Physicists interested in submillimeter fMRI, neuroscientists interested in the midbrain

To visualise BOLD signal changes in the midbrain, specifically the subthalamic nucleus and substantia nigra, with submillimeter resolution.

BOLD responses in the midbrain during tasks requiring inhibition (“stop tasks”) have been reported using 3 Tesla fMRI as arising from a relatively large complex including the subthalamic nucleus (STN) and substantia nigra (SN; Aron, 2006). The STN has been hypothesized to include functional subdivisions, including a motor, associative, and limbic subpart (Temel, Blokland, Steinbusch, & Visser-Vandewalle, 2005, but see Keuken et al., 2012;). Submillimeter fMRI would be useful in identifying such subdivisions in vivo. Both STN and SN are rich in iron, resulting in low baseline T₂^*-values and therefore require BOLD imaging with very short TE, especially at 7T. Multi-echo FLASH, which provides high spatial resolution at a cost of temporal resolution, has previously been used to study functional T₂^* changes in cortex at a submillimeter scale (Koopmans, Barth, & Norris, 2010). Here, we investigated whether submillimeter resolution BOLD data might be acquired from the STN/SN using a multi-echo FLASH acquisition.

Five healthy control subjects were scanned at 7T (Philips, Netherlands) in a session which included a high-resolution, large 3D-EPI acquisition (0.5mm isotropic, FOV 218*196*60 mm, TR/TE/α = 110 ms/30 ms/20^o, EPI factor 19) of which the magnitude data were used to confirm the location of the STN/SN for slice positioning and as anatomical background. Three functional runs of multi-echo GRE (0.75 mm isotropic, FOV 180*180*24 mm, TE=3.6/6.6/9.6/12.6/15.6 ms, TR/α = 19 ms/15^o, TR_volume = 40 s) were then acquired while subjects performed blocks of the stop signal task, interleaved with blocks of rest. The blocks were one TR long (40 seconds). During the stop task subjects indicated the direction of an arrow presented on the screen, unless they were presented a tone signalling they had to inhibit their response. The onset of tone presentation was dynamically shifted in such a way that subjects successfully stopped on 50% of the stop trials (Logan, Cowan, & Davis, 1984).

Data was analysed as follows: for each time point (n=60) a T₂^* map was obtained by combining the magnitude maps for different echo times after smoothing with a Gaussian kernel with 2.0mm full width at half maximum (FWHM). The smoothing was necessary to get stable T₂^* estimates. These T₂^* maps were submitted to a standard mass univariate GLM analysis, where the SPMs were overlaid on the anatomical scan. This 0.5 mm susceptibility-weighted anatomical scan was also used to draw masks of three clearly visible midbrain nuclei: the subthalamic nucleus (STN), substantia nigra (SN) and red nucleus (RN). These masks were then transformed to the 0.75 mm functional FLASH sequences to do a ROI-analyses on the average T₂^* per condition.

Functional data from one of the subjects had to be discarded due to discomfort, for another subject the behavioural data was lost. The remaining three subjects all performed the task well (on average 344±9 trials (mean±std), mean RT of 500±70 ms, the mean delay between onset of the arrow and the stop signal (the stop signal delay; SSD) was 234±80 ms with on average 51±10% successful stops). Average T₂^* maps computed from the functional data confirmed the very short T₂^* of the STN/SN/RN complex (rangeing from 15 to 25 ms; Figure 1). All single subject GLM results showed task related activation both in the visual cortex, due to the visual instruction of the task (Figure 2), and in small clusters in or near the STN. However, midbrain clusters were very small and often only bordered the STN or SN, rather than falling robustly inside it (Figure 3 and 4). ROI-analysis showed no robust difference in mean T₂^* between task and rest in any of the three segmented nuclei (Figure 5), emphasizing the need for high spatial resolution and a voxel-wise analysis. Multi-echo GRE can be used to acquire dynamic T₂^* data from the midbrain, although some tasks and brain regions may require a higher temporal resolution than can be achieved with this approach.