Efficient Joint Image Reconstruction of Multi-Patch Data Reusing a Single System Matrix in Magnetic Particle Imaging.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Physics

Background:

• Magnetic particle imaging (MPI) is limited by peripheral nerve stimulation, restricting its field of view (FoV) to millimeters.
• To overcome FoV limitations, MPI employs multi-patch acquisition, rapidly scanning larger volumes by shifting the excitation field.
• Joint reconstruction of these patches improves image quality but is computationally intensive due to large, coupled linear systems.

Purpose of the Study:

• Develop a computationally efficient reconstruction algorithm for MPI multi-patch data.
• Address the memory and computational demands of joint reconstruction methods.
• Enable faster and more scalable imaging of larger volumes using MPI.

Main Methods:

• Developed a novel reconstruction algorithm for MPI multi-patch data.
• Exploited the sparsity of the joint system matrix for computational efficiency.
• Utilized a highly efficient implicit matrix format for on-the-fly linear algebra operations.

Main Results:

• The new algorithm reduces computational effort to a linear dependence on the number of patches.
• Successfully reconstructed 3-D multi-patch phantom datasets.
• Achieved reconstruction of large datasets (15 patches) in under 22 seconds.

Conclusions:

• The developed algorithm significantly enhances the efficiency of MPI multi-patch reconstruction.
• This method overcomes previous computational bottlenecks, enabling faster and scalable large-volume MPI.
• Facilitates broader applications of MPI by improving its practical imaging capabilities.