Comparison of 3D orientation distribution functions measured with confocal microscopy and diffusion MRI

Abstract

The ability of diffusion MRI (dMRI) fiber tractography to non-invasively map three-dimensional (3D) anatomical networks in the human brain has made it a valuable tool in both clinical and research settings. However, there are many assumptions inherent to any tractography algorithm that can limit the accuracy of the reconstructed fiber tracts. Among them is the assumption that the diffusion-weighted images accurately reflect the underlying fiber orientation distribution (FOD) in the MRI voxel. Consequently, validating dMRI’s ability to assess the underlying fiber orientation in each voxel is critical for its use as a biomedical tool. Here, using post-mortem histology and confocal microscopy, we present a method to perform histological validation of orientation functions in 3D, which has previously been limited to two-dimensional analysis of tissue sections. We demonstrate the ability to extract the 3D FOD from confocal z-stacks, and quantify the agreement between the MRI estimates of orientation information obtained using constrained spherical deconvolution (CSD) and the true geometry of the fibers. We find an orientation error of approximately 6° in voxels containing nearly parallel fibers, and 10-11° in crossing fiber regions, and note that CSD was unable to resolve fibers crossing at angles below 60° in our dataset. This is the first time the 3D white matter orientation distribution is calculated from histology and compared to dMRI. Thus, this technique serves as a gold standard for dMRI validation studies - providing the ability to determine the extent to which the dMRI signal is consistent with the histological FOD, and to establish how well different dMRI models can predict the ground truth FOD.

Neurite dispersion: a new marker of multiple sclerosis spinal cord pathology?

“…Future validation of our preliminary findings could include: analysis of tissue from areas beyond the spinal cord; extension of the histology to the third dimension50, 51; further confirmation from in vivo data; characterization of more complex morphological features of glial cells52; more accurate diffusion MRI signal modeling43, 53; analysis of other quantitative MRI metrics (such as those from relaxometry, magnetization transfer techniques or susceptibility imaging).…”

The VALiDATe29 MRI Based Multi-Channel Atlas of the Squirrel Monkey Brain

“…For example, we anticipate future releases to be augmented with high resolution histological data (Sun et al 2015). This could, for example, be used for validating the accuracy of various diffusion MRI techniques for reconstructing an estimate of local fiber orientation distributions (Choe et al 2012; Schilling et al 2016), or confirming the validity of fiber tractography (Gao et al 2013; Dyrby et al 2007). Incorporation of fMRI, PET, and brain connectivity information is also readily performed.…”

“…Together, this allows acquisition of data with higher signal-to-noise ratios and at a higher resolution compared to in vivo studies. We chose to include ex vivo templates because of their significantly increased contrast, and the additional value of the use of ex vivo and subsequent histology as a means to validate MRI techniques and aid image interpretation (Azadbakht et al 2015; Bastiani et al 2016; Calabrese et al 2015; Choe et al 2012; Knosche et al 2015; Leergaard et al 2010; McNab et al 2009; Schilling et al 2016). …”

The link between diffusion MRI and tumor heterogeneity: Mapping cell eccentricity and density by diffusional variance decomposition (DIVIDE)

“…Human studies are scarce due to the invasive nature of resection and biopsies but initial studies have shown a relation between diffusion anisotropy and tissue microstructure in brain (Ronen et al, 2014), tumor (Szczepankiewicz et al, 2015) and prostate tissue (Bourne et al, 2012). In animals, the investigated features range across structure eccentricity and orientation (Budde and Frank, 2012; Khan et al, 2015; Schilling et al, 2016), neurodegeneration (Jelescu et al, 2016; Jespersen et al, 2010; Kamagata et al, 2016), and axonal diameter (Barazany et al, 2009). However, the link between microscopic tissue heterogeneity and diffusional variance has not yet been studied.…”