Post mortem mapping of connectional anatomy for the validation of diffusion MRI

Abstract

Diffusion MRI (dMRI) is a unique tool for the study of brain circuitry, as it allows us to image both the macroscopic trajectories and the microstructural properties of axon bundles in vivo. The Human Connectome Project ushered in an era of impressive advances in dMRI acquisition and analysis. As a result of these efforts, the quality of dMRI data that could be acquired in vivo improved substantially, and large collections of such data became widely available. Despite this progress, the main limitation of dMRI remains: it does not image axons directly, but only provides indirect measurements based on the diffusion of water molecules. Thus, it must be validated by methods that allow direct visualization of axons but that can only be performed in post mortem brain tissue. In this review, we discuss methods for validating the various features of connectional anatomy that are extracted from dMRI, both at the macro-scale (trajectories of axon bundles), and at micro-scale (axonal orientations and other microstructural properties). We present a range of validation tools, including anatomic tracer studies, Klingler's dissection, myelin stains, label-free optical imaging techniques, and others. We provide an overview of the basic principles of each technique, its limitations, and what it has taught us so far about the accuracy of different dMRI acquisition and analysis approaches.

Fig. 1.. Topographies of axon bundles shown with anatomic tracing vs. dMRI.

The projections of the dorsomedial prefrontal cortex (dmPFC), dorsal anterior cingulate cortex (dACC), and ventromedial prefrontal cortex (vmPFC) / orbitofrontal cortex (OFC) follow a dorsal-to-ventral topographic organization in the internal capsule (IC). This is shown by placing tracer injections in each of the three areas in NHP (a) and replicated by seeding dMRI probabilistic tractography in the three areas, in both ex vivo NHP data (b) and in vivo human data (c). (Adapted from Safadi et al., 2018.)

Fig. 2.. Improved in vivo human dMRI shows greater agreement with NHP anatomic studies.

Probabilistic tractography from a seed region in the frontal pole is shown for in vivo human dMRI datasets acquired with (a) a higher-resolution protocol (b_max = 10 K s/mm², 384 directions, 1.5mm resolution), and (b) a lower-resolution protocol (b = 700 s/mm², 60 directions, 2mm resolution). Based on both anatomic tracing and ex vivo dMRI tractography in NHPs, we expect frontal pole to project through the corpus callosum (CC), internal capsule (IC), and uncinate fasciculus (UF), among other pathways. These projections are present consistently in the higher-resolution human dMRI tractography, while many are missing in the lower-resolution human dMRI tractography. True positives are marked with green arrows; false negatives with yellow arrows. When discrepancies between human dMRI and NHP tracing disappear as dMRI data quality improves, these discrepancies are likely to be dMRI errors and not true inter-species differences.

Fig. 3.

Histological sections from paraffin-embedded human brain samples around the calcarine sulcus containing the stria of Gennari, stained with: a: Hematoxylin & Eosin (HE; 5 μm). Cortical layers are identified as I-VIb. b: Luxol Fast Blue (LFB; 5 μm). c: Bodian (BOD; 6 μm). d: Detail of g. (Adapted from Kleinnijenhuis, 2014.)

Fig. 4.

Fiber orientation maps acquired with 3D-PLI. Left: Entorhinal pathways and the angular bundle in the human hippocampus (reused from Zeineh et al., 2017). Right: 3D fiber architecture of the avian and mammalian primary visual regions (reused from Stacho et al., 2020). Fiber orientations are encoded in HSV color space.

Fig. 5.

Top: PSOCT retardance and optic axis orientation maps of a 15 cm² parasagittal section of the human cerebellum. Bottom: Nissl stain and Gallyas stain from the same sample. (Reused from Wang et al., 2018).