Self-Supervised Iterative Guided Filtering for Sodium MRI Denoising and aTSC Quantification

Guo C, Duan C, Nagel AM, Zhang X, Lou X (2026)

Publication Type: Journal article

Publication year: 2026

Journal

IEEE Transactions on Instrumentation and Measurement Institute of Electrical and Electronics Engineers (IEEE)

Book Volume: 75

DOI: 10.1109/TIM.2026.3697089

Abstract

Sodium (23Na) magnetic resonance imaging (MRI) provides unique metabolic insights through apparent tissue sodium concentration (aTSC) quantification; however, its clinical utility is hindered by an inherently low signal-to-noise ratio (SNR) and the fundamental difficulty of achieving effective noise suppression without compromising quantitative integrity. In this study, we propose a self-supervised iterative guided filtering network (SSIGF) tailored for in vivo 23Na MRI. By leveraging the intrinsic physical properties of the density-adapted 3D radial projection (DA-3DPR) sequence, we extract physically consistent noise from signal-free peripheral slices to construct a statistically faithful noise dataset, bypassing the need for restrictive analytical noise models. To harness this noise-modeling strategy, the SSIGF architecture integrates a learnable guided filtering mechanism that imposes explicit structural constraints, thereby preserving anatomical details and mitigating boundary artifacts. Furthermore, a focal frequency loss (FFL) is incorporated with the spatial L1 loss to enhance sensitivity to structural features and prevent over-smoothing. Extensive evaluations using synthetic phantoms and in vivo data from 119 subjects demonstrate that SSIGF significantly improves image quality and the reliability of aTSC estimates, outperforming state-of-the-art self-supervised and conventional denoising methods. Our results suggest that SSIGF offers a robust, image-domain solution for high-precision quantitative sodium imaging in clinical research.

Authors with CRIS profile

Armin Michael Nagel Institute of Radiology

Involved external institutions

Zhejiang University / Chekiang University / 浙江大学 CN China (CN) People's Liberation Army Hospital / 中国人民解放军总医院 CN China (CN)

How to cite

APA:

Guo, C., Duan, C., Nagel, A.M., Zhang, X., & Lou, X. (2026). Self-Supervised Iterative Guided Filtering for Sodium MRI Denoising and aTSC Quantification. IEEE Transactions on Instrumentation and Measurement, 75. https://doi.org/10.1109/TIM.2026.3697089

MLA:

Guo, Chenlu, et al. "Self-Supervised Iterative Guided Filtering for Sodium MRI Denoising and aTSC Quantification." IEEE Transactions on Instrumentation and Measurement 75 (2026).

BibTeX: Download