Ex vivo diffusion MRI of the human brain: Technical challenges and recent advances

Abstract

This review discusses ex vivo diffusion magnetic resonance imaging (dMRI) as an important research tool for neuroanatomical investigations and the validation of in vivo dMRI techniques, with a focus on the human brain. We review the challenges posed by the properties of post-mortem tissue, and discuss state-of-the-art tissue preparation methods and recent advances in pulse sequences and acquisition techniques to tackle these. We then review recent ex vivo dMRI studies of the human brain, highlighting the validation of white matter orientation estimates and the atlasing and mapping of large subcortical structures. We also give particular emphasis to the delineation of layered gray matter structure with ex vivo dMRI, as this application illustrates the strength of its mesoscale resolution over large fields of view. We end with a discussion and outlook on future and potential directions of the field.

Keywords: cortical layers; diffusion MRI; ex vivo; gray matter; validation; white matter.

Figure 1

The multiscale nature of human structural brain connectivity and its measurement with different techniques. The measurement of the connectivity phenomena here refers to features directly resolved by the acquired spatial resolution of the technique (not by modeling the contrast over multiple measurements and indirectly inferring statistics of such features, as in microstructure modeling of diffusion MRI). dMRI, diffusion MRI; EM, electron microscopy; LM, light microscopy

Figure 2

Diffusion‐weighted radiofrequency (RF) pulse sequences. (A) Diffusion‐weighted spin echo (dwSE) or pulsed‐gradient spin‐echo. (B) Diffusion‐weighted stimulated echo acquisition mode (dwSTE). (C) Diffusion‐weighted steady‐state free precession (dwSSFP). For each sequence, the echo time (TE), diffusion time (Δ) and periods over which T ₁, T ₂ or combined T ₁ and T ₂ decay occurs are indicated

Figure 3

An example showing amplitude and phase modulations in k‐space and navigator echo‐based phase correction for three‐dimensional (3D) diffusion‐weighted gradient and spin echo (GRASE) readout. (A) T ₂‐ and T ₂*‐dependent amplitude modulation along K _y and K _z for ‘center‐out’ k‐space sampling with echo planar imaging (EPI) factor = 3 and rapid acquisition with relaxation enhancement (RARE) factor = 4. (B) Corresponding phase modulation in k space as a result of off‐resonance spins and phase oscillations between odd‐ and even‐numbered echoes. (C) Resulting point spread function (PSF). (D) Diffusion‐weighted image from a post‐mortem human hippocampus before and after phase correction. Phase discontinuities in k space result in ghosting artifacts (arrows) in diffusion‐weighted images, which are minimized by twin‐navigator echo phase correction. (Aggarwal et al.43)

Figure 4

Validation of orientation estimates in human cortex. (A) Gray matter (A.a) and white matter (A.b) locations with multiple fiber orientations estimated with structure tensor analysis (STA) of a myelin stain (orientation color coded between red = horizontal and blue = vertical) and the diffusion tensor (green box: axes of largest diffusion projected into the section plane).58 In (A.a), the orientation of the main tensor axes (green box) is well aligned with the main radial STA (purple) in the deeper layers (orange arrow), but tensor orientation accuracy is less in cortical layers with both radial and tangential intracortical fibers (blue arrow). In (A.b), the orientation of the main tensor axes (green box) is misaligned with the STA orientations (red and blue). Two‐dimensional (2D) fiber orientation distributions (FODs) from STA of a myelin stain of human V1 superimposed on an anisotropy‐weighted 2D orientation color map (orientation color coded between dark red = horizontal/tangential and green = vertical/radial) (B) and its corresponding myelin stain (C), with six myeloarchitectural layers labeled from superficial layer 1 (bottom) to deep layer 6 (top).55 (Reproduced with permission from Budde and Annese55 and Seehaus et al.58)

Figure 5

Validation of tractography. (A) The effect of resolution on diffusion tensor imaging (DTI) streamline tractography in the human optic chiasm from downsampled isotropic resolutions: 312.5 μm, left; 625 μm, middle; 1250 μm, right.67 (B) An illustration of two‐dimensional (2D) histological section (micrographs with green and red color indicating carbocyanine dye) alignment in the human temporal lobe to diffusion magnetic resonance imaging (dMRI) data of the same tissue for validation [primary DTI eigenvectors rendered as three‐dimensional (3D) cylinders with orientation color coding (red, left–right; green, anterior–posterior; blue, inferior–superior)].68 (Reproduced with permission from Roebroeck et al.67 and Seehaus et al.68)

Figure 6

Structural organization of the human basis pontis as resolved with ex vivo high angular resolution diffusion imaging (HARDI) at 11.7 T.39 (A) Coronal slice through the fractional anisotropy (FA) map. (B) Fiber orientation distributions (FODs) reconstructed from constrained spherical deconvolution in the region corresponding to the white box in (A). Red, green and blue in color‐coded FODs indicate anterior–posterior, medial–lateral and inferior–superior orientations, respectively. The FOD map reveals interdigitating corticospinal fascicles (blue) and transverse pontine fibers (green) projecting from the pontine nuclei. (C) Zoomed‐in view of FODs in a small region (within the white box in B) shows the intravoxel crossing fiber orientations resolved with HARDI. Scale bar, 250 μm. (Data from Aggarwal et al.39)

Figure 7

Laminar structure of the human visual cortex revealed with ex vivo high angular resolution diffusion imaging (HARDI). (A) A polarized light imaging (PLI) section of the primary visual area overlaid with diffusion magnetic resonance imaging (dMRI)‐derived fiber orientation distributions (FODs) (left) and intracortical fiber tracts from streamline tracking (right). Arrows indicate tangential fibers in the stria of Gennari. (B) Delineation of cortical layers and the marked transition between primary (V1) and secondary (V2) visual areas seen with dMRI at 11.7 T. Distinct intracortical layers corresponding to the stria of Gennari and inner band of Baillarger in V1 are evident in the direction‐encoded color (DEC) map, which shows excellent agreement with the cortical myeloarchitecture seen in silver‐stained sections adjacent to the V1–V2 boundary in the same specimen. The fractional anisotropy (FA) profile across the cortical depth is shown at the top right. T2w, T ₂‐weighted. Scale bar, 0.5 mm. (Reproduced with permission from Aggarwal et al.45 and Leuze et al.81)