A framework for optimal whole-sample histological quantification of neurite orientation dispersion in the human spinal cord.
Grussu F., Schneider T., Yates RL., Zhang H., Wheeler-Kingshott CAMG., DeLuca GC., Alexander DC.
BACKGROUND: The complexity of fibre distributions in tissues is an important microstructural feature, now measurable in vivo by magnetic resonance imaging (MRI) through orientation dispersion (OD) indices. OD metrics have gained popularity for the characterisation of neurite morphology, but they still lack systematic validation. This paper demonstrates a framework for whole-sample histological quantification of OD in spinal cord specimens, potentially useful for validating MRI-derived OD estimates. NEW METHOD: Our methodological framework is based on (i) sagittal sectioning; (ii) Palmgren's silver staining; (iii) structure tensor (ST) analysis; (iv) directional statistics. Novel elements are the data-driven optimisation of the spatial scale of ST analysis, and a new multivariate, weighted directional statistical approach for anisotropy-informed quantification of OD. RESULTS: Palmgren's silver staining of sagittal spinal cord sections provides robust visualisation of neuronal elements, enabling OD quantification. The choice of spatial scale of ST analysis influences OD values, and weighted directional statistics provide OD maps with high contrast-to-noise. Segmentation of neurites prior to OD quantification is recommended. COMPARISON WITH EXISTING METHODS: Our framework can potentially provide OD even in demyelinating diseases, where myelin-based histology is not suitable. As compared to conventional univariate approaches, our multivariate weighted directional statistics improve the contrast-to-noise of OD maps and more accurately describe the distribution of ST metrics. CONCLUSIONS: Our framework enables practical whole-specimen characterisation of OD in the spinal cord. We recommend tuning the scale of ST analysis for optimal OD quantification, as well as neurite segmentation and weighted directional statistics, of which examples are provided herein.