The hippocampus, primarily associated with encoding of episodic memory, is a structure of acute interest in Alzheimer's disease where its loss of volume has become an early marker of the disease and could play an important role in its prognosis. Its complex functional and anatomical organization¹ makes it difficult to study in vivo. Ultra-high magnetic resonance imaging (UHF-MRI) provides access to its inner structure^2,3,4 in vivo, but cannot give access to its fine inner connectivity, due to gradient limitations on clinical systems. Few studies of the inner connectivity of the hippocampus have been done in the past⁵. In this work, we demonstrate that post-mortem UHF-MRI using strong gradients allows to infer the polysynaptic intrahippocampal pathway at high spatial/angular resolutions.

Acquisition - A formalin-fixed post-mortem human temporal lobe was soaked in a phosphate-buffered saline solution for rehydratation before being scanned with a preclinical 11.7T Bruker MRI system equipped with strong gradients (G_max=760mT/m / SR=9500T/m/s) using a dedicated protocol including: a) a MSME sequence performed to evaluate the T2 relaxation time (150μm isotropic;12 echoes;TE=6.40ms-76.8ms; TR=16s; 8 averages; scan time 10h14min); b) a T2-weighted spin-echo sequence depicting the anatomy (200μm isotropic; TE=45ms; TR=1.8s; 9 averages; scan time 9h13min); c) a Pulsed-Gradient-Spin-Echo sequence to collect a high angular resolution diffusion imaging (HARDI) dataset at b=4500 s/mm² along 125 uniformly distributed directions (300μm isotropic; TE=24.2ms; TR=9s; scan time 12h57min).

Segmentation – The segmentation of the hippocampus and of its surrounding structures was performed manually from the T2-weighted anatomical scan using anatomical landmarks and following the strategy described in 2-4. The segmentation was validated by an independant neuroanatomist.

Tractography – The HARDI dataset was used to compute a field of orientation distribution functions using the Sharpening Deconvolution Transform⁶ with a spherical harmonics order 8 and a regularization factor λ=0.006. Then, a streamling regularized deterministic tractography⁷ was launched on the mask corresponding to all the segmented structures, using the Connectomist toolbox⁸ with the following parameters: 8 seeds per voxel, forward step 70μm, maximum solid angle 30°, minimum/maximum fiber lengths 0.5/100mm.

Inference of the polysynaptic circuit– From the obtained tractogram and the mask of segmented regions, the connectivity matrix of regions and their corresponding individual bundles were inferred using Connectomist. The trisynaptic pathwayinvolving the entorhinal cortex-subicular area–sulcus–dentate gyrus–CA3–CA2-CA1–subicular area–alveus–fimbria circuit was finally reconstructed from the individual bundles.

The MSME sequence allowed to establish the histogram of T2 depicting a bimodal distribution showing two peaks at T2(White-Matter)=36.3ms and T2(Grey-Matter)=46.4ms, guiding the choice of TE=45/24.2ms for the anatomical/diffusion sequences respectively (Fig.1). Fig.1 also depicts one diffusion sensitized slice at 4500s/mm² highlighting its high SNR and shows the bimodal histogram of the mean diffusivity with two peaks at ADC(WM)=0.16x10^-9m²/s and ADC(GM)=0.53x10^-9m²/s. The anatomical scan presenting a very good contrast at 11.7T allowed to delineate the inner/surrounding structures of the hippocampus: the entorhinal and perirhinal cortex, the subicular area, CA1, CA3, the dentate gyrus, the sulcus, the stratum radiatum, the alveus, the parahippocampal cortex, the fimbria. Figure 2 gives a 3D rendering of them attributing a specific color to each region. Figure 3 depicts the obtained tractogram superimposed on the 2D fusion of the anatomy and the color-encoded direction map, and shows the inner connectivity of the hippocampus at very high spatial/angular resolutions characterized by trajectories in perfect agreement with the knowledge of its anatomy (eg higher level of connectivity in the head of the hippocampus). Figure 4 shows the connectivity matrix of the substructures of the hippocampus assessing the higher level of connectivity in the head, and figure 5 illustrates how the polysynaptic circuit of the hippocampus could be efficiently extracted using high field / high gradients diffusion MR-based tractography.In this work, we demonstrated that ultra-high field and ultra-high gradient diffusion-weighted MRI offers a unique opportunity to map the inner structural connectivity of the hippocampus and, in the future, to better correlate its atrophy observed at low field in Alzheimer's patients with modifications of its inner connectivity.