Quantitative Susceptibility Mapping In Multiple Sclerosis And Brain Tumors-initial Experience

Part ? Rim on QSM in Multiple Sclerosis Lesions Identifies a Subset of Lesions Susceptible to More Myelin DamageObjective:Iron accumulation within microglia/macrophages occurs at the rim of chronic active multiple sclerosis(MS)lesions.This study utilized quantitative susceptibility mapping(QSM)and myelin water fraction(MWF)to identify MS lesions with a rim of iron deposition and assess the extent of myelin damage within these lesions.Materials and Methods:Forty six MS patients had two longitudinal QSM scans reconstructed from multi-echo gradient echo(GRE)imaging with a mean interval of 28.9±11.4 months.MWF mapping reconstructed from Fast Acquisition with Spiral Trajectory and T2prep(FAST-T2)was obtained at the second time point to assess myelin damage.A total of 110 QSM hyperintense rim positive(rim+)and 404 rim negative(rim-)lesions were identified.Mixed-effects-models were utilized to assess lesion QSM and MWF values while controlling for multiple lesions per patient,T2w lesion volume,gender,patient age and longitudinal time interval.Results:The volume of rim+ lesions was higher on QSM as compared to T2w(p<0.01).QSM and MWF in rim+ lesions decreased from rim to core,which is consistent with rim iron deposition.In the mixed-effects-model.QSM were on average 8.21 ppb higher in rim+lesions compared to rim-lesions.p<0.0001.All rim+ lesions retained a hyperintense rim over time,with slightly increasing QSM values withiin rim and core regions,p<0.0001.Whole lesion MWF for rim+ and rim-lesions were 0.048 ± 0.02 and 0.064 ± 0.02,respectively.Rim + lesions had on average 0.01 lower MWF as compared to rim-,p<0.0001.The volume of the rim at initial QSM was associated with the follow-up MWF,p=0.005.Conclusions:QSM rim + lesions maintained a hyperintense rim.demonstrated an increase in susceptibility and had lower MWF.Our results are consistent with identification of chronic active MS lesions and may provide a target for therapeutic interventions to reduce myelin damage.Part II Grading of Astrocytoma on Quantitative Susceptibility Mapping with and without Textural Analysis ToolObjective:Angiogenic vascular proliferation plays an important role in tumor malignancy.Its associated micro and macro hemorrhages can be measured by quantitative susceptibility mapping(QSM).which is highly sensitive to paramagnetic iron in blood degradation products.The objective was to determine whether QSM could grade astrocytic brain tumors accurately with and without textural analysis(TA).Materials and methods:In this study,forty two patients with first diagnosed astrocytoma were recruited from two institutions(n=14 for grade ?.n=7 for grade ? and n=21 for grade IV).The 42 patients were randomly divided into training group(n=22)and test group(n=20).Multi-echo gradient echo(GRE)imaging plus standard T1,T2 and post contrast T1 weighted imaging were acquired.QSM was generated from GRE data using a morphology enabled dipole inversion(MEDI)method.Regions of interest(ROI)were defined consisting each slice from entire tumor volume on post contrast T1 or T2 weighted images if there isn't any contrast enhancement.Mean relative susceptibility was compared between different grades using independent-samples t test and ANOVA.For training data set.ROIs were loaded into MaZda package to perform texture analysis using linear discriminant analysis(LDA)on both QSM and FLAIR images and obtain the most discriminant factor(MDF1)values,respectively.Receiver operating characteristic(ROC)curves were constructed to determine the diagnostic accuracy and obtain the cut-off values.Using these cut-off values to determine the misclassification error rate of test data set.Results:Hyperintense intratumoral susceptibility based signals were found to increase with increasing tumor grade.Over the entire tumor volume,low grade(grade ?)and high grade(grade ? and grade ?)astrocytomas could be separated by mean susceptibility value(p=0.0018).Grade ? from grade ?(p=0.004).Grade ? from grade ?(p=0.048)could also be differentiated by mean susceptibility value.For texture analysis training data set on QSM,MDF1 values were significantly different for low grade and high grade astrocytomas(p<0.01)with a cutoff value of 0.004485.Grade ? and ?.Grade ? and ?.Grade ? and IV astrocytomas could also be separated with MDF1(p<0.01)with cutoff values of 0.0184.0.001835,-0.00954.For texture analysis test data set on QSM.using the same features and LDA model to do texture analysis could correctly classify 19/20(95%)cases in differentiating low grade from high grade astrocytomas using cutoff value of 0.004485.In addition,it could successfully classify 16/20(80%)patients,differentiating grades ?,?and IV in the test set using cutoff values of 0.0184 and-0.00954.For texture analysis training data set on FLAIR,MDF1 values were significantly different for low grade and high grade astrocytomas(p<0.01)with a cutoff value of 0.00003249.The FLAIR images cannot differentiate grade ? astrocytomas from grade ? ones even with texture analysis.Only Grade ? and IV,Grade ? and ? astrocytomas could be separated with MDF1(p<0.01)with cutoff values of 0.0001948 and-0.00006489.For texture analysis test data set on FLAIR,using the same features and LDA model to do texture analysis could correctly classify 17/20(95%)cases in differentiating low grade from high grade astrocytomas using cutoff value of 0.00003249.Conclusion:QSM could provide more information for grading astrocytoma.TA promises to be a useful tool for grading astrocytoma on multi-institutional QSM data sets.This method requires further validation in a larger cohort study.