A greedy regularized block Kaczmarz method for accelerating reconstruction in magnetic particle imaging.

Area of Science:

• Medical Imaging
• Computational Imaging
• Biomedical Engineering

Background:

• Magnetic particle imaging (MPI) is an emerging tomographic modality offering high sensitivity and resolution.
• Current MPI reconstruction methods, like the Kaczmarz algorithm, suffer from high computation times, limiting real-time applications.
• Accelerating MPI reconstruction is crucial for its clinical translation.

Purpose of the Study:

• To develop a novel method for fast and accurate MPI image reconstruction.
• To address the computational limitations of existing Kaczmarz-based algorithms for real-time MPI.

Main Methods:

• A greedy regularized block Kaczmarz (GRBK) method was proposed.
• GRBK incorporates a greedy partition strategy for system matrix preprocessing into well-conditioned blocks.
• A regularized block Kaczmarz algorithm is employed for efficient reconstruction.

Main Results:

• GRBK demonstrated reconstruction speed improvements of orders of magnitude compared to prevalent methods (Tikhonov, Fused Lasso, wavelet-based sparse models).
• Evaluations on simulated and real MPI data (OpenMPIData) showed GRBK outperforms state-of-the-art algorithms in speed and image quality.
• The method maintained accuracy across various noise levels (20-40 dB).

Conclusions:

• The proposed GRBK method significantly enhances MPI reconstruction speed.
• GRBK offers a promising solution for real-time MPI, improving both speed and image quality.
• This method is expected to be pivotal for the clinical adoption of real-time MPI.