A robust deconvolution method to disentangle multiple water pools in diffusion MRI.

Area of Science:

• Magnetic Resonance Imaging
• Biomedical Engineering
• Computational Neuroscience

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) signals reflect multiple water diffusion processes.
• Current models often assume a fixed number of diffusion components, leading to biases due to partial volume effects in metrics like DTI and DKI.
• Existing spectral analysis methods for dMRI require high signal-to-noise ratios (SNR) and have limitations in component identification.

Purpose of the Study:

• To develop a robust method for automatic component identification in dMRI spectral analysis.
• To improve the accuracy of diffusion imaging metrics by accounting for partial volume effects.
• To enable reliable dMRI analysis at clinically relevant SNR levels.

Main Methods:

• Developed a method for automatic component identification in dMRI spectral analysis, incorporating outlier rejection and data-driven regularization for noise robustness.
• Applied the method to account for partial volume effects in Diffusion Tensor Imaging (DTI) and Diffusion Kurtosis Imaging (DKI) model fitting.
• Validated the method using numerical simulations and in vivo 3T MRI data.

Main Results:

• The method reliably decomposed three diffusion components from simulated data with SNR=30.
• Biases in DTI and DKI metrics were significantly reduced when accounting for multiple diffusion components.
• Analysis of in vivo data identified three macro-compartments (hindered diffusion, free water, pseudo-diffusion).
• Incorporating free water and pseudo-diffusion in DKI reduced mean diffusivity and increased fractional anisotropy in gray and white matter.

Conclusions:

• The proposed method accurately determines co-existing diffusion compartments without prior assumptions on their number.
• This approach effectively accounts for signal contaminations, improving dMRI analysis at clinically achievable SNR levels.
• The method enhances the reliability and accuracy of diffusion imaging metrics, particularly DTI and DKI.