Electromagnetic Interference Removal in Low-Field MRI via EDITER, Structured Low-Rank Modeling, and Denoising Diffusion Model

Motivation: Low-field MRI often suffers from severe electromagnetic interference (EMI), substantially deteriorating image quality. Existing methods, including EDITER (External Dynamic InTerference Estimation and Removal), struggle to completely eliminate EMI, especially at high EMI levels. Goal(s): To develop a robust framework that enhances low-field MRI by effectively eliminating EMI artifacts and improving signal-to-noise ratio (SNR). Approach: We propose integrating LORAKS, a structured low-rank matrix modeling method, and DDNM, a zero-shot denoising diffusion model, into the EDITER framework. Results: Our method successfully eliminates residual EMI artifacts and improves SNR, leading to substantial enhancements in low-field MRI and thereby overcoming limitations of existing methods. Impact: We propose a novel framework for enhanced low-field MRI by integrating LORAKS and DDNM into EDITER, which effectively eliminates electromagnetic interference (EMI) and improves SNR. The proposed method substantially improves low-field MRI, overcoming limitations of existing methods.