In cooperation with the University of Cambridge (MRC Cognition and Brain Sciences Unit and the Wolfson Brain Imaging Centre, WBIC), we planned and conducted an elaborate memory study using high-resolution, ultra-high field fMRI at 7 Tesla. Over several piloting sessions, we refined an fMRI sequence that would allow us to scan the entorhinal cortex and surrounding medial temporal lobe (MTL) at sub-millimetre resolution, achieving a voxel size of 0.8mm cubed. At only 1.9% of the volume of “standard” 3mm voxels in typical 3 Tesla experiments, this voxel size allows us to tease apart deep and superficial entorhinal cortex layers. In fMRI, individual brain images are not scanned all at once, but slice by slice. Advanced multiband scanning allowed us to scan 3 slices at once, meaning we could scan the entire MTL while still retaining a comparatively high acquisition speed (<1.8s between consecutive images). The anatomical T1- and T2-weighted scans were optimised, returning incredibly detailed images of the MTL structure at a resolution of down to 0.4mm. We also prepared the preprocessing pipeline, an important step prior to statistical analysis. Here, the data are corrected for differences in acquisition time between slices, head movements throughout the fMRI time-series, and image distortions caused by magnetic field inhomogeneities.
Furthermore, we refined the memory task that participants would solve in the MRI scanner. First, participants learned word-object and word-scene pairs. Then, they engaged in a mathematical task that kept them distracted. This phase was included to investigate off-line reactivation of the just-learned content as a key component of consolidation. Finally, during retrieval, participants saw only the words and, via button presses, they indicated whether they could vividly recall the associated object or scene, or not. Hence, the task allows us to investigate brain activity associated with memory for objects and scenes during three memory stages: Encoding, consolidation, and retrieval.
During preparation of the 7T study, we also worked on an existing (previously unpublished) 3T dataset acquired using a similar behavioural task. Here, we could show not only that major input/output regions of the entorhinal cortex are active during object/scene recall – providing an excellent starting point for our 7T study - we also developed a number of data analysis protocols, which included a novel approach testing for perception-retrieval overlap (PRO), indicating correlations across MTL voxels between object/scene perception and object/scene recall.
After meticulous preparation of the 7T study, we invited 27 healthy volunteers between the ages of 18 and 35. Each completed a 7T MRI session in which they solved the memory task while we scanned their brain activity. Afterwards, they completed a number of additional tasks on the computer which measured different aspects of their long-term memory. Results are highly promising. In line with our hypotheses, we could show that the entorhinal cortex, along with other MTL regions, is active during successful recall, and that anterolateral and posteriormedial entorhinal subregions prefer object and scene recall, respectively. Moreover, despite the known difficulties scanning this brain region, we have yielded good quality signal. This is an excellent base for our ongoing analyses on layer and subregional differences.
These first results of the 7T study have been well-received at two talks presented at the Universities of Birmingham and Cambridge. They will be shared with a wider audience at the Neuroscience 2019 conference in Chicago, USA. During the action, we also developed a novel analysis pipeline (PRO) based on the 3T data that has recently been published in eNeuro.
