The superior colliculus (SC) is a paired layered structure located in the midbrain. It consists of 7 layers: the most superficial layer receives input mainly from the retina; the intermediate layers contain neurons mostly involved in oculomotor control; and the deep layers contain multisensory and visuomotor neurons [1]. In this study we investigated the benefits of using 9.4T MRI for humans to study small brain structures in the brainstem, like the SC, in comparison with 3T data. The increase in sensitivity in the blood oxygen level dependent size allowed us to use higher resolution than at 3T and to get functional consistency across subjects on a short (20 min) stimulation paradigm.

Left or right–sided half circles of a checkerboard were shown in a block design (15s duration) interleaved with fixation baseline. For the 9.4T, a custom-made head coil was used [2] and 9 healthy volunteers participated and underwent two sessions of 10 stimulation cycles (one cycle: fixation baseline, right check. and left check.). fMRI data was acquired using GR EPI with 1mm isotropic resolution and TR of 1s.For the 3T analysis we randomly selected 9 datasets from Linzenbold et al. 2011 [3] with 3 sessions of 10 cycles of visual stimulation each. fMRI data was acquired with 2 mm slice thickness, 1.5 x 1.5 mm2 in-plane resolution and TR of 1.51s. To compare the results obtained for both fields when the same resolution is used, the 9.4T images were downsampled in k-space (reducing the k-space matrtix) to 3T resolution. In order to directly compare the three datasets (original 9.4T, downsampled 9.4T in k-space and the original 3T datasets), both the 9.4T downsampled and the 3T datasets were then upsampled to 1mm (in image space using cubic interpolation).Regions-of-Interest (ROI) of the left and right SC were drawn manually for each subject using anatomical landmarks. Using a custom/made algorithm, the SC was automatically divided in three zones (following the SC anatomy as compared to histology) that we called: superficial zone (SZ), intermediate zone (IZ) and deep zone (DZ)) (Fig.1). Averages of the beta estimates of the GLM model were calculated within each of these zones for both the 3T and the 9.4T datasets.

For the 9.4T data we detected signal increases in the left and right SC in individual whole volume analyses of each participant (Fig. 2). With the 9.4T data the highest signals were significantly localized in the SZ of the contralateral colliculus as expected. There were no signal increases outside the structure (CSF) evidencing the localizing capacity of our technique. For the 3T data there is also a higher value for the SZ compared to the other zones. However, this difference was significantly less clear compared to the results obtained for the 9.4T data. After downsampling the 9.4T data in k-space we still obtain higher beta values than with 3T (Fig.3).We found a substantial benefit of the 9,4T in comparison with the 3T. For the same number of stimulation cycles there is a higher response lateralization and a significative higher depth-dependence of visual signals in the 9.4T data of the SC with highest signal in the SZ in good agreement with its functional architecture known from non-human primates. The results show that the duration of data acquisition and the number of stimulation cycles have a critical impact on the low signal level at the SC at 3T. By downsamplig the 9.4T data in kspace there is a loss in spatial resolution as the difference in signal between the SZ and the IZ reduces considerably compared to the original 9.4T data, but the signal amplitude remains the same (also higher than the 3T data for the same resolution). This result is expected because even for the same reduced resolution of the 3T there is an increase of the blood oxygen level dependent size in the 9.4T. We also showed that upsampling the images for the required voxel size doesn’t bring the same information compared to the data acquired originally with the required resolution (3T upsampled in comparison with the original 9.4T data). In contrast to current 3T standards our 9.4T measurements represent a start point of further considerable improvements of spatial and temporal resolution. These developments will help to close the gap between neurophysiology in animal models and brain activity measurements in humans.