Three-Dimensional Magnetic Particle Imaging Resolution Enhancement Method Based on Structured Distillation Contrast Learning

Magnetic Particle Imaging (MPI) is a novel imaging technique for visualizing the spatial distribution of magnetic nanoparticles. Due to variations in gradient field strength and scanning trajectories, MPI resolution shows anisotropy. This paper presents CSDNet, a model based on structured distillation contrast learning. It extracts low-resolution directional features from a two-dimensional isotropic teacher network to guide the training of the student network and improve the resolution in three dimensions through deblurring. The introduced contrast loss significantly improves the ability to extract image details. Experimental results confirm CSDNet’s superior performance in detail recovery and accuracy.