Dynamic Sparse Mask Spatial-Frequency System Matrix Compression for Accelerated Reconstruction in Narrowband Magnetic...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Narrowband magnetic particle imaging (MPI) offers high signal-to-noise ratio (SNR) for large-field-of-view (FOV) human-scale imaging.
• System-matrix (SM)-based reconstruction is crucial for suppressing artifacts in MPI.
• Large FOV scanners result in substantial SM size, causing memory and time constraints.

Purpose of the Study:

• To develop a method for reducing SM size and accelerating reconstruction in large-FOV narrowband MPI.
• To maintain or enhance image quality despite SM compression.

Main Methods:

• Proposed a Spatial-Frequency System Matrix (SF-SM) approach utilizing Discrete Cosine Transform (DCT) for matrix reduction.
• Introduced a Dynamic Sparse Mask (DSM) strategy combining global prior and signal-driven masks to select informative frequency components.
• Extracted a task-adaptive SM by intersecting masks to retain key components and reduce noise.

Main Results:

• The SF-SM approach compressed the SM to approximately 1% of its original size.
• Reconstruction time was accelerated by nearly two orders of magnitude.
• Image quality was maintained or enhanced compared to the Raw-SM, validated across simulations, a public dataset, and an in-house scanner.

Conclusions:

• The DSM-based SF-SM effectively enables significant SM compression and highly efficient reconstruction while preserving image quality.
• This method offers a scalable solution for accelerating narrowband MPI reconstruction, especially for large-FOV systems and real-time applications.