Positive Contrast Susceptibility MR Imaging Using GPU-based Primal-Dual Algorithm.

Area of Science:

• Medical Imaging
• Computational Physics
• Materials Science

Background:

• Metallic interventional devices pose challenges for accurate magnetic susceptibility imaging.
• Existing susceptibility-based positive contrast MR techniques utilize kernel deconvolution with L-1 minimization.
• The first-order primal-dual (PD) algorithm offers faster reconstruction for L-1 minimization problems.

Purpose of the Study:

• To accelerate the PD algorithm for positive contrast magnetic susceptibility imaging of metallic devices.
• To leverage multi-core, multi-thread capabilities of graphics processor units (GPUs) for enhanced computational speed.

Main Methods:

• Implementation of a GPU-based accelerated PD algorithm for susceptibility-based positive contrast MRI.
• Application of the technique to estimate magnetic susceptibility distributions of metallic interventional devices.
• Comparison of computational time and accuracy against a CPU-based scheme.

Main Results:

• The GPU-based PD algorithm achieved comparable accuracy in imaging metallic devices.
• The GPU approach demonstrated a significant speed-up, being 4-15 times faster than the CPU-based method.
• This acceleration enhances the processing efficacy for estimating magnetic susceptibility distributions.

Conclusions:

• GPU acceleration of the PD algorithm significantly improves the efficiency of susceptibility-based positive contrast MRI for metallic devices.
• The developed method offers a faster and effective solution for characterizing magnetic susceptibility distributions.
• This advancement holds clinical relevance for improved imaging of metallic implants and devices.