Date Published: September 25, 2019
Publisher: Public Library of Science
Author(s): Jose M. Guerrero, Nagesh Adluru, Barbara B. Bendlin, H. Hill Goldsmith, Stacey M. Schaefer, Richard J. Davidson, Steven R. Kecskemeti, Hui Zhang, Andrew L. Alexander, Pew-Thian Yap.
NODDI is widely used in parameterizing microstructural brain properties. The model includes three signal compartments: intracellular, extracellular, and free water. The neurite compartment intrinsic parallel diffusivity (d∥) is set to 1.7 μm2⋅ms−1, though the effects of this assumption have not been extensively explored. This work investigates the optimality of d∥ = 1.7 μm2⋅ms−1 under varying imaging protocol, age groups, sex, and tissue type in comparison to other biologically plausible values of d∥.
Model residuals were used as the optimality criterion. The model residuals were evaluated in function of d∥ over the range from 0.5 to 3.0 μm2⋅ms−1. This was done with respect to tissue type (i.e., white matter versus gray matter), sex, age (infancy to late adulthood), and diffusion-weighting protocol (maximum b-value). Variation in the estimated parameters with respect to d∥ was also explored.
Results show d∥ = 1.7 μm2⋅ms−1 is appropriate for adult brain white matter but it is suboptimal for gray matter with optimal values being significantly lower. d∥ = 1.7 μm2⋅ms−1 was also suboptimal in the infant brain for both white and gray matter with optimal values being significantly lower. Minor optimum d∥ differences were observed versus diffusion protocol. No significant sex effects were observed. Additionally, changes in d∥ resulted in significant changes to the estimated NODDI parameters.
The default (d∥) of 1.7 μm2⋅ms−1 is suboptimal in gray matter and infant brains.
In diffusion weighted magnetic resonance imaging (dMRI), biophysical models are used for relating the dMRI signal to microstructural properties in white and gray matter [1–7]. Neurite orientation dispersion and density imaging (NODDI) , separates the brain tissue microstructure landscape into three compartments: intracellular space or neurites (axons, dendrites), extracellular tissue matrix, and a free water compartment. In spite of its shortcomings, much like the case of other techniques such as diffusion tensor imaging (DTI), NODDI offers useful information and has been widely used in the investigation of brain tissue microstructure as a function of early development, cognitive function and aging as well as a number of neurological conditions [8–13].
The results are organized as follows. (1) We first show how variation in d∥ translates to variability in the estimated parameters. (2) Then, the model RMS residuals, with respect to d∥ are shown to differ between tissue types. (3) This is followed by the presentation of voxel-wise optimized d∥ maps and the ways in which the optimality of d∥ = 1.7 μm2⋅ms−1 is influenced by age, sex, protocol and tissue type.
In this work we studied the implications of diverse multi-shell dMRI data on the optimality of the NODDI parallel intrinsic diffusivity d∥ = 1.7 μm2⋅ms−1. The results suggests model assumptions for d∥ may be suboptimal for specific ages (i.e., infants) and also in gray matter. Although not examined, the optimality of d∥ = 1.7 μm2⋅ms−1 may also vary with pathology. We also observed that suboptimal d∥ leads to biases in the estimated NODDI parameters. Of particular interest would be a drop of neurite density in gray matter, a result that is consistent with findings in a recent study .
In this work, dependence of the estimated NODDI parameters on the parallel intrinsic diffusivity d∥ was observed. Optimum d∥ in white matter of the adult brain is similar to the currently used value d∥ = 1.7 μm2⋅ms−1 but significantly lower in gray matter. Optimal d∥ is also lower than the default value for the newborn brain in white and gray matter. Effects of imaging protocol on the optimum d∥ were also observed. Finally, it is important to consider that, despite its limitations, recent analysis suggests that NODDI metrics provide information that is congruent with histologically equivalent metrics .