Related Experiment Video
Updated: Jan 13, 2026
10:06
High-resolution Functional Magnetic Resonance Imaging Methods for Human Midbrain
Published on: May 10, 2012
13.4K
Repeatability of Susceptibility Source Separation Methods in Human Brain: A Single-Site Study at 3 T
Nashwan Naji1, Peter Seres1, Gerald Moran2
1Department of Radiology and Diagnostic Imaging, University of Alberta, Edmonton, Alberta, Canada.
NMR in Biomedicine
|January 6, 2026
Summary
Susceptibility source separation provides subvoxel brain insights but algorithm choice impacts repeatability. APART and χ-SepNet showed better performance, though reliability varied by brain region and was lower than conventional QSM.
Area of Science:
- Neuroimaging
- Biophysics
- Medical Physics
Background:
- Conventional quantitative susceptibility mapping (QSM) is limited to net susceptibility effects.
- Susceptibility source separation aims to differentiate iron and myelin distribution.
- Repeatability of R2'-based susceptibility separation algorithms requires thorough evaluation.
Purpose of the Study:
- To investigate the repeatability of three R2'-based susceptibility separation algorithms: χ-Separation, χ-SepNet, and APART.
- To assess the reliability and contrast of paramagnetic (χ+) and diamagnetic (χ-) maps generated by these algorithms.
- To compare algorithm performance across different brain regions.
Main Methods:
- Utilized 3-T scan-rescan data from 21 healthy subjects.
- Assessed repeatability using intraclass correlation coefficient (ICC) and repeatability coefficient (RC).
- Evaluated map contrast using average values and compared performance across brain regions.
Main Results:
- Repeatability varied significantly between algorithms and brain regions.
- APART and χ-SepNet demonstrated superior repeatability compared to χ-Separation.
- Paramagnetic (χ+) and diamagnetic (χ-) maps showed moderate to good reliability, generally lower than conventional QSM due to R2' input reliability.
- Repeatability was reduced in the frontal lobe and near air-tissue interfaces.
- Average RCs for APART were 4 ppb (7 ppb in iron-rich regions for χ+), and for χ-SepNet were 5 ppb (10 ppb in iron-rich regions for χ+).
Conclusions:
- Susceptibility source separation offers valuable subvoxel insights into brain iron and myelin.
- Algorithm choice critically influences the repeatability and reliability of susceptibility separation results.
- APART and χ-SepNet present promising options for susceptibility source separation, with moderate to good repeatability in most brain regions.
- Further research may focus on improving the reliability of R2' input for enhanced separation accuracy.
