Application Study Of MRI Subtraction In Multiple Sclerosis

IntroductionMultiple Sclerosis is a kind of recurrence,gradual progression and worse central neural immune disease.Now MRI is only one tool to impersonality monitor and evaluate MS.MRI subtraction can partly cancel out normal anatomic structure and stable lesion and directly show lesion changes.However,this measurement based on pixel has some limitation.Patient motion,MR instrument change,magnetic filed,MRI sequence,MRI thickness,MRI artifact,imaging interpolaton method,rescan patient's head position and angle result in coregistered image bias.Therefore,these limit partly application of MRI subtraction.The studies,which used same MRI parameter and MRI scanner,have improved reproducibility and accuracy in MRI subtraction.The good reproducibility of identifying lesion change between independent readers has been reported by research,especially in the new lesion.But some biases were found in the enlarging and resolving lesion.How to reduce bias in MRI subtraction is a key issue in MRI subtraction..MRI subtraction can overcome defect and disadvantage of T2 weighted MR image in detecting lesion change,and explore to some extent diffusion pathological characters(such as inflammation,gliosis,edema,demyelination,Wallerian degeneration and axon loss).Directly to segment MS lesion change in MRI subtraction has not been reported yet.To improve accuracy of MS lesion detection in subtraction image and reduce bias in subtraction image,MR subtraction image artifacts were analyzed.To compared with accuracy and reproducibility of two measurements -conventional semi-automated segmentation(CSEG)and subtraction images segmentation(SSEG)and analyze correlationship between lesion changes and brain parenchymal fraction(BPF),disease duration and MRI intervals.To assess the sensitivity and reproducibility of two image analysis strategies:conventional imaging(CSEG)single-time point segmentation for determination of numeric difference between total lesion volumes versus the segmentation of subtraction images(SSEG).To analyze correlation between their lesion volume change and brain parenchymal fraction(BPF),disease duration and MRI intervals.MethodsTen patients with MS(6 women;9 relapsing-remitting,1 secondary progressive; mean age 32.2 years,MR interval 30 minutes)were studied as scan-rescan group,and twenty-one MS patients(17 women;20 relapsing-remitting,1 secondary progressive; mean age 43.6 years,average MR interval:average±SD=3.1±0.89 years)were severed as follow-up group.Two groups were retrospectively studied.Each participant underwent MR imaging on a 1.5 T MR system(Signa,GE Medical Systems, Milwaukee,Wisconsin).Dual-echo PD/T2 weighted MR images(TE=30/80ms, TR=3000ms,192 phase-encoding steps,0.93×0.93×3 mm~3 nominal voxel size,with no interslice gaps)were acquired using a standard circular polarized head coil.Enlarging, resolving and new lesion needed to be greater than or equal to 4 pixels,less than or equal 4 and greater than 3 mm respectively per slice and confirmed on at least two contiguous images.Two blinded radiologists identified lesion changes using Yin Yang artifacts in the comparison between registered,original and subtracted MR images,and then one of them marked lesion change by using Yin Yang artifact and not using it. Template-driven segmentation plus partial volume effect correction(TDS+)was performed and manually edited by an expert radiologist on each of the original images to obtain lesion volumes for each time-point.Each of 31 patients' paired proton density weighted serial images were co-registered,intensity normalized,filtered(Gaussion) and subtracted to produce MRI subtraction.New,enlarging and resolving lesions were marked by using boxed ROIs,a semi-automated 3D segmentation of change was performed in ROIs boxes.Each automated result was then edited by an expert operator and adjusted manually where necessary.In scan-rescan group,positive ratio of Yin Yang artifact was studied on basis of total MS lesions.Yin Yang artifact classification and ratio was done in subtraction imaging by comparing original,registered and subtracted MRI.Intra agreement of two measurement(using Yin Yang artifact and not)done by one radiologist was compared by statistics analysis.Difference between two radiologists using Yin Yang artifact identification method was compared,and its reproducibility was tested by Bland-Altman.To compute the reproducibility of the CSEG and SSEG methods,we analyzed the lesion volume differences in the scan/rescan group(10 patients)using the coefficient of variability(COV)and the Bland and Altman method[25].For the CSEG method,the Bland and Altman analysis was performed by comparing the lesion volume measurements from the scan/rescan group.For the SSEG method the Bland and Altman analysis was performed using the baseline lesion volume(measured with CSEG)and the sum of baseline lesion volume and the net change in lesion volume measured by the SSEG method.The same strategy was used to calculate the COV(defined as the ratio between the standard deviation and the mean).In addition,the percentage systematic bias of the scan/rescan was calculated by dividing net change by the baseline scan total lesion volume.A Wilcoxon test was performed to assess the average differences bias between CSEG and SSEG results. Brain parenchymal fraction defined as ratio between brain parenchymal and intracranial cavity(ICC),i.e.:BPF=1-CSF/ICC(CSF,cranial spinal fluid).For the 21 MS patients,the correlation coefficients between the lesion changes(lesion volume change,lesion volume yearly change and lesion volume percentage yearly change)and clinical characters(BPF,disease duration and MRI interval time)were performed by Spearmen rank correlation test.Lesion volume percentage yearly change was ratio between annual lesion change and basal line lesion volume.Results Yin Yang artifact was a majority artifact in the subtraction image,occurred at 24.6%in total MS lesion.Yin Yang artifact was classified as four type according to image pixel shift orientation(x,y,z,xy,xz,yz and xyz axial).Ⅰ,Ⅱ,ⅢandⅣtype occurred respectively 15.12%,5.06%,53.16%and 26.58%.In the subtraction image, typeⅢoccurred more than three and half times than typeⅠdid,and nearly two times than typeⅣdid.The typeⅡwas found as the minimum one.The agreement between differential readers(Kappa value=0.488)was batter using Yin Yang artifact than one reader did using Yin Yang artifact or not(Kappa value=0.454).Reproducibility of CSEG was better than one of SSEG.Average yearly lesion percentage change,average lesion change and average yearly lesion change with CSEG/SSEG method were 11.3±16.62%/26.7±42.16%,0.895±1.625cm~3/1.160±1.564cm~3 and 0.279±0.474cm~3 /0.356±0.447cm~3 respectively.A significant statistics difference between SSEG and CSEG at the average yearly percentage lesion change was tested by using Wilcoxon test(p＜0.05).SSEG had a negative correlation with BPF change,and CSEG did not. Negative correlation between average yearly percentage lesion change with CSEG (r=-0.514,p=0.0215)/SSEG(r=-0.552,p=0.0135)and disease duration was shown.A negative correlation between yearly lesion volume change with SSEG and disease duration was founded(r=-0.508 p=0.0231),and that was not showed in CSEG(r=-0.360 p=0.1077).MRI intervals had no relation with lesion volume change in both measurementsConclusionsYin Yang artifact showed counterposed white and black borders on the lesion periphery.A perfect registration image combined with Yin Yang artifacts identification measurement contributed to detect MS lesion change.Yin-Yang artifact was a main artifact in subtraction image.According to image pixel shift orientation,it defined four types,simple type(Ⅰ,Ⅱ,Ⅲtype)and complicated type(ⅣType).Yin Yang artifact identification measurement had a good reproducibility and improved accuracy of MS lesion detection to reduce bias.Yin Yang artifact measurement has played a potential role in identifying MS lesion in MRI subtraction.SSEG mothed was more sensitive in the MS lesion size and intensity than CSEG's did,and had a better reproducibility. SSEG provided an impotant role measurement for clinical therapy and monitoring in MS.