Neurite density imaging versus imaging of microscopic anisotropy in diffusion MRI: A model comparison using spherical...

Area of Science:

• Neuroimaging
• Diffusion MRI
• Biophysics

Background:

• Microscopic diffusion anisotropy in dMRI is crucial for probing elongated structures like neurites.
• Conventional single diffusion encoding (SDE) methods, like NODDI, entangle anisotropy with dispersion and diffusivity, requiring model assumptions.
• These assumptions may limit the accuracy of neurite density estimation.

Purpose of the Study:

• To introduce and validate the constrained diffusional variance decomposition (CODIVIDE) method for quantifying microscopic diffusion anisotropy.
• To compare CODIVIDE with NODDI in estimating neurite properties in healthy volunteers and glioma patients.
• To investigate the impact of b-tensor shape variability on dMRI parameter estimation.

Main Methods:

• CODIVIDE jointly analyzes data from linear tensor encoding (LTE) and spherical tensor encoding (STE).
• Comparison of neurite density (NODDI) and microscopic anisotropy (CODIVIDE) in gray matter, white matter, and gliomas.
• Simulations were used to assess parameter bias in NODDI under varying assumptions.

Main Results:

• NODDI and CODIVIDE showed significant discrepancies in gray matter and gliomas, with NODDI overestimating neurite fraction.
• The NODDI tortuosity assumption was found to be invalid, leading to inconsistent neurite density estimates between LTE and STE data.
• CODIVIDE revealed distinct levels of microscopic anisotropy across brain tissues: high in white matter, intermediate in thalamus/putamen, and low in cortex/gliomas.

Conclusions:

• The NODDI assumption linking neurite density and mean diffusivity is flawed, biasing neurite density mapping.
• Accurate mapping of microscopic diffusion anisotropy necessitates dMRI data acquired with variable b-tensor shapes (LTE and STE).
• CODIVIDE provides a more robust method for characterizing microscopic anisotropy in the brain.