High-Field MRI
One of the goals of the RIBS research program is to understand what current imaging measures of brain function relate can tell us about how the brain processes information. We have compared spatial patterns of fMRI BOLD signal changes to electro-cortical brain stimulation (ECS) in epilepsy patients who have electrode grids implanted for localization of the source of seizures (papers). We use a special fMRI scan technique that is less sensitive to signal from inflowing and draining blood vessels to improve accuracy of localizing neuronal tissue. This technique, 3D PRESTO, was first conceived by Peter van Gelderen, Guoying Liu Bin and Chrit Moonen at the NIH (Bethesda, MD, USA), and has been refined and improved upon at the UMC Utrecht. In comparing PRESTO to ECS we found that fMRI identifies all language foci in an individual subject, but also as many foci where ECS does not affect language performance. This discrepancy was attributed to either artifacts in fMRI image acquisition and analysis, or to the notion that part of the brain activity seen with fMRI reflects regions that are involved in processing language, but are not indispensible (involved but not critical). We are currently investigate the second option by comparing fMRI activity patterns to electrical neuronal activity in epilepsy patients. Signals from the cortical electrodes are obtained while the patient performs the same tasks as those performed during an fMRI session before the implant, and is converted to the frequency domain (notably bandwidths in the higher frequencies such as 55-95 Hz). These electrocorticographic (ECoG) are also the basis for the BCI projects. The first option, attributing discrepancies between imaging measures to artifacts in fMRI, is now the subject of investigation with the 7 Tesla human MRI system that was installed in the UMC in 2007.
Data acquisition and data analysis at 7T are quite different from 3T and lower field scanners. To obtain powerful fMRI imaging at 7T we are developing robust EPI scan techniques for high-resolution imaging at high speed. Speed has always been high on the priority list because I believe that effects of motion (both cyclic related to arterial pulsation and to respiration, and incidental head motion) are best corrected for by reducing volume acquisition time as much as possible. At 3T our standard volume acquisition time is 0.6 seconds (whole brain). The amount of motion reduces with scan time within a single volume acquisition, and is best corrected with volume registration in data analysis. The optimal volume acquisition time would be 0.2 s, because at this rate we get 4 measurements per heartbeat, allowing for near-complete motion correction of arterial pulsation in data processing. High resolution imaging improves our ability to map brain activity within the grey matter layer. Most of the draining veins are located at the pial surface of the cortex, so better localization of neuronal activity is obtained with small voxels within the grey matter layer. We currently use EPI for our functional studies, which has been developed by Natalia Petridou at the University of Nottingham (with Richard Bowtell and Penny Gowland). However, 3D EPI and 3D PRESTO are under development with Natalia who now works in the UMC 7T group.
Several projects are currently conducted to elucidate the relationship between fMRI BOLD activity and neuronal activity (ECoG), and to map neuronal representations of different components of specific cognitive functions within known cortical nodes at mm resolution. Jeroen Siero is investigating how we can use high-resolution BOLD fMRI to get spatially accurate maps of brain activity. He found that the shape of the BOLD curve differs between voxels at the surface of the cortical grey matter layer and those located deeper within the layer. The primary difference is the width of the BOLD curve. We expect this knowledge to be of use in identifying the exact location of neuronal tissue involved in a particular function. It will allow for highly detailed topographical maps of neuronal functions at a mm resolution because we expect to be able to exclude even smaller draining veins from blurring the BOLD activity maps within a particular cortical patch. He is also conducting a project with epilepsy patients who have a high-density electrode grid implanted after fMRI scans, together with Dora Hermes. Here, we compare 7T BOLD maps at 1.5 mm resolution with ECoG maps at 1.5 mm resolution.
Detailed brain mapping of brain functions is conducted for two reasons. First, we want to identify patches of cortex where we can decode activity for future use in icBCI, these locations will be targets for neuronavigation-based implant of epidural electrodes. Projects are by Martin Bleichner and Gert Kristo. They use a newly developed 16-channel surface coil (Dennis KLomp and Natalia Petridou) to map motor cortex representations of hand gestures, and of language production respectively. Pieter van de Vijver has completed a study where we map different aspects of Working Memory in dorsolateral prefrontal cortex and superior parietal cortex in the left hemisphere. The result indicate that information processing and maintenance are served by distinctly different regions, and that different stimulus modalities (auditory, verbal, object, spatial) are served also by distinct regions. These result are based on group-averaged data, which by definition reduces the spatial resolution. A next challenge is to display activity patterns in the original high resolution in individual subjects, and to compare within-region patterns across subjects preserving the high spatial detail. For this, Mathijs Raemaekers is working on translating cortical flattening techniques from the field of visual cortex research for the 7T anatomy scans.
