Using graph signal processing in model-based compressive sensing of MRI brain image.

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Biology

Background:

• Magnetic Resonance Imaging (MRI) acquisition is often prolonged, limiting patient tolerance due to the need for multiple breath-holds.
• Compressive Sensing (CS) offers a promising approach to reduce MRI scan times by enabling lower sampling rates.

Purpose of the Study:

• To develop and evaluate a novel graph-based signal model for enhancing Compressive Sensing MRI (CS-MRI) of brain images.
• To improve the quality and efficiency of image reconstruction in CS-MRI by incorporating image-specific structural information.

Main Methods:

• A graph-based signal model was constructed using a general brain MRI template to capture interdependencies between signal elements.
• A convex optimization problem was formulated by integrating image template, node value sparsity, and graph total variation.
• The optimization problem was solved using the Alternating Direction Method of Multipliers (ADMM) applied to a system of Lagrange equations.

Main Results:

• The proposed graph model effectively limits the search space for CS-MRI image recovery.
• Numerical simulations demonstrated that the recovery process achieves higher image quality at a considerably reduced sampling rate.
• The graph model parameters were successfully utilized in the recovery of CS-MRI images.

Conclusions:

• The developed graph-based signal model enhances the precision of CS-MRI brain image reconstruction.
• This approach offers a viable strategy for significantly reducing MRI acquisition time while maintaining or improving image quality.
• The method holds potential for clinical applications where faster MRI scans are desirable.