
==== Front
Neuroimage Clin
Neuroimage Clin
NeuroImage : Clinical
2213-1582
Elsevier

S2213-1582(24)00037-8
10.1016/j.nicl.2024.103598
103598
Review
Quantitative susceptibility mapping in multiple sclerosis: A systematic review and meta-analysis
Voon Cui Ci ab
Wiltgen Tun ab
Wiestler Benedikt c
Schlaeger Sarah c
Mühlau Mark mark.muehlau@tum.de
ab⁎
a Dept. of Neurology, School of Medicine and Health, Technical University of Munich, Munich, Germany
b TUM-Neuroimaging Center, School of Medicine and Health, Technical University of Munich, Munich, Germany
c Dept. of Neuroradiology, School of Medicine and Health, Technical University of Munich, Munich, Germany
⁎ Corresponding author at: Department of Neurology, School of Medicine, Technical University of Munich, Ismaninger Str. 22, D-81675 Munich, Germany. mark.muehlau@tum.de
25 3 2024
2024
25 3 2024
42 10359821 12 2023
7 3 2024
24 3 2024
© 2024 The Authors
2024
https://creativecommons.org/licenses/by-nc-nd/4.0/ This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Graphical abstract

Highlights

• Systematic review on magnetic susceptibility values (QSM values) in MS.

• Compared to controls, QSM values are increased in the basal ganglia but reduced in the thalamus.

• QSM values of white matter lesions change over time, increasing first, plateauing later, and declining finally.

• High methodological heterogeneity across studies.

Background

Quantitative susceptibility mapping (QSM) is a quantitative measure based on magnetic resonance imaging sensitive to iron and myelin content. This makes QSM a promising non-invasive tool for multiple sclerosis (MS) in research and clinical practice.

Objective

We performed a systematic review and meta-analysis on the use of QSM in MS.

Methods

Our review was prospectively registered on PROSPERO (CRD42022309563). We searched five databases for studies published between inception and 30th April 2023. We identified 83 English peer-reviewed studies that applied QSM images on MS cohorts. Fifty-five included studies had at least one of the following outcome measures: deep grey matter QSM values in MS, either compared to healthy controls (HC) (k = 13) or correlated with the score on the Expanded Disability Status Scale (EDSS) (k = 7), QSM lesion characteristics (k = 22) and their clinical correlates (k = 17), longitudinal correlates (k = 11), histological correlates (k = 7), or correlates with other imaging techniques (k = 12). Two meta-analyses on deep grey matter (DGM) susceptibility data were performed, while the remaining findings could only be analyzed descriptively.

Results

After outlier removal, meta-analyses demonstrated a significant increase in the basal ganglia susceptibility (QSM values) in MS compared to HC, caudate (k = 9, standardized mean difference (SDM) = 0.54, 95 % CI = 0.39–0.70, I2 = 46 %), putamen (k = 9, SDM = 0.38, 95 % CI = 0.19–0.57, I2 = 59 %), and globus pallidus (k = 9, SDM = 0.48, 95 % CI = 0.28–0.67, I2 = 60 %), whereas thalamic QSM values exhibited a significant reduction (k = 12, SDM = −0.39, 95 % CI = −0.66–−0.12, I2 = 84 %); these susceptibility differences in MS were independent of age. Further, putamen QSM values positively correlated with EDSS (k = 4, r = 0.36, 95 % CI = 0.16–0.53, I2 = 0 %). Regarding rim lesions, four out of seven studies, representing 73 % of all patients, reported rim lesions to be associated with more severe disability. Moreover, lesion development from initial detection to the inactive stage is paralleled by increasing, plateauing (after about two years), and gradually decreasing QSM values, respectively. Only one longitudinal study provided clinical outcome measures and found no association. Histological data suggest iron content to be the primary source of QSM values in DGM and at the edges of rim lesions; further, when also considering data from myelin water imaging, the decrease of myelin is likely to drive the increase of QSM values within WM lesions.

Conclusions

We could provide meta-analytic evidence for DGM susceptibility changes in MS compared to HC; basal ganglia susceptibility is increased and, in the putamen, associated with disability, while thalamic susceptibility is decreased. Beyond these findings, further investigations are necessary to establish the role of QSM in MS for research or even clinical routine.

Keywords

Systematic review
Meta-analysis
Multiple sclerosis
Quantitative susceptibility mapping
White matter lesion
Brain atrophy
Abbreviations

ARR annualized relapse rate

CI confidence interval

CIS clinically isolated syndrome

DD disease duration

DGM deep grey matter

EDSS expanded disability status scale

FLAIR fluid-attenuated inversion recovery

g Hedges’ g

Gd Gadolinium

GdE gadolinium-enhancing lesion

GRE gradient recalled echo

HC healthy control

I2 I2 heterogeneity

k number of studies

MRI magnetic resonance imaging

MS multiple sclerosis

NAWM normal-appearing white matter

PMS progressive multiple sclerosis

ppm parts-per million

PPMS primary progressive multiple sclerosis

Q Q-statistics

QSM quantitative susceptibility mapping

r correlation coefficient

RRMS relapsing-remitting multiple sclerosis

SD standard deviation

SMD Hedges’ g standardized mean difference

SPMS secondary progressive multiple sclerosis

SWI susceptibility weighted imaging

T Tesla

T1w T1 weighted

T2w T2 weighted

TLV total lesion volume

WM white matter

WML white matter lesion

β beta-coefficient
==== Body
pmc1 Introduction

Multiple sclerosis (MS) is a progressive neuroinflammatory and neurodegenerative disease characterized by the formation of demyelinated lesions and later neurodegeneration in the central nervous system (Lassmann et al., 2007). Focal lesions seen in magnetic resonance imaging (MRI) are commonly used to diagnose MS, initiate treatment, and monitor disease activity (Filippi et al., 2019, Rovira et al., 2015, Thompson et al., 2018). However, conventional MRI measures, such as lesion load on T2-weighted sequences (T2w), are only weakly associated with MS clinical outcomes (Barkhof, 1999, Valizadeh et al., 2021). To better understand and monitor pathophysiological changes in MS, new (quantitative) MRI sequences have been proposed. A promising one of those is quantitative susceptibility mapping (QSM).

Previous reviews have consistently described QSM as a non-invasive measure to investigate microstructural damage and neuroinflammation in MS through estimating concentration changes in iron, involved in neurodegeneration, and myelin, the primary target of acute inflammatory activity in MS (Alexander et al., 2011, Deistung et al., 2017, Granziera et al., 2021, Reichenbach et al., 2015). QSM allows for the quantification of magnetic susceptibility within an imaging voxel (Liu et al., 2015). Quantitative susceptibility maps can be reconstructed through GRE's phase images, assuming that phase shift is mainly driven by susceptibility-induced field inhomogeneity (Liu et al., 2015). The sophisticated processing steps of QSM involve estimating the field distribution from the phase image and removing background field contributions originating from non-brain regions. Finally, the process includes solving the inverse problem to determine magnetic susceptibility (Reichenbach et al., 2015, Wang and Liu, 2015). Due to the removal of the background phase, QSM does not provide absolute values but values relative to a reference region within the target area. This relative susceptibility value (QSM value), often expressed through parts per million (ppm), approximates the overall paramagnetic and diamagnetic effects within a voxel or brain region. In other words, brain regions dominated by paramagnetic components, such as iron and deoxyhemoglobin, show higher QSM values and appear brighter compared to surrounding tissues. In contrast, regions that are dominated by diamagnetic components, such as myelin, show decreased QSM values and appear darker on QSM compared to surrounding tissues.

Histological findings in healthy aging have shown a positive correlation between QSM values and iron content. This positive correlation is stronger in deep grey matter (DGM) than in white matter (WM) (Langkammer et al., 2012). While iron is deemed a dominant susceptibility source in DGM, the primary susceptibility source in WM is less obvious, especially in MS, as both neuroinflammatory and neurodegenerative processes occur that potentially alter the content and ratio of paramagnetic iron and diamagnetic myelin. Moreover, the confounding effects of fiber-to-field orientation decrease the reliability of QSM values in inferring iron/myelin content in highly myelinated fiber bundles (Lancione et al., 2017; Li et al., 2012).

The use of QSM values to assess disrupted iron homeostasis within the central nervous system has sparked interest in investigating QSM’s clinical utility in MS. Studies have demonstrated increased QSM values in the basal ganglia in MS (Langkammer et al., 2013, Zivadinov et al., 2018) compatible with the idea of a marker of early neurodegeneration in MS. Along the same line, DGM QSM values have shown a positive correlation with Expanded Disability Status Scale (EDSS) scores (Langkammer et al., 2013) in contrast to other susceptibility-based imaging, e.g., R2*. Other than QSM values in DGM, evidence points to increased QSM values in the normal-appearing white matter (NAWM) of MS patients compared to the WM of healthy controls (HC) (Yu et al., 2019). This suggests the prognostic potential of QSM values through detecting abnormal iron hemostasis or changes in myelin possibly related to chronic microglial activation (Lassmann et al., 2007). Of note, a recent study even suggested differential effects of disease-modifying therapies on QSM measures by demonstrating that rim lesion QSM values significantly decreased under dimethyl fumarate treatment in comparison to glatiramer acetate. Notably, the reduction in QSM values of rim lesions correlated with decreased microglial cell activation states, underscoring QSM’s potential as an outcome measure for treatment monitoring (Zinger et al., 2022).

Other MRI modalities, such as myelin water fraction (MWF), offer additional insights into the underlying pathophysiological states, such as myelin integrity. Recent research by Rahmanzadeh et al. (2022) reported that more than half of the initially hyperintense QSM lesions, which later turned iso- and hypointense, showed significantly increased MWF values during the two-year follow-up. This observation suggests the occurrence of remyelination within these iso- and hypointense lesions, which was substantiated histologically. Furthermore, gadolinium-enhanced T1-weighted imaging (Gd-enhanced T1w) offers valuable insights, specifically in the context of newly developed lesions and their corresponding QSM values. Studies combining Gd-enhanced T1w and QSM images have reported significantly lower QSM values in Gd-enhancing lesions when compared to non-enhancing lesions (Zhang et al., 2016b, Zhang et al., 2016c). The reduction in QSM values observed in Gd-enhancing lesions has led Zhang et al., (2016c) to propose QSM as a potential tool for detecting newly formed lesions without the need for gadolinium-based contrast agent.

Given the potential of QSM for MS research and clinical practice, coupled with the plethora of MS research utilizing QSM, there is a lack of systematically reviewed evidence. This review aims to address this gap by achieving the following objectives: 1) summarizing and evaluating QSM values in different brain regions, including lesions, 2) examining the clinical correlates of QSM, 3) investigating the prognostic value of QSM through longitudinal studies, 4) exploring other imaging correlates of QSM, and 5) assessing the histological correlates of QSM.

This review is confined to brain MRI as the QSM algorithm for the spinal cord is yet to be developed (Granziera et al., 2021). Neither does this review cover the central vein sign (CVS), since, according to the Consensus Statement by North American Imaging in Multiple Sclerosis Cooperative (NAIMS), the hypointense CSV is best detected by T2*-based MRI, which are T2*-weighted images, by susceptibility-weighted imaging (SWI), or by combining FLAIR and T2* images (Sati et al., 2016). Moreover, the pooled incidence and diagnostic performance of CVS using the recommended sequences have already been systematically reviewed and meta-analyzed by Suh et al. (2019).

2 Methods

This systematic review and meta-analysis of QSM in multiple sclerosis was registered on PROSPERO (registration number: CRD42022309563) and was reported following PRISMA guidelines (Moher et al., 2016).

2.1 Selection of studies

A systematic literature search on PubMed, Science Direct, Scopus, Web of Science, and Wiley Online Library was conducted for publications in the English language until 30th April 2023. The search string ((“multiple sclerosis” OR “clinically isolated syndrome” OR “radiologically isolated syndrome”) AND “quantitative susceptibility mapping”) was applied to all databases.

Publications were included if they met the following criteria: 1) original peer-reviewed publication written in English; 2) participants with MS; 3) participants who underwent an MRI scan with QSM image reconstruction; 4) provision of sufficient information on QSM-related findings. Thus, case reports, reviews, posters, book sections, publications in other languages, and publications without relevant outcome measures were excluded.

Two researchers (CV and MM) independently screened articles and decided on eligible studies using Covidence (Veritas Health Innovation, 2023), a screening and data extraction tool that automatically removes duplicate references; another four duplicates were later identified through full-text screening by CV and MM. Any disagreement between researchers was resolved through discussion.

2.2 Data extraction

The following characteristics were extracted from the full text: first author, publication year, country, sample size, characteristics of participants, magnetic field strength, QSM reconstruction approach, and outcome measures of QSM in MS.

Diverse quantitative findings were extracted and were grouped into the following domains of correlation of QSM in MS: 1) lesion characteristics (lesion counts and lesion types); 2) QSM values in brain regions or QSM lesions; 3) clinical severity; 4) longitudinal design (aiming at any prognostic properties of QSM); 5) other MRI correlates; 6) histopathology.

Previous studies have delineated different lesion characteristics of QSM respective to their contrast or locations in the brain (Cronin et al., 2016, Li et al., 2016). Lesions are named differently across studies even though the studies refer to the same type of lesion, e.g., rimmed lesion (Harrison et al., 2016) and paramagnetic rim lesion (Meaton et al., 2022). For clarity, the following conventions of QSM lesions are applied throughout this review. First, bright lesions on QSM that have higher QSM values than the neighboring normal-appearing white matter (NAWM) are termed hyperintense QSM lesions. Second, dark lesions on QSM with lower QSM values than the neighboring NAWM are termed hypointense QSM lesions. Third, lesions that have been detected on other MRI images but are non-apparent on QSM are termed isointense QSM lesions. Fourth, rim lesions are characterized by their hyperintense or hypointense rim at the edge of the lesion on QSM images. Fifth, Gadolinium (Gd)-enhancing lesions with ring-like pattern, independent of their appearance on QSM, are termed rim Gd-enhancing lesions. Sixth, susceptibility values derived from QSM are referred to as QSM values.

2.3 Data synthesis and statistical analysis

2.3.1 Data summary and synthesis

The extracted outcome measures were synthesized quantitatively or qualitatively. When the number of studies was sufficient (k ≥ 3), we aimed to conduct meta-analyses (section 2.3.3). However, despite many studies reporting the findings on clinical correlates of QSM lesions, meta-analysis was not feasible because of the varying clinical outcome measures and statistical tests employed across studies. Therefore, vote counting, as described in Vancampfort et al. (2012), was used to synthesize the findings on clinical correlates of QSM lesions. This way, studies could be included if they provided information on the direction of effect (i.e., positively or negatively associated) and the state of statistical significance (i.e., the p value was provided). Outcome measures with an insufficient number of studies (k < 3) were reported in tables, visual graphs, or summarized narratively.

2.3.2 Duplication of samples

To avoid multiple data entries, we contacted two authors of studies with longitudinal designs and follow-up publications (i.e., Elkady and Hagemeier). Four studies were removed from the meta-analysis comparing QSM value differences in MS and HC, as author(s) confirmed through email correspondence the recruitment of overlapping samples in their follow-up studies (Elkady et al., 2019, Elkady et al., 2018, Hagemeier et al., 2018b, Schweser et al., 2021); thus, only data from baseline studies were included in the meta-analysis (Elkady et al., 2017, Hagemeier et al., 2018a). Nonetheless, Elkady et al. (2019) and Hagemeier et al., (2018b) were included in the second meta-analysis, as these two studies reported correlational findings on DGM susceptibility and EDSS in their respective sample cohort. Lastly, seven studies were likely to have recruited samples from the same database (Chen et al., 2017, Chen et al., 2014, Huang et al., 2022, Zhang et al., 2019, Zhang et al., 2016a, Zhang et al., 2016b, Zhang et al., 2016c), risking bias due to multiple inclusions of the same patients; those studies were summarized descriptively.

2.3.3 Meta-analysis

Two major meta-analyses were conducted using meta package (Balduzzi et al., 2019) in R-4.2.0 (R Core Team, 2021). The first meta-analysis examined the difference in QSM values between MS and HC groups in different DGM regions using random effects Hedges’ g standardized mean difference (SMD). SMD was calculated using the mean difference between the QSM values of the MS and HC groups divided by the pooled standard deviation (SD) of the two groups, while random effects models were used to account for the variances resulting from between-study heterogeneity (Harrer et al., 2022). In addition, if a study provided more than one mean QSM value and SD from different MS subgroups (i.e., clinically isolated syndrome (CIS), relapsing-remitting MS (RRMS), and progressive MS (PMS)), we combined all the means and SDs using the formulae provided in the Cochrane’s Handbook for Systematic Review (Higgins et al., 2021). This procedure ensured that, in the main meta-analysis, MS and HC groups from each study were represented by only one mean and one SD.

The second meta-analysis investigated the correlation between DGM susceptibility and the score on the Expanded Disability Status Scale (EDSS), which is the most established clinical score in MS with values ranging from zero, indicating a normal neurological examination, to 10, indicating death from MS (Kurtzke, 1983). Pearson’s correlation coefficient was used as a common effect estimate. Studies that reported Spearman’s correlation coefficient were also included in the meta-analysis as both Spearman’s and Pearson’s correlation coefficients do not differ substantially (Gilpin, 1993). The correlation coefficients were pooled using a random-effects model to weigh different studies based on their sample size. Each correlation coefficient was first transformed into Fisher’s z coefficient, then z values were weighted with the inverse of the variance of the correlation coefficient. Finally, the pooled Fisher’s z coefficient was converted back into a pooled estimate of Pearson’s correlation coefficient (Cooper et al., 2019). As combining multiple r values of different subgroups (e.g., MS subgroups) is impossible without assessing the whole dataset of a study (Charter and Alexander, 1993), total sample correlation was preferred when both total sample and MS subgroup correlations of QSM values and EDSS were provided. However, when studies only reported correlations based on MS subgroups, only the r value derived from the largest subgroup (i.e., RRMS) was included in the meta-analysis.

Sources for heterogeneity were assessed using Q-statistic and the I2 value. If the Q-statistic is statistically significant (p < 0.05), it signifies the presence of heterogeneity (Huedo-Medina et al., 2006). An I2 larger than 50 % indicates substantial heterogeneity, while I2 lower than 50 % indicates that heterogeneity is unlikely to be problematic (Higgins and Thompson, 2002). Potential publication bias was assessed by funnel plots and Egger’s regression. As SMDs may increase false positives in Egger’s regression, the corrected standard error was used as the predictor in the model (Harrer et al., 2022).

2.4 Quality assessment

An eight-item checklist for risk of bias assessment was customized by CV and MM to assess the quality of studies included in the meta-analyses. Most items from the list were adapted from the Cochrane risk of bias tool for randomized trials 2 (Sterne et al., 2019) and the Downs and Black checklist for randomized and non-randomized studies (Downs and Black, 1998). Additional items were included to assess the quality of the QSM post-processing and image segmentation procedure, while non-applicable domains and items from the two established assessment tools, such as randomization and blinding of participants, were excluded.

The risk of bias assessment checklist includes the following items: 1) Was QSM reconstruction procedure clearly stated? 2) Was an established DGM segmentation approach used? 3) Were the characteristics of participants clearly described? 4) Was the sample size adequate (≥50)? 5) Did any participant withdraw, or was data excluded from the analysis? 6) Was the interval between MRI and EDSS assessment within 12 months? 7a). Was the measure outcome (M and SD) valid and reliable? (first meta-analysis comparing DGM susceptibility between MS and HC) 7b). Was the measure outcome (r) valid and reliable? (second meta-analysis correlating DGM susceptibility and EDSS scores) 8). Were key confounders (e.g., age and sex) controlled? Items were scored as low risk, some concerns, and high risk. Two researchers (CV, MM) independently assessed the risk of bias, and any disagreement was resolved by discussion.

3 Results

3.1 Characteristics of included studies

After removing duplicates, our systematic search identified 868 publications. Seven hundred twenty were excluded by screening the title or abstract. Of the remaining 148, full-text manuscripts were reviewed, and finally, 55 studies with 3582 MS patients were included in the review. Fig. 1 illustrates the PRISMA flow chart and screening procedures for study inclusion. An overview of all included studies and study characteristics is given in Inline Supplementary Tables 1 and 2. In addition, Appendix 1 offers a summary of the cohort characteristics and QSM reconstruction techniques employed by the included studies. The frequency of outcome measures of the included studies is presented in Fig. 2.Fig. 1 PRISMA chart flow of study selection process.

Fig. 2 Outcome measures: absolute frequency and percentage (with multiple counts) across studies (n = 55).

3.2 Meta-analyses on DGM susceptibility

3.2.1 Included studies and risk of bias

A total of 17 studies, comprising 1903 MS patients and 900 HC, were eligible for meta-analysis. Two primary meta-analyses were performed: one on the QSM values difference in different DGM regions in MS and HC (k = 13), and one on the correlation between DGM susceptibility and EDSS (k = 8).

The bias assessment found more than 80 % of studies (k = 14–17 of 17, and 13 of 15, respectively) to be low risk of bias for the questions regarding sample size, patient characteristics, reliability on reporting outcome measures, transparency on reporting participant exclusion, as well as DGM segmentation methods. However, there were three studies rated with a high risk of bias due to small sample size (Hamdy et al., 2022, Rudko et al., 2014, Schmalbrock et al., 2016), and two were rated high risk due to selective reporting in the mean and standard deviation of QSM values. The QSM values reported were either limited to the thalamus (Pontillo et al., 2017) or had to be estimated from a graph (Fujiwara et al., 2017).

Two items showed moderate concerns on the reporting of the interval between MRI and EDSS and the clarity in describing the QSM reconstruction pipeline. About 47 % to 82 % of these studies were rated with some concerns, as they did not report the interval (k = 14 of 17) (Bergsland et al., 2018; Burgetova et al., 2017; Cho et al., 2022, Cobzas et al., 2015, Elkady et al., 2019; Hagemeier et al., 2018a, Hagemeier et al., 2018b Hamdy et al., 2022, Langkammer et al., 2013, Pontillo et al., 2019; Pudlac et al., 2020; Rudko et al., 2014, Schmalbrock et al., 2016, Schweser et al., 2018) and/or they did not specify the reference brain region to which QSM value was related (k = 8 of 17) (Burgetova et al., 2017, Elkady et al., 2017, Elkady et al., 2019, Fujiwara et al., 2017, Hamdy et al., 2022, Langkammer et al., 2013, Pontillo et al., 2019, Schmalbrock et al., 2016). Additionally, two other items showed high-risk proportions ranging from 23 % to 38 %. This was observed in studies that did not report correlational coefficients when reporting non-significant relationships between QSM values and EDSS (k = 3 of 8) (Elkady et al., 2019, Hagemeier et al., 2018a, Hagemeier et al., 2018b) and/or in studies that did not control for key confounders (i.e., age and sex) (k = 4 of 17) (Hamdy et al., 2022, Langkammer et al., 2013, Pontillo et al., 2019, Schmalbrock et al., 2016). Further sensitivity tests were conducted to address the potential bias because of the underreported non-significant findings and the lack of control on key confounders (see Appendix 2 for detailed ratings per study).

3.2.2 Results of meta-analyses on DGM susceptibility

Thirteen studies (N MS = 1640, N HC = 833) were included in the meta-analyses, comparing the mean difference in DGM susceptibility between MS and HC. Among the 13 studies, 12 reported mean QSM values and SDs, one provided values in graphs from which mean and SD values were estimated (Fujiwara et al., 2017). The studies included in the meta-analysis are marked in Inline Supplementary Table 1 on study characteristics.

To ensure comparability between studies, effect estimates of different DGM subregions, namely thalamus, caudate, putamen, and globus pallidus, were assessed separately. In addition, subgroup analyses were conducted on the RRMS subtype, while subgroup analyses for CIS and PMS were excluded due to the small sample size (k < 3). In the RRMS subgroup analysis, we included seven studies that provided data on thalamic QSM values and six studies that provided data on basal ganglia QSM values. The main meta-analyses on DGM difference between MS and HC reported a significantly higher susceptibility in the caudate (g = 0.68, 95 % CI = 0.25–1.11, p < 0.01) and pallidum (g = 0.63, 95 % CI = 0.14–1.13, p = 0.01), while marginally higher susceptibility was reported in the putamen for MS (g = 0.72, 95 % CI = − 0.03–1.47, p = 0.06). No significant difference in QSM values between MS and HC in the thalamus was found (g = −0.11, 95 % CI = −0.70–0.48, p = 0.72) (see forest plots in Appendix 3). Study heterogeneity for all four meta-analyses was substantial for all pooled effect sizes (I2 between 78 % and 91 %). Although some funnel plots suggested asymmetry, none of the Egger’s tests yielded significant results (see Appendix 4).

The findings of main meta-analyses yielded one outlier with an effect size larger than three (Rudko et al., 2014). Fig. 3 illustrates the sensitivity analysis after the removal of this outlier, resulting in a significant decrease in thalamic QSM values in MS, despite a high between-study heterogeneity, with I2 = 84 %. Compared to the findings of main meta-analysis, the pooled effect sizes decreased slightly in basal ganglia subregions (caudate, putamen, and globus pallidus) but remained significant, and the between-study heterogeneity dropped substantially (Δ I2 = 22–32 %). As the meta-analytical findings show sensitivity to the outlier study (Rudko et al., 2014), it was excluded from the remaining sensitivity tests.Fig. 3 Forest plots of sensitivity analysis after outlier removal, illustrating the difference in QSM values between MS and HC in different DGM subregions. After removing the outlier study (Rudko et al., 2014), the findings illustrated consistent increase in QSM values across all basal ganglia regions (caudate, putamen, and globus pallidum) and a significant decrease in thalamic susceptibility in MS. p values less than 0.05 represent statistical significance. Red boxes indicate significant mean differences, while grey boxes indicate non-significant mean differences. Abbreviations: 95 % CI = 95 % confidence intervals, I2 = I2 heterogeneity, Chi2 = Q-statistic. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Additional sensitivity analyses were performed on studies with age-matched HC. The findings showed a significant increase in QSM values in all basal ganglia subregions in MS compared to age-matched HC (g ranged between 0.35 and 0.56, p < 0.01, I2 ranged between 45 % and 63 %), while a significant decrease in thalamic QSM values in MS was observed (g = −0.46, 95 % CI = −0.78–−0.13, p < 0.01, I2 = 86 %) (see Appendix 7 for forest plots). Further sensitivity analyses were conducted on studies that acquired images from 3 T MRI and below, findings showed a significant reduction in thalamic QSM values (g = −0.64, 95 % CI = −0.81–−0.47, p < 0.01, I2 = 51 %), and a significant increase in basal ganglia QSM values, with heterogeneity levels ranging between 53 % and 62 % (see Appendix 8 for forest plots). Lastly, sensitivity analyses were carried out on studies that did not specify the brain region to which QSM values were referenced. The findings showed a similar significant increase in basal ganglia QSM values in MS, while no significant difference in thalamic QSM values was detected (see Appendix 9 for forest plots).

Subgroup analysis for RRMS demonstrated a significant increase in QSM values in MS in caudate, putamen, and globus pallidus with a relatively lower heterogeneity (I2 ranged from 39 % to 61 %). Significant difference was again not observed in the thalamus (g = −0.33, 95 % CI = −0.8–0.13, p = 0.16), and the heterogeneity level remained high, with I2 = 83 % (see Appendix 5 for forest plots). Despite the lack of significant change in thalamic QSM values, both findings from main meta-analysis and the RRMS subgroup analysis showed a consistent direction towards a decrease in thalamic QSM values in RRMS and MS as a group.

3.2.3 Meta-analysis on the correlation between EDSS and DGM susceptibility

Of the 10 studies that examined the relationship between EDSS and DGM susceptibility, five were included in the meta-analyses with four studies focusing on the thalamus (Burgetova et al., 2017, Pontillo et al., 2019; Pudlac et al., 2020; Rudko et al., 2014), and four on basal ganglia subregions (Burgetova et al., 2017; Pudlac et al., 2020; Rudko et al., 2014, Schmalbrock et al., 2016). In addition, three of the included studies reported Pearson’s correlation coefficients (Pontillo et al., 2019, Rudko et al., 2014, Schmalbrock et al., 2016) and two reported Spearman’s correlation coefficients (Burgetova et al., 2017, Pudlac et al., 2020). Another five studies were excluded from the meta-analyses for the following reasons: 1) only unstandardized β values were reported (Zivadinov et al., 2013), 2) only correlation with basal ganglia QSM values as a whole was provided (Langkammer et al., 2013), 3) missing r values for non-significant correlation between EDSS and DGM susceptibility (Elkady et al., 2019, Hagemeier et al., 2018b, Hamdy et al., 2022). The latter three studies, however, were included in sensitivity analyses, see below.

Fig. 4 presents the pooled effects on the relationship between EDSS and DGM susceptibility in four subregions. There was a significant effect on the correlation between EDSS and putamen QSM values (r = 0.36, 95 % CI = 0.16–0.53, p = 0.01, I2 = 0 %). In contrast, no significant effect was observed on the correlation between EDSS the QSM values in the thalamus, caudate, and globus pallidus, with heterogeneity levels varying from low (I2 = 0 %) to high (I2 = 85 %). Visual examination of funnel plots (see Appendix 11) did not show noticeable asymmetry, none of the Egger’s test showed significant results (see Appendix 4B).Fig. 4 Forest plots of meta-analyses of correlation between EDSS and QSM values in different DGM subregions. The correlation between EDSS and putamen QSM values was found to be significant. However, no significant correlation was observed between EDSS and QSM values in the thalamus, caudate, and globus pallidus. p values less than 0.05 represent statistical significance. Red boxes indicate significant correlations, and grey boxes indicate non-significant correlations. Abbreviations: 95 % CI = 95 % confidence intervals, I2 = I2 heterogeneity, Chi2 = Q-statistic. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

We conducted sensitivity analyses by including studies that did not provide effect estimates (r values) but reported non-significant relationships between EDSS and DGM susceptibility. The missing effect estimates were imputed with a value of zero. Despite a decrease in pooled effect sizes, the findings were mostly insensitive to the inclusion of these imputed effect sizes. Correlation between EDSS and QSM values in thalamus (r = −0.01, 95 % CI = −0.28–0.26, p = 0.91, I2 = 71 %), caudate (r = 0.13, 95 % CI = −0.04–0.3, p = 0.11, I2 = 23 %), and globus pallidus (r = 0.1, 95 % CI = −0.25–0.42, p = 0.49, I2 = 71 %) remained non-significant. As for the relationship between putamen susceptibility and EDSS, the pooled effect size decreased, but the effect remained almost significant (r = 0.25, 95 % CI = 0.00–0.47, p = 0.05, I2 = 61 %) (see Appendix 12 for forest plots).

A second sensitivity test was specifically conducted for globus pallidus, as there was an outlier from the study by Rudko et al. (2014). The confidence interval from Rudko et al. (2014) did not overlap with the confidence interval of the pooled effect (see Appendix 12 for forest plot). After the removal of the outlier study, similar findings were observed. There was still a non-significant correlation between EDSS and globus pallidus susceptibility, with a pooled effect of r = −0.02, 95 % CI = −0.10–0.05, p = 0.39, I2 = 0 % (see Appendix 13 for forest plots).

3.2.4 Other clinical and structural correlates of DGM susceptibility

In addition to EDSS, associations of DGM susceptibility with disease duration (DD) and clinical assessments were reported in the literature. Of the six studies that examined this relationship, only three provided correlation coefficients, while the other three mentioned that no significant association was found (represented as ‘n.s.’ in Fig. 5). As summarized in Fig. 5, the majority of the seven studies reported non-significant correlations between DD and QSM values in all DGM subregions. While the relationship between DD and QSM values in caudate, putamen, and globus pallidus were positively inclined, negative correlations of DD and QSM values were reported in the thalamic region by Zivadinov et al. (2018).Fig. 5 Overview of findings of clinical and structural correlates of susceptibility values in DGM subregions. Displayed are effect sizes derived from correlation analyses of DGM subregion QSM values with clinical and structural measures. The different clinical and structural correlates are labeled on top of the figure, and the respective DGM subregions are denoted on the right. As for the direction of correlation effects, negative correlations indicate (A) shorter disease duration, (B) decreased performance, (C) smaller lesion volume, and (D) smaller regional volume in relation to increased DGM subregion QSM values and vice versa. For clarity, Schmalbrock et al. (2017) positive correlations between high Flanker’s test (decreased performance) and QSM values are flipped towards the negative direction but represented with unfilled bars. Abbreviations: * and bold letters = significant correlation, p < 0.05, n.s. = non-significant correlation, p = Pearson’s correlation coefficient, s = Spearman’s correlation coefficient, u = unstandardized regression coefficient, b = standardized regression coefficient (beta), DGM = deep grey matter.

The three studies that investigated the association between clinical assessments and DGM susceptibility covered different clinical tests ranging from neuropsychological tests (e.g., Symbol Digit Modalities Test, and Flanker’s test) used in the studies by Fujiwara et al. (2017) and Schmalbrock et al. (2016) to a mobility and leg function performance test (e.g., Timed 25-Foot Walk) in the study from Hamdy et al. (2022). Most findings showed a non-significant association between clinical assessments and DGM susceptibility. Only three significant associations were reported. First, Hamdy et al. (2022) showed a negative correlation between thalamic QSM values and mobility, quantified by Timed 25-Foot Walk. Second, Schmalbrock et al. (2016) found that Flanker’s test positively correlated with caudate QSM values, indicating worsening inhibition as caudate QSM values increase. Third, Fujiwara et al. (2017) showed a negative correlation between combined neuropsychological test performance (recall tasks and verbal fluency) and globus pallidus QSM values after controlling for demographic influences (i.e., age and sex) and globus pallidus volume.

Six studies examined structural correlates of DGM susceptibility. Among the six, two reported the correlation between QSM values and total lesion volume (TLV), quantified through the total T2w lesion volume (Burgetova et al., 2017, Chiang et al., 2018). Chiang et al. (2018) reported a positive correlation between TLV and thalamic QSM values, while Burgetova et al. (2017) reported a negative correlation. In the regions of caudate, putamen, and globus pallidus, no significant relationship between TLV and QSM values was found (Burgetova et al., 2017). Regarding brain volume loss, four studies examined its relationship with DGM susceptibility (Fujiwara et al., 2017, Pontillo et al., 2019, Schweser et al., 2018, Zivadinov et al., 2018). Findings on basal ganglia QSM values and volume loss were inconsistent; only one study reported significant negative correlation between basal ganglia volume and QSM values (Zivadinov et al., 2018), whereas two other studies reported no significant associations. As for thalamic QSM values and volume, three studies reported a significant positive correlation of thalamic volume and QSM values (Pontillo et al., 2019, Schweser et al., 2018, Zivadinov et al., 2018).

3.3 QSM lesions

3.3.1 Features and proportions of QSM lesions

Of the 22 studies reporting QSM lesion counts (N MS = 1118), 17 studies reported only QSM white matter lesion (WML) count, while two reported QSM lesions across cortical and WM regions; another two studies focused solely on cortical QSM lesions. Only one study reported the proportion of QSM WML (80.3 %), cortical lesions (14.8 %), and mixed WM/ cortical lesions (4.9 %) (Bian et al., 2016).

The number of hyperintense QSM lesions in reference to total lesion count on conventional scans (e.g., T2, FLAIR) was most frequently documented. Seventeen studies provided data on the percentages of hyperintense QSM lesions, showing varying proportions that ranged from 13 % to 88 % (Guo et al., 2021, Liao et al., 2023; Haacke et al., 2021; Castellaro et al., 2017, Chawla et al., 2016, Chawla et al., 2018, Chen et al., 2014, Cronin et al., 2016, Gillen et al., 2021, Harrison et al., 2016, Kakeda et al., 2015, Li et al., 2016, Pelizzari et al., 2020, Rahmanzadeh et al., 2022, Tolaymat et al., 2020, Zhang et al., 2016a, Zhang et al., 2016c). Additionally, rim lesion counts on QSM were recorded in 12 studies, reporting proportions that ranged from 4 % to 44 % (Chawla et al., 2018, Cronin et al., 2016, Guo et al., 2021, Gillen et al., 2021, Li et al., 2016, Harrison et al., 2016, Kaunzner et al., 2019, Huang et al., 2022, Rahmanzadeh et al., 2022, Yao et al., 2018, Zhang et al., 2022).

3.3.2 Clinical correlates of QSM in lesions and normal-appearing white/grey matter

Table 1 provides an overview of the findings on the clinical correlates of QSM lesions. Our review identified five clinical correlates from 17 studies (N MS = 902). EDSS is the most studied clinical variable. A significant positive association between rim lesions and EDSS was reported in 4 out of 7 studies, constituting 73 % of the total sample (N total = 438). The findings on EDSS and other QSM correlates, quantified by QSM lesion frequency, lesion QSM values, and NAWM QSM values, were inconsistent (Lesion frequency N = 159, k = 5; Lesion QSM values N = 93, k = 3, NAWM QSM values N = 203, k = 3).Table 1 Summary of findings of clinical correlates of QSM lesions. Correlations of QSM variables with clinical severity are indicated by ‘+’ and ‘-’ signifying positive and negative relations, respectively; ‘n.a.’ denotes not applicable as the QSM lesion correlates were categorical variables (i.e., MS phenotypes). Abbreviations: Assoc = association, NAWM = normal-appearing white matter, NAGM = normal-appearing grey matter.

Clinical and demographical variables in relation to QSM variables	Studies with significant results	Studies without significant results	Proportion of significant findings (%)	
Reference	Assoc	Reference	
Expanded Disease Status Scale	
Lesion frequency	Castellaro et al. (2017)	+	Guo et al. (2021)	2/5 (40 %)	
Tolaymat et al. (2020)		Haacke et al. (2021)	
		Pelizzari et al. (2020)	
Lesion QSM values	Coffman et al. (2022)	+	Guo et al. (2021)	1/3 (33 %)	
Harrison et al. (2016)	
Rim lesion frequency	Huang et al. (2022)	+	Cronin et al. (2016)	4/7 (57 %)	
Marcille et al. (2022)		Jang et al. (2020)	
Coffman et al. (2022)		Yao et al. (2018)	
Tolaymat et al. (2020)	
NAWM QSM values	Pietroboni et al. (2022)	+	Pontillo et al. (2023)	1/3 (33 %)	
Hamdy et al. (2022)	


	
Neurological Assessments	
Lesion frequency			Harrison et al. (2016)	0/1 (0 %)	
Lesion QSM values			Harrison et al. (2016)	0/1 (0 %)	
Rim lesion frequency	Marcille et al. (2022)	+	Tolaymat et al (2020)	1/2 (50 %)	
QSM lesion volume			Haacke et al. (2021)	0/1 (0 %)	


	
Disease Duration		
Lesion frequency	Castellaro et al. (2017)	–	Yao et al. (2018)	2/3 (60 %)	
Pelizzari et al. (2020)	
QSM lesion volume			Pelizzari et al. (2020)	0/1 (0 %)	
NAGM QSM values	Straub et al. (2023)	+		1/1 (100 %)	
NAWM QSM values			Chen et al. (2017)	0/1 (0 %)	


	
Annualized Relapse Rate	
Lesion QSM values			Guo et al. (2021)	0/1 (0 %)	
Rim lesion frequency	Guo et al. (2021)	+		1/1 (100 %)	


	
MS Phenotypes	
Lesion frequency	Harrison et al. (2016)	n.a.	Castellaro et al. (2017)	1/4 (25 %)	
Pelizzari et al. (2020)	
Tolaymat et al. (2020)	
Lesion QSM values	Harrison et al. (2016)	n.a.		1/1 (100 %)	
Rim lesion frequency	Harrison et al. (2016)	n.a.		1/1 (100 %)	
QSM lesion volume			Pelizzari et al. (2020)	0/2 (0 %)	
Tolaymat et al. (2020)	
NAWM QSM values	Pietroboni et al. (2022)	n.a.		1/1 (100 %)	

Different neuropsychological tests (e.g., Symbol Digit Modalities Test and California Verbal Learning Test-II) and overall functional assessments (e.g., Multiple Sclerosis Functional Composite) were also studied. However, the majority of the findings (k = 3 out of 4) reported no association between neurological assessment and QSM lesion frequency, volume, and QSM values (Haacke et al., 2021, Harrison et al., 2016, Tolaymat et al., 2020). Furthermore, the only longitudinal analysis by Tolaymat et al. (2020) found that MS patients with rim and without rim lesions only differed in one (Paced Auditory Serial Addition Test) out of five clinical tests after a one-year follow-up (Modified Fatigue Impact Scale, Symbol Digit Modalities Test, Timed 25-Foot Walk, 9-Hole Peg Test, and EDSS). Additionally, Tolaymat et al. (2020) also reported no clinical score difference between MS patients with and without rim lesions at baseline and after two and three years.

The few studies, which examined DD and annualized relapse rate (ARR) as clinical correlates of QSM lesions yielded inconsistent findings. As for MS phenotypes, Harrison et al. (2016) found the PMS subtype to have significantly higher numbers of QSM lesions and rim lesions than the RRMS subtype. In contrast, Pelizzari et al. (2020) and Tolaymat et al. (2020) reported that MS subtypes did not differ in total QSM lesion volume. Of note, neither of the latter two studies accounted for the presence of Gd-enhancing lesions.

3.4 Longitudinal MS studies using QSM

3.4.1 DGM susceptibility

We found only one study that investigated the correlation between QSM and EDSS over time. No significant correlation was observed in all DGM subregions (Hagemeier et al., 2018b). Three studies reported longitudinal changes in QSM values in DGM, and the findings varied across studies. Two reported an increase in caudate QSM values after two-year follow-up (Elkady et al., 2018, Hagemeier et al., 2018b), with Elkady et al. (2018) observing the rise in caudate QSM values in RRMS (N = 27), but not in PMS (N = 17). A similar trend was reported by Hagemeier et al., (2018b), showing a significant increase in caudate QSM values in RRMS (N = 98) and SPMS (N = 22). For the other DGM subregions, there was no significant change in QSM values in MS as compared to HC over the two-year follow-up period. A five-year longitudinal study observed a significant increase in QSM values only in the putamen (Elkady et al., 2019).

3.4.2 Lesion susceptibility

Among the eight studies reporting longitudinal findings on QSM lesion changes in MS, none investigated the correlation of QSM lesion change and clinical parameters over time. Regarding lesion size, Chawla et al. (2018) reported that most hyperintense QSM lesions remained stable over two years, while only 4 out of 191 lesions shrank and 10 enlarged. Coffman et al. (2022) studied rim lesions over two years and reported no significant change in rim lesion volume (p = 0.2) or the lesion volume of the inner and outer diameter of the rim (p > 0.3). Other than lesion size, the visibility of lesions remained constant over two years (Coffman et al., 2022, Jang et al., 2020) and decayed after four years (Zhang et al., 2019).

Six studies recorded changes in lesion QSM values over time (Fig. 6). At the time of Gd-enhancement, lesions exhibited lower QSM values, with rim Gd-enhancing lesions having higher QSM values than nodular Gd-enhancing lesions (Zhang et al., 2016a). The QSM values of lesions increased over time as the lesions turned into non-enhancing lesions, as shown in the three studies by Zhang et al., 2016a, Zhang et al., 2016b, and Chen et al. (2014). QSM values reached their maximum within the first half of the year upon detection, and the QSM values of lesions only started to fall drastically after about two years for non-rim lesions and a bit later for rim lesions (Zhang et al., 2019). This finding was also consistent with Chen et al. (2014), who reported lesion QSM values peaking after two years and declining after six years.Fig. 6 Longitudinal change in lesion QSM values. The left panel shows aggregate lesion QSM values reported in all studies over time, the lesion QSM values were normalized within-study by subtracting the mean QSM value and dividing by the standard deviation of the corresponding study. The right panel illustrates lesion QSM values change per individual study. Figures display lower QSM values for lesions at the time of Gd-enhancement. Subsequently, QSM values increase over time when lesions become non-enhancing, as shown in the three studies by Zhang et al., 2016a, Zhang et al., 2016b, and Chen et al. (2014). QSM values of lesions only started to fall drastically after about two years for non-rim lesions and a bit later for rim lesions. Error bars represent standard deviations. Abbreviations: GdE = Gadolinium-enhancing lesion, nonGdE = non-gadolinium-enhancing lesion, ppm = parts per million, rim = rim lesion, QSM lesion = all QSM hyperintense lesions, including both rim and non-rim lesions.

3.5 Other MRI modalities

The associations between QSM lesions in relation to other MRI sequences were examined across 12 studies (N MS = 616), which are summarized in Appendix 14. For the correlation with Gd-enhancing lesions, see section 3.4.2 on longitudinal studies. For the correlation with histology (also see section 3.6), the study of Rahmanzadeh et al. (2022) used MWF measures to investigate myelin integrity and found low MWF values in rim lesions and high MWF values in iso- and hypointense QSM lesions.

3.6 Correlates of QSM with histology

Table 2 summarizes all seven studies which combined both histological analysis and QSM images to investigate the paramagnetic and diamagnetic source of QSM values in MS brain specimens. Galbusera et al. (2022) investigated the correlation between lesion QSM values and myelin content; they reported a negative correlation between QSM values and myelin (rho = −0.58, p < 0.01), while QSM values positively correlated with astrocyte immunoreactivity (rho = 0.34, p < 0.001). On the contrary, Wiggermann et al. (2017) reported no significant correlation between QSM values and myelin (R2 = 0.001, p = 0.93).Table 2 Summary of Studies Using Histological Methods. Abbreviations: BCAS1 = breast carcinoma amplified sequence 1, BG = basal ganglia, CN = caudate nucleus, DAB = 3,3′- diaminobenzidine, DGM = deep grey matter, F = female, GFAP(IF) = glial fibrillary acidic protein (immunofluorescence), GP = global pallidus, Iba1 = ionized calcium-binding adapter molecule1, iNOS = inducible nitric oxide synthase, LFB-PAS = Luxol fast blue and periodic acid-schiff myelin stain, M = male, MHC -II = major histocompatibility complex class II, MPB = myelin basic protein, NF(IF) = neurofilament cocktail (immunofluorescence), PPMS = primary progressive MS, PUT = putamen, RRMS = relapsing-remitting MS, SPMS = secondary progressive MS, TBB = Turnbull bell iron staining, TSPO = mitochondrial translocator protein, WML = white matter lesion.

No	Reference	N (MS Type)	Mean Age at Death	Sex	Number of Lesion (Brain Region)	Histological Stains	Analysis	Findings	
1	Gillen et al., 2021	16(RRMS: 3; SPMS: 11; PPMS: 1)	57	F: 8M: 8	20(WML)	DBA-Perl’s, MBP, CD68, iNOS	Correlation	QSM signal intensity of rim lesions correlated with the density of Perl’s and CD68 + cells (R2 = 0.77), whereas non-rim lesions showed no Perl’s staining and minimally activated microglia. 	
2	Kaunzner et al., 2019	7(RRMS: 4; SPMS: 1; PPMS: 1; unknown: 1)	55	F: 3M: 4	7(WML)	DBA-Perl’s,CD68, TSPO	Descriptive	Iron was present in CD68 + myeloid cells throughout the hyperintense rims. In non-rim lesions, iron-positive cells were not present at the border and only a few CD68 + microglia cells were seen.	
3	Wiggermann et al., 2017	5(SPMS: 3; PPMS: 2)	50	F: 3M: 2	15(WML)	TBB, LFB-PAS, Iba1	Correlation	No direct correlation between QSM values and iron content quantified via TBB staining (R2 = 0.002, p > 0.35), and no significant correlation between QSM values and myelin (R2 = 0.001, p = 0.93).	
4	Wisnieff et al., 2015	1 (unknown)	68	M: 1	5(WML)	Perl’s, MBP, CD68	Descriptive	Higher QSM values correspond to substantial iron content and vice versa. CD68 labelling showed evidence of microglia associated with iron at the border of QSM lesion; in contrast, lesions without trace of iron showed minimal sign of CD 68	
5	Rahmanzadeh et al., 2022	3
(SPMS: 2;
PPMS: 1)	63	M: 3	63(WML)	MPB, TBB, MHC II, and BCAS1	Descriptive	Hyperintense rim lesions on QSM were chronic active lesions with iron-laden macrophages at the lesion edge and hyperintense QSM lesions as chronic inactive lesions with low level of staining of iron and macrophages and showing extensive demyelination. 71.4 % of hyperintense QSM lesions appeared to be chronic inactive, 92.8 % of the rim lesions were chronic active, and 88.9 % of hypo-/iso-intense lesions were remyelinated lesions.	
6	Galbusera et al., 2022	3(RRMS: 2, SPMS: 1)	58	F: 1M: 2	65(WML)	LFB - PAS, MBP, BCAS1, HLA-DP, clone CR3/43, NF(IF), GFAP cocktail (IF)	Correlation, Logistic regression, Linear mixed-effect models	Depending on the extent of demyelination/remyelination, lesions were categorized as active (9), chronic active (35), inactive (12), and remyelinated (9) through histological staining and image analyses. QSM values correlated with myelin content (MBP: rho = −0.58, p < 0.01; LFB: -0.41, p < 0.01) as well as astrocyte immunoreactivity (rho = 0.34, p < 0.01). In terms of lesion type identification, QSM values were effective in discriminating active lesions from other lesion types (OR = 1.03, = 0.02, AIC = 50.47), but less ideal in predicting remyelinated lesion from other lesion types (OR = 0.97, p = 0.045, AIC = 51.280), and lastly QSM values of inactive and chronic active lesions did not differ (OR = 1, p = 0.75, AIC = 57.3).	
7	Sun et al., 2015	3;(SPMS: 2; RRMS: 1)	56	M: 3	(DGM, BG, and Thalamus)	Perl’s	Correlation	Significant linear correlation of QSM values to Perl’s iron stain was found for all three subjects (R2 = 0.75, 0.62, 0.86, p < 0.05).	

Regarding iron content, only one study (NSample = 3) focused on the DGM region, showing a strong association between iron content and QSM values, with R2 = 0.62 (Sun et al., 2015). Studies focusing on QSM rim lesions demonstrated that QSM signals within the rims were positively correlated with iron content, as well as the expression of major histocompatibility complex class II cells and macrophage-associated antigen CD 68 (Gillen et al., 2021, Kaunzner et al., 2019, Rahmanzadeh et al., 2022, Wisnieff et al., 2015). In contrast, non-rim lesions showed only minimal presence of CD68+ and iron positive cells. Wiggermann et al. (2017) found no correlation between QSM values and iron-positive cells (R2 = 0.002, p > 0.35) in 15 chronic (assumed non-active) lesions. Similarly, findings by Rahmanzadeh et al. (2022) showed 71.4 % of hyperintense QSM lesions were chronic inactive lesions with only minimal staining of iron and/or macrophages, while iso- and hypointense QSM lesions were mostly (88.9 %) remyelinated. These histopathological findings of QSM not only suggest that an increased iron content drives QSM hyperintensity of active rim lesions; these findings are also compatible with the idea that iso- and hypointense QSM lesions indicate remyelination.

4 Discussion

4.1 Summary of evidence and general considerations

This systematic review and meta-analysis have summarized a broad range of literature on the use of QSM in MS and yielded four main findings: 1. After outlier removal, QSM values increase in basal ganglia (caudate, putamen, and globus pallidus), while thalamic QSM values decrease. In addition, QSM values of the putamen positively correlate with EDSS. 2. QSM rim lesions are associated with more severe disability. 3. Lesion development from initial detection to the inactive stage is paralleled by increasing, plateauing (after about two years), and decreasing QSM values, respectively. 4. Histological findings support the association between QSM values and iron in the edges of rim lesions; in concert with data from myelin water imaging, the decrease of diamagnetic myelin is likely to drive the increase of QSM values within WM lesions. However, despite a significant number of studies on the topic, the meta-analytic yield of our pre-registered study was unexpectedly low.

4.2 Sources of bias and heterogeneity

For the limited yield of our systematic review, we see three potentially important reasons: 1) too little available data, 2) the heterogeneous study designs and reporting styles, and 3) the lack of methodological standards of QSM itself. We observed highly variable outcome measures and approaches to analyze them. Some, but by far not all, resulted from specific objectives plausibly. On the other hand, it has not become clear why measurement choices, reporting styles, and statistical tests had to be as heterogeneous as observed in this review, prompting the need for standardized approaches.

Finally, different processing pipelines, including different reference regions, have been applied to generate QSM images, indicating no commonly accepted methodological standard (Appendix 1). Given the impact of QSM reconstruction algorithms and parameters on image clarity (Langkammer et al., 2018), the need to detail QSM reconstruction processes and to adopt a standardized QSM pipeline is evident. Notably, the QSM Consensus Organization Committee has provided guidelines on acquiring sequences, reconstructing QSM images, analyzing findings, and presenting results (Bilgic et al., 2023), which we believe are crucial for improving the reliability and comparability of QSM findings. In conclusion, the heterogeneous nature of the data analyzed must be considered when further discussing them.

4.3 DGM susceptibility

Histopathological evidence shows strong positive correlation between QSM values and iron content in DGM of HCs (Langkammer et al., 2012, Stüber et al., 2014, Sun et al., 2015) so that iron can be regarded as the primary driver of DGM susceptibility. Further, robust evidence shows that DGM susceptibility increases with normal aging (Haider et al., 2014, Taege et al., 2019, Wang and Liu, 2015) likely reflecting higher concentrations of iron. In MS patients, however, our analyses on DGM revealed different behaviors of QSM values in the thalamus and basal ganglia; thus, these two regions are discussed separately.

The higher basal ganglia QSM values found in MS indicate higher iron concentration in this brain region. However, the precise physiological cause of elevated iron concentration remains elusive. Schweser et al. (2021) suggested that the increase in QSM values is partially driven by neuronal loss, which subsequently leads to an increase in regional concentration instead of an actual increase in the overall content of iron. Along this line, studies on the relationship between basal ganglia QSM values and their respective volumes demonstrated a trend in the association between high QSM values and lower regional volume (Fig. 5D) (Fujiwara et al., 2017, Pontillo et al., 2019, Zivadinov et al., 2018). As our meta-analysis could not account for regional volume as a confound, we could not clarify if the difference in basal ganglia QSM values between MS and HC is independent of volume loss. In terms of correlations of QSM with clinical scores, we observed a correlation between QSM values and EDSS only for the putamen, although it was only close to significant (p = 0.05) after taking into account studies with missing data. The putamen is relatively large and clearly demarcated, facilitating precise measurements. Another explanation for the non-significant association across parts of DGM is that iron accumulation, the driving factor behind the elevated QSM values in MS, may occur early in the course of MS (Al-Radaideh et al., 2013, Khalil et al., 2015) without necessarily leading to a substantial increase in EDSS scores. Alternatively, clinical scores, such as fatigue, better reflecting basal ganglia function, may be necessary to demonstrate such a relation. As only univariate correlational effect measures were available for meta-analysis on DGM susceptibility and EDSS, our analysis did not account for the effect of age, which is not only a covariate of DGM susceptibility (Li et al., 2021, Liu et al., 2016), but also of EDSS (Manouchehrinia et al., 2017).

Compared to the results on basal ganglia, we observed lower thalamic QSM values in MS with high heterogeneity levels, after controlling for outlier and age. The reduction in thalamic QSM values likely indicates a decrease in iron content or a disproportionate decay of iron-containing structures (Schweser et al., 2021). In contrast to the basal ganglia, the thalamus consists not only of nuclei containing dense grey matter but it also contains a considerable amount of WM separating and connecting these nuclei (Herrero et al., 2002). This distinctive anatomical structure containing grey and white matter may lead to opposing effects regarding QSM values. Possibly different QSM techniques may cover these effects differently and, hence, contribute to the heterogenous results across studies particularly in the thalamus.

4.4 Lesion susceptibility

According to histological evidence, both iron and myelin content drive the change in QSM lesion susceptibility (Deh et al., 2018, Hametner et al., 2018, Langkammer et al., 2012, Stüber et al., 2014). Further evidence from myelin imaging has also demonstrated the association between myelin density and changes in QSM values in rim and non-rim lesions (Huang et al., 2022, Yao et al., 2018). Thus, depending on the lesion type, the iron to myelin density ratio varies, and so does QSM lesion susceptibility (Gillen et al., 2021, Wisnieff et al., 2015, Zhang et al., 2019). These findings, however, do not explain the nature of hypointense QSM lesions, the least frequent QSM lesion type. Again, we believe that the disproportional decay of iron-containing structures best explains this phenomenon. Of note, we found evidence based on histology and myelin water imaging, that hypo- and isointense QSM lesions may indicate remyelination (Rahmanzadeh et al., 2022). Nevertheless, the results of the included MRI studies were again very heterogeneous. For instance, hyperintense QSM lesions were the most frequent lesion type reported, yet its frequency ranged widely across studies, from only 13 % to nearly 88 %.

Regarding longitudinal change of lesion QSM values, our graphical syntheses of six studies from the same institution yielded a coherent picture well compatible with the suggestions of Kuhlmann et al.(2017). Shortly after lesion formation corresponding to the active lesion stage, there is a rapid and sharp increase in lesion QSM values, in which inflammation and iron accumulation occur (Hametner et al., 2013). Then, lesion QSM values maintain high for a couple of years. The maintenance of lesion QSM values corresponds to mixed active and inactive lesions, wherein demyelination and inflammation occur. Lastly, lesion QSM values decrease gradually after year four to six, which corresponds to the depletion of all inflammation and demyelinating activities in the inactive lesion stage.

Regarding rim lesions, histological studies demonstrated an association of QSM values in the rim (of rim lesions) with the iron content contained in CD68-positive cells such as macrophages (Gillen et al., 2021, Kaunzner et al., 2019, Rahmanzadeh et al., 2022). There is also an association between the presence of rim lesions with more severe disability in the studies representing the majority of all patients (73 %). However, rim lesion frequency varied across studies, ranging between 4 % and 44 %. This wide variability has also been reported in other reviews on rim lesions in MS that considered not only rim lesions identified by QSM but also by SWI or phase imaging (Altokhis et al., 2020, Kwong et al., 2021). Notably, transient rim lesions have been described (Weber et al., 2022), and only a few studies covered by this review have accounted for rim Gd-enhancing lesions. This is important as the transient occurrence of QSM rims close to the stage of lesion development (and hence contrast-enhancement) may have contributed to the wide variability of the frequency of rim lesions across studies. In a cross-sectional study on MRI scans of 159 patients with RRMS acquired in the context of routine yearly follow-up examinations, only 171 (3.8 %) rim lesions and only 20 (0.4 %) Gd-enhancing lesions were observed (Marcille et al., 2022). In contrast, a longitudinal study over a mean follow-up of 16 months by Jang et al. (2020) reported a considerable overlap of rim lesions and contrast enhancement among 22 patients with MS. Nine out of 19 rim lesions (47.3 %) showed contrast enhancement at baseline; while the enhancement disappeared on follow-up, rim lesions remained visible on T2w images. Given the unclear proportion of rim Gd-enhancing lesions and their degree of persistency over time, the overlap of QSM rim lesions and contrast enhancement deserves further investigation since accounting for this source of variability is likely to improve the informative value of rim lesions.

4.5 Limitations

There are several limitations that have potentially influenced our results. First, the overarching theme of this review, QSM in MS research, is broad, and our a-priori decisions on subdividing this theme into different domains (section 2.2) may not have been the best choice. Second, although we decided to include and exclude studies based on commonly accepted standards, these had to be selected and modified for applicability reasons, resulting in arbitrary choices (e.g., risk of bias assessment criteria on sample size ≥50, adequate interval between MRI and EDSS of 12 months, and reliable outcome measure, i.e., no selective reporting detected). Third, only 17 out of 55 studies were included in the meta-analyses, while the remaining studies, which mainly focused on QSM lesions, were synthesized through graphs and tables. Fourth, in the meta-analysis on the correlational effect of DGM susceptibility and EDSS, Egger’s test was applied despite the small study sizes (k < 7); therefore, sensitivity to detect potential publication bias may have been minor (Higgins et al., 2021). Fifth, the meta-analytical findings derived from correlational effect estimates (Pearson’s r and Spearman’s rho) could not be age-corrected.

5 Conclusion

Our systematic review and meta-analysis on QSM in MS research revealed a significant increase in basal ganglia QSM values in MS compared to HC, and, in the putamen, an association of QSM values with disability. In contrast, thalamic QSM values were decreased.

However, the observed high heterogeneity across studies employing QSM in MS research emphasizes the need for standardization in QSM. This includes scanning procedures, processing pipelines, and reference regions to enhance the comparability and reliability of findings. Furthermore, the limited number of studies exploring the clinical correlates of QSM and its longitudinal aspects prompts further investigations to substantiate the role of QSM in MS for research or even clinical routine.

Ethics approval

Not applicable.

Funding

This work was supported by the German Federal Ministry of Education and Research (MS Lesion) grant: Radiomics: Next Generation of Biomedical Imaging; project number 428223038.

Disclosure of generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) used Grammarly, v6.8.263 and ChatGPT, v 3.5, in order to improve language and readability. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

CRediT authorship contribution statement

Cui Ci Voon: Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft. Tun Wiltgen: Conceptualization, Investigation, Writing – review & editing. Benedikt Wiestler: Writing – review & editing. Sarah Schlaeger: Writing – review & editing. Mark Mühlau: Writing – original draft, Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A Supplementary data

The following are the Supplementary data to this article:Supplementary data 1

Supplementary data 2

Data availability

Data will be made available on request.

Appendix A Supplementary data to this article can be found online at https://doi.org/10.1016/j.nicl.2024.103598.
==== Refs
References

Alexander A.L. Hurley S.A. Samsonov A.A. Adluru N. Hosseinbor A.P. Mossahebi P. Tromp do P.M. Zakszewski E. Field A.S. Characterization of cerebral white matter properties using quantitative magnetic resonance imaging stains Brain Connect 2011 10.1089/brain.2011.0071
Al-Radaideh A.M. Wharton S.J. Lim S.Y. Tench C.R. Morgan P.S. Bowtell R.W. Constantinescu C.S. Gowland P.A. Increased iron accumulation occurs in the earliest stages of demyelinating disease: An ultra-high field susceptibility mapping study in Clinically Isolated Syndrome Mult. Scler. J. 19 2013 896 903 10.1177/1352458512465135
Altokhis A.I. Alotaibi A.M. Felmban G.A. Constantinescu C.S. Evangelou N. Iron rims as an imaging biomarker in ms: A systematic mapping review Diagnostics 10 2020 1 16 10.3390/diagnostics10110968
Balduzzi S. Rücker G. Schwarzer G. How to perform a meta-analysis with R: A practical tutorial Evid. Based. Ment. Health 22 2019 10.1136/ebmental-2019-300117
Barkhof F. MRI in multiple sclerosis: correlation with expanded disability status scale (EDSS) Mult. Scler. J. 5 1999 283 286 10.1177/135245859900500415
Bergsland N. Schweser F. Dwyer M.G. Weinstock-Guttman B. Benedict R.H.B. Zivadinov R. Thalamic white matter in multiple sclerosis: A combined diffusion-tensor imaging and quantitative susceptibility mapping study Hum. Brain Mapp. 39 2018 4007 4017 10.1002/hbm.24227 29923266
Bian W. Tranvinh E. Tourdias T. Han M. Liu T. Wang Y. Rutt B. Zeineh M.M. In Vivo 7T MR quantitative susceptibility mapping reveals opposite susceptibility contrast between cortical and white matter lesions in multiple sclerosis Am. J. Neuroradiol. 37 2016 1808 1815 10.3174/ajnr.A4830 27282860
Bilgic, B., Costagli, M., Chan, K.-S., Duyn, J., Langkammer, C., Lee, J., Li, X., Liu, C., Marques, J.P., Milovic, C., Robinson, S.D., Schweser, F., Shmueli, K., Spincemaille, P., Straub, S., van Zijl, P., Wang, Y., Group, I.E.-M.T.P.S., 2023. Recommended Implementation of Quantitative Susceptibility Mapping for Clinical Research in The Brain: A Consensus of the ISMRM Electro-Magnetic Tissue Properties Study Group.
Burgetova A. Dusek P. Vaneckova M. Horakova D. Langkammer C. Krasensky J. Sobisek L. Matras P. Masek M. Seidl Z. Thalamic iron differentiates primary-progressive and relapsing-remitting multiple sclerosis Am. J. Neuroradiol. 38 2017 1079 1086 10.3174/ajnr.A5166 28450431
Castellaro M. Magliozzi R. Palombit A. Pitteri M. Silvestri E. Camera V. Montemezzi S. Pizzini F.B. Bertoldo A. Reynolds R. Monaco S. Calabrese M. Heterogeneity of cortical lesion susceptibility mapping in multiple sclerosis Am. J. Neuroradiol. 38 2017 1087 1095 10.3174/ajnr.A5150 28408633
Charter R.A. Alexander R.A. A note on combining correlations Bull. Psychon. Soc. 31 1993 10.3758/BF03334158
Chawla S. Kister I. Wuerfel J. Brisset J.C. Liu S. Sinnecker T. Dusek P. Haacke E.M. Paul F. Ge Y. Iron and non-iron-related characteristics of multiple sclerosis and neuromyelitis optica lesions at 7T MRI Am. J. Neuroradiol. 2016 1223 1230 10.3174/ajnr.A4729 27012298
Chawla S. Kister I. Sinnecker T. Wuerfel J. Brisset J.C. Paul F. Ge Y. Longitudinal study of multiple sclerosis lesions using ultra-high field (7T) multiparametric MR imaging PLoS One 13 2018 1 16 10.1371/journal.pone.0202918
Chen W. Gauthier S.A. Gupta A. Comunale J. Liu T. Wang S. Pei M. Pitt D. Wang Y. Quantitative susceptibility mapping of multiple sclerosis lesions at various ages Radiology 271 2014 183 192 10.1148/radiol.13130353 24475808
Chen W. Zhang Y. Mu K. Pan C. Gauthier S.A. Zhu W. Wang Y. Quantifying the susceptibility variation of normal-appearing white matter in multiple sclerosis by quantitative susceptibility mapping Am. J. Roentgenol. 209 2017 889 894 10.2214/AJR.16.16851 28705068
Chiang G.C. Hu J. Morris E. Wang Y. Gauthier S.A. Quantitative susceptibility mapping of the thalamus: relationships with thalamic volume, total gray matter volume, and T2 Lesion Burden Am. J. Neuroradiol. 39 2018 467 472 10.3174/ajnr.A5537 29371258
Cho J. Nguyen T.D. Huang W. Sweeney E.M. Luo X. Kovanlikaya I. Zhang S. Gillen K.M. Spincemaille P. Gupta A. Gauthier S.A. Wang Y. Brain oxygen extraction fraction mapping in patients with multiple sclerosis J. Cereb. Blood Flow Metab. 42 2022 338 348 10.1177/0271678X211048031 34558996
Cobzas D. Sun H. Walsh A.J. Lebel R.M. Blevins G. Wilman A.H. Subcortical gray matter segmentation and voxel-based analysis using transverse relaxation and quantitative susceptibility mapping with application to multiple sclerosis J. Magn. Reson. Imaging 42 2015 1601 1610 10.1002/jmri.24951 25980643
Coffman C.H. White R. Subramanian K. Buch S. Bernitsas E. Haacke E.M. Quantitative susceptibility mapping of both ring and non-ring white matter lesions in relapsing remitting multiple sclerosis Magn. Reson. Imaging 91 2022 45 51 10.1016/j.mri.2022.05.009 35605736
Cooper H. Hedges L.V. Valente J.C. The Handbook of research synthesis and meta-analysis 2019 Sage
Cronin M.J. Wharton S. Al-Radaideh A. Constantinescu C. Evangelou N. Bowtell R. Gowland P.A. A comparison of phase imaging and quantitative susceptibility mapping in the imaging of multiple sclerosis lesions at ultrahigh field. Magn. Reson. Mater. Physics Biol. Med. 29 2016 543 557 10.1007/s10334-016-0560-5
Deh K. Ponath G.D. Molvi Z. Parel G.C.T. Gillen K.M. Zhang S. Nguyen T.D. Spincemaille P. Ma Y. Gupta A. Gauthier S.A. Pitt D. Wang Y. Magnetic susceptibility increases as diamagnetic molecules breakdown: Myelin digestion during multiple sclerosis lesion formation contributes to increase on QSM J. Magn. Reson. Imaging 48 2018 1281 1287 10.1002/jmri.25997 29517817
Deistung A. Schweser F. Reichenbach J.R. Overview of quantitative susceptibility mapping NMR Biomed. 30 2017 10.1002/nbm.3569
Downs S.H. Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions J. Epidemiol. Community Heal. 52 1998 377 384 10.1136/jech.52.6.377
Elkady A.M. Cobzas D. Sun H. Blevins G. Wilman A.H. Progressive iron accumulation across multiple sclerosis phenotypes revealed by sparse classification of deep gray matter J. Magn. Reson. Imaging 46 2017 1464 1473 10.1002/jmri.25682 28301067
Elkady A.M. Cobzas D. Sun H. Blevins G. Wilman A.H. Discriminative analysis of regional evolution of iron and myelin/calcium in deep gray matter of multiple sclerosis and healthy subjects J. Magn. Reson. Imaging 48 2018 652 668 10.1002/jmri.26004
Elkady A.M. Cobzas D. Sun H. Seres P. Blevins G. Wilman A.H. Five year iron changes in relapsing-remitting multiple sclerosis deep gray matter compared to healthy controls Mult. Scler. Relat. Disord. 33 2019 107 115 10.1016/j.msard.2019.05.028 31181540
Filippi M. Brück W. Chard D. Fazekas F. Geurts J.J.G. Enzinger C. Hametner S. Kuhlmann T. Preziosa P. Rovira À. Schmierer K. Stadelmann C. Rocca M.A. Association between pathological and MRI findings in multiple sclerosis Lancet Neurol. 18 2019 198 210 10.1016/S1474-4422(18)30451-4 30663609
Fujiwara E. Kmech J.A. Cobzas D. Sun H. Seres P. Blevins G. Wilman A.H. Cognitive implications of deep gray matter iron in multiple sclerosis Am. J. Neuroradiol. 38 2017 942 948 10.3174/ajnr.A5109 28232497
Galbusera R. Bahn E. Weigel M. Schaedelin S. Franz J. Lu P.J. Barakovic M. Melie-Garcia L. Dechent P. Lutti A. Sati P. Reich D.S. Nair G. Brück W. Kappos L. Stadelmann C. Granziera C. Postmortem quantitative MRI disentangles histological lesion types in multiple sclerosis Brain Pathol 2022 10.1111/bpa.13136
Gillen K.M. Mubarak M. Park C. Ponath G. Zhang S. Dimov A. Levine-Ritterman M. Toro S. Huang W. Amici S. Kaunzner U.W. Gauthier S.A. Guerau-de-Arellano M. Wang Y. Nguyen T.D. Pitt D. QSM is an imaging biomarker for chronic glial activation in multiple sclerosis lesions Ann. Clin. Transl. Neurol. 8 2021 877 886 10.1002/acn3.51338 33704933
Gilpin A.R. Table for conversion of kendall’s tau to spearman’s rho within the context of measures of magnitude of effect for meta-analysis Educ. Psychol. Meas. 53 1993 87 92 10.1177/0013164493053001007
Granziera C. Wuerfel J. Barkhof F. Calabrese M. De Stefano N. Enzinger C. Evangelou N. Filippi M. Geurts J.J.G. Reich D.S. Rocca M.A. Ropele S. Rovira À. Sati P. Toosy A.T. Vrenken H. Gandini Wheeler-Kingshott C.A.M. Kappos L. Quantitative magnetic resonance imaging towards clinical application in multiple sclerosis Brain 144 2021 1296 1311 10.1093/brain/awab029 33970206
Guo Z. Long L. Qiu W. Lu T. Zhang L. Shu Y. Zhang K. Fang L. Chen S. The distributional characteristics of multiple sclerosis lesions on quantitative susceptibility mapping and their correlation with clinical severity Front. Neurol. 12 2021 1 8 10.3389/fneur.2021.647519
Haacke E.M. Bernitsas E. Subramanian K. Utriainen D. Palutla V.K. Yerramsetty K. Kumar P. Sethi S.K. Chen Y. Latif Z. Jella P. Gharabaghi S. Wang Y. Zhang X. Comley R.A. Beaver J. Luo Y. A comparison of magnetic resonance imaging methods to assess multiple sclerosis lesions: implications for patient characterization and clinical trial design Diagnostics 12 2021 77 10.3390/diagnostics12010077 35054244
Hagemeier J. Ramanathan M. Schweser F. Dwyer M.G. Lin F. Bergsland N. Weinstock-Guttman B. Zivadinov R. Iron-related gene variants and brain iron in multiple sclerosis and healthy individuals NeuroImage Clin. 17 2018 530 540 10.1016/j.nicl.2017.11.003 29201641
Hagemeier J. Zivadinov R. Dwyer M.G. Polak P. Bergsland N. Weinstock-Guttman B. Zalis J. Deistung A. Reichenbach J.R. Schweser F. Changes of deep gray matter magnetic susceptibility over 2 years in multiple sclerosis and healthy control brain NeuroImage Clin. 18 2018 1007 1016 10.1016/j.nicl.2017.04.008 29868452
Haider L. Simeonidou C. Steinberger G. Hametner S. Grigoriadis N. Deretzi G. Kovacs G.G. Kutzelnigg A. Lassmann H. Frischer J.M. Multiple sclerosis deep grey matter: The relation between demyelination, neurodegeneration, inflammation and iron J. Neurol. Neurosurg. Psychiatry 85 2014 1386 1395 10.1136/jnnp-2014-307712 24899728
Hamdy E. Galeel A.A. Ramadan I. Gaber D. Mustafa H. Mekky J. Iron deposition in multiple sclerosis: overall load or distribution alteration? Eur. Radiol. Exp. 6 2022 10.1186/s41747-022-00279-9
Hametner S. Wimmer I. Haider L. Pfeifenbring S. Brück W. Lassmann H. Iron and neurodegeneration in the multiple sclerosis brain Ann. Neurol. 74 2013 848 861 10.1002/ana.23974 23868451
Hametner S. Endmayr V. Deistung A. Palmrich P. Prihoda M. Haimburger E. Menard C. Feng X. Haider T. Leisser M. Köck U. Kaider A. Höftberger R. Robinson S. Reichenbach J.R. Lassmann H. Traxler H. Trattnig S. Grabner G. The influence of brain iron and myelin on magnetic susceptibility and effective transverse relaxation - A biochemical and histological validation study Neuroimage 179 2018 117 133 10.1016/j.neuroimage.2018.06.007 29890327
Harrer M. Cuijpers P. Furukawa T.A. Ebert D.D. Doing meta-analysis with R - A Hands-On Guide J. Stat. Softw. 102 2022 10.18637/jss.v102.b02
Harrison D.M. Li X. Liu H. Jones C.K. Caffo B. Calabresi P.A. Van Zijl P. Lesion heterogeneity on high-field susceptibility MRI Is associated with multiple sclerosis severity Am. J. Neuroradiol. 37 2016 1447 1453 10.3174/ajnr.A4726 26939635
Herrero M.T. Barcia C. Navarro J.M. Functional anatomy of thalamus and basal ganglia Child’s Nerv. Syst. 18 2002 386 404 10.1007/s00381-002-0604-1 12192499
Higgins J.P.T. Thomas J. Chandler J. Cumpston M. Li T. Page M.J. Welch V.A. Cochrane handbook for systematic reviews of interventions version 6.2 [updated February 2021] Cochrane Handbook for Systematic Reviews of Interventions 2021
Higgins J.P.T. Thompson S.G. Quantifying heterogeneity in a meta-analysis Stat. Med. 21 2002 10.1002/sim.1186
Huang W. Sweeney E.M. Kaunzner U.W. Wang Y. Gauthier S.A. Nguyen T.D. Quantitative susceptibility mapping versus phase imaging to identify multiple sclerosis iron rim lesions with demyelination J. Neuroimaging 1–9 2022 10.1111/jon.12987
Huedo-Medina T.B. Sánchez-Meca J. Marín-Martínez F. Botella J. Assessing heterogeneity in meta-analysis: Q statistic or I 2 Index? Psychol. Methods 11 2006 10.1037/1082-989X.11.2.193
Jang J. Nam Y. Choi Y. Shin N.Y. An J.Y. Ahn K.J. Kim B.S. Lee K.S. Kim W. Paramagnetic rims in multiple sclerosis and neuromyelitis optica spectrum disorder: A quantitative susceptibility mapping study with 3-t mri J. Clin. Neurol. 16 2020 562 572 10.3988/jcn.2020.16.4.562 33029961
Kakeda S. Futatsuya K. Ide S. Watanabe K. Miyata M. Moriya J. Ogasawara A. Sato T. Narimatsu H. Okada K. Uozumi T. Liu T. Wang Y. Korogi Y. Improved detection of cortical gray matter involvement in multiple sclerosis with quantitative susceptibility mapping Acad. Radiol. 22 2015 1427 1432 10.1016/j.acra.2015.08.003 26342769
Kaunzner U.W. Kang Y. Zhang S. Morris E. Yao Y. Pandya S. Hurtado Rua S.M. Park C. Gillen K.M. Nguyen T.D. Wang Y. Pitt D. Gauthier S.A. Quantitative susceptibility mapping identifies inflammation in a subset of chronic multiple sclerosis lesions Brain 142 2019 133 145 10.1093/brain/awy296 30561514
Khalil M. Langkammer C. Pichler A. Pinter D. Gattringer T. Bachmaier G. Ropele S. Fuchs S. Enzinger C. Fazekas F. Dynamics of brain iron levels in multiple sclerosis A longitudinal 3T MRI study Neurology 84 2015 2396 2402 10.1212/WNL.0000000000001679 25979698
Kuhlmann T. Ludwin S. Prat A. Antel J. Brück W. Lassmann H. An updated histological classification system for multiple sclerosis lesions Acta Neuropathol. 133 2017 13 24 10.1007/s00401-016-1653-y 27988845
Kurtzke J.F. Rating neurologic impairment in multiple sclerosis: An expanded disability status scale (EDSS) 1444–1444 Neurology 33 1983 10.1212/WNL.33.11.1444
Kwong K.C.N.K. Mollison D. Meijboom R. York E.N. Kampaite A. Thrippleton M.J. Chandran S. Waldman A.D. The prevalence of paramagnetic rim lesions in multiple sclerosis: A systematic review and meta-analysis PLoS One 16 2021 1 14 10.1371/journal.pone.0256845
Lancione M. Tosetti M. Donatelli G. Cosottini M. Costagli M. The impact of white matter fiber orientation in single-acquisition quantitative susceptibility mapping NMR Biomed. 30 2017 10.1002/nbm.3798
Langkammer C. Schweser F. Krebs N. Deistung A. Goessler W. Scheurer E. Sommer K. Reishofer G. Yen K. Fazekas F. Ropele S. Reichenbach J.R. Quantitative susceptibility mapping (QSM) as a means to measure brain iron? A post mortem validation study Neuroimage 62 2012 1593 1599 10.1016/j.neuroimage.2012.05.049 22634862
Langkammer C. Liu T. Khalil M. Enzinger C. Jehna M. Fuchs S. Fazekas F. Wang Y. Ropele S. Quantitative susceptibility mapping in multiple sclerosis Radiology 267 2013 551 559 10.1148/radiol.12120707 23315661
Langkammer C. Schweser F. Shmueli K. Kames C. Li X. Guo L. Milovic C. Kim J. Wei H. Bredies K. Buch S. Guo Y. Liu Z. Meineke J. Rauscher A. Marques J.P. Bilgic B. Quantitative susceptibility mapping: Report from the 2016 reconstruction challenge Magn. Reson. Med. 79 2018 1661 1673 10.1002/mrm.26830 28762243
Lassmann H. Brück W. Lucchinetti C.F. The immunopathology of multiple sclerosis: An overview Brain Pathol. 17 2007 210 218 10.1111/j.1750-3639.2007.00064.x 17388952
Li J. Chang S. Liu T. Wang Q. Cui D. Chen X. Jin M. Wang B. Pei M. Wisnieff C. Spincemaille P. Zhang M. Wang Y. Reducing the object orientation dependence of susceptibility effects in gradient echo MRI through quantitative susceptibility mapping Magnetic resonance in medicine 68 5 2012 1563 1569 10.1002/mrm.24135 22851199
Li X. Harrison D.M. Liu H. Jones C.K. Oh J. Calabresi P.A. Van Zijl P.C.M. Magnetic susceptibility contrast variations in multiple sclerosis lesions J. Magn. Reson. Imaging 43 2016 463 473 10.1002/jmri.24976 26073973
Li Y. Sethi S.K. Zhang C. Miao Y. Yerramsetty K.K. Palutla V.K. Gharabaghi S. Wang C. He N. Cheng J. Yan F. Haacke E.M. Iron content in deep gray matter as a function of age using quantitative susceptibility mapping: A multicenter study Front. Neurosci. 14 2021 1 13 10.3389/fnins.2020.607705
Liao H. Cai Z. Ye H. Chen Q.L. Zhang Y. Shaghaghi M. Lutz S.E. Chen W. Cai K. Combining in vivo proton exchange rate (kex) MRI with quantitative susceptibility mapping to further stratify the gadolinium-negative multiple sclerosis lesions Front. Neurosci. 16 2023 10.3389/fnins.2022.1105376
Liu M. Liu S. Ghassaban K. Zheng W. Dicicco D. Miao Y. Habib C. Jazmati T. Haacke E.M. Assessing global and regional iron content in deep gray matter as a function of age using susceptibility mapping J. Magn. Reson. Imaging 44 2016 59 71 10.1002/jmri.25130 26695834
Liu C. Wei H. Gong N.-J. Cronin M. Dibb R. Decker K. Quantitative susceptibility mapping: contrast mechanisms and clinical applications Tomography 1 2015 3 17 10.18383/j.tom.2015.00136 26844301
Manouchehrinia A. Westerlind H. Kingwell E. Zhu F. Carruthers R. Ramanujam R. Ban M. Glaser A. Sawcer S. Tremlett H. Hillert J. Age related multiple sclerosis severity score: disability ranked by age Mult. Scler. J. 23 2017 1938 1946 10.1177/1352458517690618
Marcille M. Hurtado Rúa S. Tyshkov C. Jaywant A. Comunale J. Kaunzner U.W. Nealon N. Perumal J.S. Zexter L. Zinger N. Bruvik O. Wang Y. Sweeney E. Kuceyeski A. Nguyen T.D. Gauthier S.A. Disease correlates of rim lesions on quantitative susceptibility mapping in multiple sclerosis Sci. Rep. 12 2022 1 10 10.1038/s41598-022-08477-6 34992227
Meaton I. Altokhis A. Allen C.M. Clarke M.A. Sinnecker T. Meier D. Enzinger C. Calabrese M. De Stefano N. Pitiot A. Giorgio A. Schoonheim M.M. Paul F. Pawlak M.A. Schmidt R. Granziera C. Kappos L. Montalban X. Rovira À. Wuerfel J. Evangelou N. Paramagnetic rims are a promising diagnostic imaging biomarker in multiple sclerosis Mult. Scler. J. 2212–2220 2022 10.1177/13524585221118677
Moher D. Shamseer L. Clarke M. Ghersi D. Liberati A. Petticrew M. Shekelle P. Stewart L.A. Estarli M. Barrera E.S.A. Martínez-Rodríguez R. Baladia E. Agüero S.D. Camacho S. Buhring K. Herrero-López A. Gil-González D.M. Altman D.G. Booth A. Chan A.W. Chang S. Clifford T. Dickersin K. Egger M. Gøtzsche P.C. Grimshaw J.M. Groves T. Helfand M. Higgins J. Lasserson T. Lau J. Lohr K. McGowan J. Mulrow C. Norton M. Page M. Sampson M. Schünemann H. Simera I. Summerskill W. Tetzlaff J. Trikalinos T.A. Tovey D. Turner L. Whitlock E. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement Rev. Esp. Nutr. Humana y Diet. 20 2016 148 160 10.1186/2046-4053-4-1
Pelizzari L. Bergsland N. Utriainen D. Viotti S. Baglio F. Mendozzi L. Cecconi P. Haacke E.M. Zamboni P. Laganà M.M. Semi-automatic detection of increased susceptibility in multiple sclerosis white matter lesions imaged with 1.5T MRI Biomed. Signal Process. Control 61 2020 10.1016/j.bspc.2020.102043
Pietroboni A.M. Colombi A. Contarino V.E. Russo F.M.L. Conte G. Morabito A. Siggillino S. Carandini T. Fenoglio C. Arighi A. De Riz M.A. Arcaro M. Sacchi L. Fumagalli G.G. Bianchi A.M. Triulzi F. Scarpini E. Galimberti D. Quantitative susceptibility mapping of the normal-appearing white matter as a potential new marker of disability progression in multiple sclerosis Eur. Radiol. 2022 10.1007/s00330-022-09338-6
Pontillo G. Cocozza S. Lanzillo R. Borrelli P. De Rosa A. Morra V.B. Tedeschi E. Palma G. Brain susceptibility changes in a patient with natalizumab-related progressive multifocal leukoencephalopathy: A longitudinal quantitative susceptibility mapping and relaxometry study Front. Neurol. 8 2017 1 5 10.3389/fneur.2017.00294 28138322
Pontillo, G., Petracca, M., Monti, S., Quarantelli, M., Lanzillo, R., Costabile, T., Carotenuto, A., Tortora, F., Elefante, A., Morra, V.B., Brunetti, A., Palma, G., Cocozza, S., 2023. Correction to: Clinical correlates of R1 relaxometry and magnetic susceptibility changes in multiple sclerosis: a multi-parameter quantitative MRI study of brain iron and myelin (European Radiology, (2022), 33, 3, (2185-2194), 10.1007/s00330-022-09154-y). Eur. Radiol. https://doi.org/10.1007/s00330-022-09279-0.
Pontillo G. Lanzillo R. Russo C. Stasi M.D. Paolella C. Vola E.A. Criscuolo C. Borrelli P. Palma G. Tedeschi E. Morra V.B. Elefante A. Brunetti A. Cocozza S. Determinants of deep gray matter atrophy in multiple sclerosis: A multimodal MRI study Am. J. Neuroradiol. 40 2019 99 106 10.3174/ajnr.A5915 30573464
Pudlac A. Burgetova A. Dusek P. Nytrova P. Vaneckova M. Horakova D. Krasensky J. Lambert L. Deep gray matter iron content in neuromyelitis optica and multiple sclerosis Biomed Res. Int. 2020 2020 10.1155/2020/6492786
R Core Team A Language and Environment for Statistical Computing R Found Stat. Comput. 2 2021 https://www.R–project.org
Rahmanzadeh R. Galbusera R. Lu P.J. Bahn E. Weigel M. Barakovic M. Franz J. Nguyen T.D. Spincemaille P. Schiavi S. Daducci A. La Rosa F. Absinta M. Sati P. Bach Cuadra M. Radue E.W. Leppert D. Kuhle J. Kappos L. Brück W. Reich D.S. Stadelmann C. Wang Y. Granziera C. A new advanced MRI biomarker for remyelinated lesions in multiple sclerosis Ann. Neurol. 92 2022 486 502 10.1002/ana.26441 35713309
Reichenbach J.R. Schweser F. Serres B. Deistung A. Quantitative susceptibility mapping: concepts and applications Clin. Neuroradiol. 25 2015 225 230 10.1007/s00062-015-0432-9
Rovira Á. Wattjes M.P. Tintoré M. Tur C. Yousry T.A. Sormani M.P. De Stefano N. Filippi M. Auger C. Rocca M.A. Barkhof F. Fazekas F. Kappos L. Polman C. Miller D. Montalban X. Evidence-based guidelines: MAGNIMS consensus guidelines on the use of MRI in multiple sclerosis - Clinical implementation in the diagnostic process Nat. Rev. Neurol. 11 2015 471 482 10.1038/nrneurol.2015.106 26149978
Rudko D.A. Solovey I. Gati J.S. Kremenchutzky M. Menon R.S. Multiple sclerosis: Improved identification of disease-relevant changes in gray and white matter by using susceptibility-based MR imaging Radiology 272 2014 851 864 10.1148/radiol.14132475 24828000
Sati P. Oh J. Todd Constable R. Evangelou N. Guttmann C.R.G. Henry R.G. Klawiter E.C. Mainero C. Massacesi L. McFarland H. Nelson F. Ontaneda D. Rauscher A. Rooney W.D. Samaraweera A.P.R. Shinohara R.T. Sobel R.A. Solomon A.J. Treaba C.A. Wuerfel J. Zivadinov R. Sicotte N.L. Pelletier D. Reich D.S. The central vein sign and its clinical evaluation for the diagnosis of multiple sclerosis: A consensus statement from the North American Imaging in Multiple Sclerosis Cooperative Nat. Rev. Neurol. 12 2016 714 722 10.1038/nrneurol.2016.166 27834394
Schmalbrock P. Prakash R.S. Schirda B. Janssen A. Yang G.K. Russell M. Knopp M.V. Boster A. Nicholas J.A. Racke M. Pitt D. Basal ganglia iron in patients with multiple sclerosis measured with 7T quantitative susceptibility mapping correlates with inhibitory control Am. J. Neuroradiol. 37 2016 439 446 10.3174/ajnr.A4599 26611996
Schweser F. Martins R.D. Raffaini Duarte Martins A.L. Hagemeier J. Lin F. Hanspach J. Weinstock-Guttman B. Hametner S. Bergsland N. Dwyer M.G. Zivadinov R. Mapping of thalamic magnetic susceptibility in multiple sclerosis indicates decreasing iron with disease duration: A proposed mechanistic relationship between inflammation and oligodendrocyte vitality Neuroimage 167 2018 438 452 10.1016/j.neuroimage.2017.10.063 29097315
Schweser F. Hagemeier J. Dwyer M.G. Bergsland N. Hametner S. Weinstock-Guttman B. Zivadinov R. Decreasing brain iron in multiple sclerosis: The difference between concentration and content in iron MRI Hum. Brain Mapp. 42 2021 1463 1474 10.1002/hbm.25306 33378095
Sterne, J.A.C., Savović, J., Page, M.J., Elbers, R.G., Blencowe, N.S., Boutron, I., Cates, C.J., Cheng, H.-Y., Corbett, M.S., Eldridge, S.M., Emberson, J.R., Hernán, M.A., Hopewell, S., Hróbjartsson, A., Junqueira, D.R., Jüni, P., Kirkham, J.J., Lasserson, T., Li, T., McAleenan, A., Reeves, B.C., Shepperd, S., Shrier, I., Stewart, L.A., Tilling, K., White, I.R., Whiting, P.F., Higgins, J.P.T., 2019. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ l4898. https://doi.org/10.1136/bmj.l4898.
Straub S. El-Sanosy E. Emmerich J. Sandig F.L. Ladd M.E. Schlemmer H.P. Quantitative magnetic resonance imaging biomarkers for cortical pathology in multiple sclerosis at 7 T NMR Biomed. 36 2023 10.1002/nbm.4847
Stüber C. Morawski M. Schäfer A. Labadie C. Wähnert M. Leuze C. Streicher M. Barapatre N. Reimann K. Geyer S. Spemann D. Turner R. Myelin and iron concentration in the human brain: A quantitative study of MRI contrast Neuroimage 93 2014 95 106 10.1016/j.neuroimage.2014.02.026 24607447
Suh C.H. Kim S.J. Jung S.C. Choi C.G. Kim H.S. The “Central Vein Sign” on T2*-weighted images as a diagnostic tool in multiple sclerosis: a systematic review and meta-analysis using individual patient data Sci. Rep. 9 2019 18188 10.1038/s41598-019-54583-3 31796822
Sun H. Walsh A.J. Lebel R.M. Blevins G. Catz I. Lu J.Q. Johnson E.S. Emery D.J. Warren K.G. Wilman A.H. Validation of quantitative susceptibility mapping with Perls’ iron staining for subcortical gray matter Neuroimage 105 2015 486 492 10.1016/j.neuroimage.2014.11.010 25462797
Taege Y. Hagemeier J. Bergsland N. Dwyer M.G. Weinstock-Guttman B. Zivadinov R. Schweser F. Assessment of mesoscopic properties of deep gray matter iron through a model-based simultaneous analysis of magnetic susceptibility and R2* - A pilot study in patients with multiple sclerosis and normal controls Neuroimage 186 2019 308 320 10.1016/j.neuroimage.2018.11.011 30445148
Thompson A.J. Banwell B.L. Barkhof F. Carroll W.M. Coetzee T. Comi G. Correale J. Fazekas F. Filippi M. Freedman M.S. Fujihara K. Galetta S.L. Hartung H.P. Kappos L. Lublin F.D. Marrie R.A. Miller A.E. Miller D.H. Montalban X. Mowry E.M. Sorensen P.S. Tintoré M. Traboulsee A.L. Trojano M. Uitdehaag B.M.J. Vukusic S. Waubant E. Weinshenker B.G. Reingold S.C. Cohen J.A. Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria Lancet Neurol. 17 2018 162 173 10.1016/S1474-4422(17)30470-2 29275977
Tolaymat B. Zheng W. Chen H. Choi S. Li X. Harrison D. Sex-specific differences in rim appearance of multiple sclerosis lesions on quantitative susceptibility mapping Mult. Scler. Relat. Disord. 45 2020 1 18 10.1016/j.msard.2020.102317
Valizadeh A. Moassefi M. Barati E. Ali Sahraian M. Aghajani F. Fattahi M.R. Correlation between the clinical disability and T1 hypointense lesions’ volume in cerebral magnetic resonance imaging of multiple sclerosis patients: A systematic review and meta-analysis CNS Neurosci. Ther. 27 2021 1268 1280 10.1111/cns.13734 34605190
Vancampfort D. Knapen J. Probst M. Scheewe T. Remans S. De Hert M. A systematic review of correlates of physical activity in patients with schizophrenia Acta Psychiatr. Scand. 2012 10.1111/j.1600-0447.2011.01814.x
Veritas Health Innovation, M.A., n.d. Covidence systematic review software [WWW Document]. URL https://www.covidence.org (accessed 4.30.23).
Wang Y. Liu T. Quantitative susceptibility mapping (QSM): Decoding MRI data for a tissue magnetic biomarker Magn. Reson. Med. 73 2015 82 101 10.1002/mrm.25358 25044035
Weber C.E. Wittayer M. Kraemer M. Dabringhaus A. Bail K. Platten M. Schirmer L. Gass A. Eisele P. Long-term dynamics of multiple sclerosis iron rim lesions Mult. Scler. Relat. Disord. 57 2022 103340 10.1016/j.msard.2021.103340
Wiggermann V. Hametner S. Hernández-Torres E. Kames C. Endmayr V. Kasprian G. Höftberger R. Li D.K.B. Traboulsee A. Rauscher A. Susceptibility-sensitive MRI of multiple sclerosis lesions and the impact of normal-appearing white matter changes NMR Biomed. 30 2017 1 12 10.1002/nbm.3727
Wisnieff C. Ramanan S. Olesik J. Gauthier S. Wang Y. Pitt D. Quantitative susceptibility mapping (QSM) of white matter multiple sclerosis lesions: Interpreting positive susceptibility and the presence of iron Magn. Reson. Med. 74 2015 564 570 10.1002/mrm.25420 25137340
Yao Y. Nguyen T.D. Pandya S. Zhang Y. Hurtado Rúa S. Kovanlikaya I. Kuceyeski A. Liu Z. Wang Y. Gauthier S.A. Combining Quantitative Susceptibility Mapping with Automatic Zero Reference (QSM0) and Myelin Water Fraction Imaging to Quantify Iron-Related Myelin Damage in Chronic Active MS Lesions Am. J. Neuroradiol. 39 2018 303 310 10.3174/ajnr.A5482 29242359
Yu F.F. Chiang F.L. Stephens N. Huang S.Y. Bilgic B. Tantiwongkosi B. Romero R. Characterization of normal-appearing white matter in multiple sclerosis using quantitative susceptibility mapping in conjunction with diffusion tensor imaging Neuroradiology 61 2019 71 79 10.1007/s00234-018-2137-7 30539215
Zhang Y. Gauthier S.A. Gupta A. Chen W. Comunale J. Chiang G.C.Y. Zhou D. Askin G. Zhu W. Pitt D. Wang Y. Quantitative susceptibility mapping and R2∗ measured changes during white matter lesion development in multiple sclerosis: Myelin breakdown, myelin debris degradation and removal, and iron accumulation Am. J. Neuroradiol. 37 2016 1629 1635 10.3174/ajnr.A4825 27256856
Zhang Y. Gauthier S.A. Gupta A. Comunale J. Chia-Yi Chiang G. Zhou D. Chen W. Giambrone A.E. Zhu W. Wang Y. Longitudinal change in magnetic susceptibility of new enhanced multiple sclerosis (MS) lesions measured on serial quantitative susceptibility mapping (QSM) J. Magn. Reson. Imaging 44 2016 426 432 10.1002/jmri.25144 26800367
Zhang Y. Gauthier S.A. Gupta A. Tu L. Comunale J. Chiang G.C.Y. Chen W. Salustri C.A. Zhu W. Wang Y. Magnetic susceptibility from quantitative susceptibility mapping can differentiate new enhancing from nonenhancing multiple sclerosis lesions without gadolinium injection Am. J. Neuroradiol. 37 2016 1794 1799 10.3174/ajnr.A4856 27365331
Zhang X.S. Nguyen T.D. Ruá S.M.H. Kaunzner U.W. Pandya S. Kovanlikaya I. Spincemaille P. Wang Y. Gauthier S.A. Quantitative susceptibility mapping of time-dependent susceptibility changes in multiple sclerosis lesions Am. J. Neuroradiol. 40 2019 987 993 10.3174/ajnr.A6071 31097429
Zhang H. Nguyen T.D. Zhang J. Marcille M. Spincemaille P. Wang Y. Gauthier S.A. Sweeney E.M. QSMRim-Net: Imbalance-aware learning for identification of chronic active multiple sclerosis lesions on quantitative susceptibility maps NeuroImage Clin. 34 2022 102979 10.1016/j.nicl.2022.102979
Zinger N. Ponath G. Sweeney E. Nguyen T.D. Lo C.H. Diaz I. Dimov A. Teng L. Zexter L. Comunale J. Wang Y. Pitt D. Gauthier S.A. Dimethyl fumarate reduces inflammation in chronic active multiple sclerosis lesions Neurol. Neuroimmunol. Neuroinflammation 9 2022 1 10 10.1212/NXI.0000000000001138
Zivadinov R. Havrdová E. Bergsland N. Tyblova M. Hagemeier J. Seidl Z. Dwyer M.G. Vaneckova M. Krasensky J. Carl E. Kalincik T. Horáková D. Thalamic atrophy is associated with development of clinically definite multiple sclerosis Radiology 268 2013 831 841 10.1148/radiol.13122424 23613615
Zivadinov R. Tavazzi E. Bergsland N. Hagemeier J. Lin F. Dwyer M.G. Carl E. Kolb C. Hojnacki D. Ramasamy D. Durfee J. Weinstock-Guttman B. Schweser F. Brain iron at quantitative MRI is associated with disability in multiple sclerosis Radiology 289 2018 487 496 10.1148/radiol.2018180136 30015589
