RF-pulse interval of 3 to 7 echoes provides an increase in imaging speed by a factor of about 3, and a TR-related increase of functional contrast by about 2.
Advanced, local dynamic shimming
This project aims to design dedicated shim arrays adapted to the requirements of brain imaging based on acquired sets of representative human brain field maps for cutting-edge applications at 3T and 9.4T. The main objective is to investigate field inhomogeneities created by anatomic susceptibility variations in the human head as a basis for the design and construction of dedicated local shim systems. The final design will be a one-in all solution that consists of a three-layer design starting with the innermost receive coil helmet, followed by the transmit coils and finally (after rf shield) the shim coils mounted on a close-fitting cylinder. In our current initial version 16 loops with 10 cm diameter were distributed equally on the surface of a cylinder. Self-built amplifiers with a ramp time of 50 ms between ±1 A were connected to each loop allowing for fast dynamic switching of shim currents. We could not detect any artifacts or increase in temporal SNR with and without attached shims. Also, the rapid linear amplifiers were able to compensate currents induced in the loop by switched liner gradients. The resulting shim quality corresponds to a shim achieved with full order 3rd harmonics. Using the modular design of stacked cylinders allows us to easily change the shape and arrangement of current loops towards a more dedicated topology compared to simple 16 loops.
Signal modelling based on realistic vessel networks
Monte Carlo simulations will be used to analyze oxygenation-related signal changes in balanced steady state free precession (bSSFP) as well as in gradient echo (GE) and spin echo (SE) sequences. Signal changes will be calculated for artificial cylinders and neurovascular networks acquired from the mouse parietal cortex by with two-photon laser scanning microscopy at 1 µm isotropic resolution. Signal changes as a function of vessel size, blood, volume, vessel orientation to the main magnetic field B0 as well as relations of intra- and extravascular and of micro- and macrovascular contributions will be investigated .
At 9.4T, the intravascular contribution is negligible for GE, and depending on the BV, increases to about 5% - 10% of the overall BOLD signal change for SE and bSSFP. Except for larger veins that occupy a huge volume fraction of the imaging voxel, bSSFP is mostly sensitive to extravascular contributions from small vessels. The BOLD signal ratio between microvessels and intracortical veins is highest for deep layers (about 20:1) and decreases to about 4:1 for the surface layer for bSSFP. Thus, the contribution of larger cortical veins is in the range of 5% to 20%, indicating a strong sensitivity of bSSFP to the microvasculature. All sequences show a strong dependence on the orientation of the cortical layers to B0 due to perpendicularly arranged feeding and draining vessels (Fig. 3).