Progressive conditional diffusion for system matrix synthesis in magnetic particle imaging

Abstract

Acquiring the system matrix (SM) under high-gradient fields in Magnetic Particle Imaging remains challenging, as it requires prolonged high-current operation. To overcome this limitation, we introduce a novel approach that synthesizes the high-gradient SM from its easier-to-obtain low-gradient counterpart using a conditional diffusion model. A progressive conditioning decay strategy is employed to effectively steer the generation of the high-gradient SM. Simulation results demonstrate that our method can successfully synthesize high-gradient SMs, thereby reducing the acquisition burden.