Comparison of basis functions and q-space sampling schemes for robust compressed sensing reconstruction accelerating...

Area of Science:

• Neuroimaging
• Biophysics
• Medical Physics

Background:

• Magnetic Resonance Imaging (MRI) time constraints limit advanced diffusion MRI (dMRI) protocols in clinical and research settings.
• Diffusion Spectrum Imaging (DSI) offers high-resolution intra-voxel microstructure but is time-intensive.
• Compressed Sensing (CS) theory provides a framework to accelerate data acquisition by collecting fewer samples and reconstructing the full dataset.

Purpose of the Study:

• To investigate accelerated Diffusion Spectrum Imaging (DSI) using Compressed Sensing (CS) theory.
• To evaluate different q-space undersampling schemes and CS reconstruction frameworks (Fourier vs. SHORE basis functions).
• To determine optimal acquisition and reconstruction strategies for time-limited DSI studies.

Main Methods:

• Exploration of three q-space undersampling schemes.
• Implementation and comparison of two CS reconstruction frameworks: Fourier-based and SHORE-based.
• Estimation of diffusion, microstructural, and orientational parameters from reconstructed DSI data.
• Validation using simulations, diffusion phantoms, and in vivo data.

Main Results:

• An isotropic distribution of q-space samples was found to be optimal for sparse DSI acquisition.
• Fourier-based CS-DSI demonstrated superior reconstruction performance compared to the SHORE-based approach.
• The developed methods enable robust recovery of DSI data under time constraints.

Conclusions:

• Accelerated DSI using CS is feasible and effective for time-limited studies.
• Fourier-based CS with isotropic sampling is recommended for optimal DSI acquisition and reconstruction.
• This approach facilitates high-quality microstructural analysis in clinical and research settings.