Comparing repeatability metrics for quantitative susceptibility mapping in the head and neck.

Area of Science:

• Medical Imaging
• Biophysics

Background:

• Quantitative susceptibility mapping (QSM) is a repeatable neuroimaging technique.
• Extending QSM to head and neck (HN) regions requires assessing repeatability metrics.
• Optimizing QSM reconstruction pipelines beyond the brain is essential.

Purpose of the Study:

• To investigate and quantify the repeatability of QSM in the HN region.
• To compare different QSM reconstruction techniques for their repeatability.
• To establish optimal metrics for assessing QSM repeatability in non-brain regions.

Main Methods:

• Acquired MRI data from 10 healthy volunteers in the HN region across two sessions.
• Reconstructed QSM using six state-of-the-art dipole inversion techniques.
• Assessed repeatability using voxel-wise (NRMSE, XSIM) and ROI-based metrics (SD, CV, ICC).

Main Results:

• Within- and between-subject variations were less than variations between QSM methods in most ROIs.
• autoNDI demonstrated the highest repeatability (ICC > 0.75 in 3/6 HN ROIs, average ICC 0.66).
• Combining standard deviation and ICC provided the best metric for comparing QSM method repeatability.

Conclusions:

• QSM repeatability in the HN region is highly dependent on the chosen dipole inversion method.
• The most repeatable methods (autoNDI, QSMnet, TFI) exhibit only moderate repeatability in most HN ROIs.
• Further optimization of QSM pipelines is needed for reliable HN applications.