Exploiting structural redundancy in q-space for improved EAP reconstruction from highly undersampled (k, q)-space in...

Area of Science:

• Medical Imaging
• Diffusion MRI Analysis
• Computational Neuroscience

Background:

• Accurate reconstruction of ensemble average propagators (EAPs) from undersampled diffusion MRI (dMRI) is crucial for advanced neuroimaging.
• Compressed sensing (CS) methods accelerate dMRI acquisition by exploiting signal sparsity, but often involve multi-step reconstruction pipelines.
• Existing CS-based EAP reconstruction typically reconstructs diffusion images first, then computes EAPs via Fourier transform.

Purpose of the Study:

• To develop a novel method for direct EAP reconstruction from undersampled (k, q)-space dMRI data.
• To incorporate geometric constraints, specifically the parallelism of level-sets in diffusion images, into the CS framework.
• To enhance reconstruction accuracy and reduce sample complexity for faster dMRI acquisition.

Main Methods:

• Direct reconstruction of the 6D EAP (P(x, r)) from partial (k, q)-space measurements.
• Utilizing geometric constraints based on the parallelism of diffusion image level-sets from proximal q-space points.
• Implementing a compressed sensing framework that directly exploits the incoherence between sensing and reconstruction domains.

Main Results:

• The proposed method demonstrates successful direct reconstruction of EAPs from undersampled dMRI data.
• Exploiting structural similarity (level-set parallelism) in q-space leads to reduced sample complexity.
• Comparative analysis on simulated, phantom, and real dMRI data shows advantages over state-of-the-art CS methods.

Conclusions:

• Direct EAP reconstruction from (k, q)-space data maximizes the benefits of compressed sensing principles.
• Incorporating geometric constraints like level-set parallelism significantly improves dMRI reconstruction efficiency.
• The developed method offers a promising pathway for accelerated and more accurate dMRI acquisition and analysis.