Non-parametric graphnet-regularized representation of dMRI in space and time

Area of Science:

• Biophysics
• Neuroimaging
• Magnetic Resonance Imaging

Background:

• Representing the four-dimensional diffusion MRI signal in q-space and diffusion time (τ) remains a challenge.
• Current methods often require numerous diffusion-weighted images (DWIs) for accurate signal depiction.

Purpose of the Study:

• To develop a novel functional basis approach, termed qτ-dMRI, for representing the dMRI signal in qτ-space.
• To enable the estimation of time-dependent qτ-indices for studying nervous tissue diffusion.

Main Methods:

• Utilized a functional basis approach tailored for qτ-space signal representation.
• Employed GraphNet regularization for signal smoothness and sparsity, reducing the number of required DWIs.
• Validated the method using in-silico Monte-Carlo simulations and in-vivo test-retest studies in mice.

Main Results:

• Demonstrated effective representation of the dMRI signal in qτ-space.
• Achieved good reproducibility of estimated qτ-index values and trends in in-vivo studies.
• Showcased the ability to estimate time-dependent qτ-indices without biophysical assumptions.

Conclusions:

• qτ-dMRI offers a new framework for interpreting the qτ-diffusion signal, specifically designed for τ-dependent analysis.
• This method provides a novel means for studying diffusion in nervous tissue.
• The approach reduces the data requirements for dMRI signal representation.