Sparse Blind Spherical Deconvolution of diffusion weighted MRI.

Area of Science:

• Neuroimaging
• Biophysics
• Computational Neuroscience

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) is crucial for mapping white matter pathways.
• Accurate characterization of white matter microstructure is limited by current dMRI analysis techniques.
• Spherical deconvolution methods rely on accurate response functions to estimate the Orientation Distribution Function (ODF).

Purpose of the Study:

• To develop and evaluate a blind spherical deconvolution algorithm that does not require prior knowledge of the response function.
• To enable robust estimation of ODF peaks and associated signals in dMRI voxels.
• To assess the performance of the proposed algorithm against state-of-the-art methods for white matter fiber orientation retrieval.

Main Methods:

• Proposed a blind spherical deconvolution algorithm assuming only axial symmetry of the response function.
• Developed methods for estimating ODF peaks and signals without an explicit response function.
• Validated the algorithm using Monte Carlo simulations and real in vivo dMRI data.

Main Results:

• The blind spherical deconvolution algorithm achieved lower angular errors on synthetic data compared to constrained spherical deconvolution.
• Performance on in vivo data was surpassed by existing state-of-the-art spherical deconvolution algorithms.
• The method demonstrated potential for deriving per-voxel, per-direction metrics when combined with advanced direction estimation techniques.

Conclusions:

• The proposed blind spherical deconvolution offers a novel approach to ODF estimation in dMRI.
• Further optimization is needed for superior performance on complex in vivo white matter structures.
• The algorithm shows promise for quantitative microstructure analysis in both simulated and real-world neuroimaging datasets.