Assessing the reliability of accelerated quantitative susceptibility mapping sequences in memory clinic patients

Alterations of iron and other chemical species in brain tissue, measurable with Quantitative susceptibility mapping (QSM), are associated with various neurodegenerative diseases. QSM relies on 3D multi-echo gradient-echo sequences with long acquisition times, potentially reducing patient comfort and increasing susceptibility to motion artifacts. This study evaluated the reliability of susceptibility maps obtained from an accelerated QSM sequence compared with a conventional protocol in memory clinic patients.

Thirty-six patients underwent a 3 T brain MRI protocol including conventional and accelerated 3D-T1 weighted and QSM acquisitions. The standard QSM acquisition was itself accelerated using parallel imaging SENSE, whereas the accelerated QSM used CS-SENSE; Image preprocessing comprised registration, segmentation, and susceptibility map reconstruction, followed by extraction of magnetic susceptibility metrics from the cerebral cortex and deep grey matter nuclei. Within-subject differences were assessed using the Wilcoxon signed-rank test, while equivalence was evaluated using a paired two one-sided tests (TOST) procedure with a predefined ± 0.025 ppm margin.

Consistency and absolute agreement were assessed using ICC(C,1) and ICC(A,1), respectively, while Bland–Altman analysis and interscan coefficient of variation evaluated agreement and reproducibility. Image similarity was evaluated using the structural similarity index and mean square error. No significant differences were observed between conventional and accelerated acquisitions for any metric.

TOST demonstrated equivalence for most ROI-metric combinations, with exceptions limited to weighted susceptibility measures in the pallidum and putamen. Overall, measurements showed good consistency, low interscan variability, and high image similarity between methods. Sensitivity analyses using alternative QSM map reconstruction and common T1-weighted segmentation yielded consistent results.

These findings support accelerated QSM as a reliable alternative to conventional acquisitions. Although further clinical validation is warranted, the shorter acquisition time may facilitate integration into routine dementia imaging protocols. This work was within the project PNRR-MCNT2-2023-12378268, CUP I63C24000490001, funded by NextGenerationEU within the Program “PNRR Mission 6/Component 2/Investment: 2.1 “Rafforzamento e potenziamento della ricerca biomedica del SSN” of the Italian Ministry of Health and “Ricerca Corrente” program.

The authors gratefully acknowledge the IRCCS Italian Neuroscience and Rehabilitation Network (RIN), promoted by the Italian Ministry of Health, for its valuable contribution in promoting collaborative activities and knowledge sharing across the neuroscience research community. IRCCS SYNLAB SDN, Via Gianturco 113, Naples, 80143, Italy Federica Di Nardo, Riccardo Iandolo, Carlo Cavaliere, Angelica Di Cecca, Sabina Pappatà, Marco Salvatore & Marco Aiello Department of Advanced Biomedical Sciences, Università degli Studi di Napoli Federico II, Naples, 80133, Italy The authors declare no competing interests. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.