Texture Analysis Of Quantitative Apparent Diffusion Coefficient Maps To Differentiate Between High-Grade Gliomas And Metastatic Brain Tumors

BACKGROUNDHigh-grade gliomas(HGG)and metastatic brain tumors(METS)are considered to be the most commonly seen adult brain tumors.These tumors show vast differences in metabolism,vasculature,micro-environmental factors,structure and function which make it imperative to differentiate them.Moreover,they appear to be indistinguishable on conventional MRI as these two entities show similar imaging characteristics and contrast enhancement patterns which display lack of differentiation in many cases.An approach to develop a better treatment plan for these tumors can be initiated if an appropriate preoperative differentiation between these lesions is possible.Previously,innumerable studies have been undertaken in order to distinguish between these tumors non-invasively using advanced medical imaging techniques but with minimal results as most of them included either intratumoral or peritumoral regions or very small ROIs which underestimated intratumoral heterogeneity significantly.Magnetic resonance texture analysis(MRTA)is a novel image analyzing technique that helps us acquire up to hundreds of parameters to evaluate the tumor as well as to assess the tumor heterogeneity based on the entire tumor comprehensively.Furthermore,it helps to describe all aspects of texture images such as intensity,shape,size,volume,tumorphenotype,genetic and micro-environmental factors which are not visible to the naked eye in detail with increased accuracy.Therefore,it is inevitable to develop methods to ease the process of differentiation between HGG and METS by using non-invasive techniques based on whole tumor volume to attain positive results.PURPOSEThe purpose of this study is to analyze high grade gliomas(HGG)and metastatic brain tumors(METS)using texture analysis(TA)parameters of apparent diffusion coefficient(ADC)maps based on entire tumor region and thereby to use its diagnostic value of differentiating them.MATERIALS AND METHODSA retrospective study was carried out at the Department of Radiology at The First Affiliated Hospital of Dalian Medical University.The magnetic resonance images(MRI)including sagittal T1 WI,axial T1 WI,axial T2 WI,axial T2 FLAIR,axial diffusion weighted images(DWI)and axial contrasted T1 WI of 50 patients with HGG and 32 patients with METS which were previously confirmed by pathology were evaluated.All the scans were performed using GE Signa HDxt 3.0 tesla MR machine(8-channel head coil).The ADC maps of all cases were processed using ADW 4.6 workstation Functool2 software.Each slice of the ADC maps obtained from the DWI images was saved as digital imaging and communications in medicine(DICOM)images for further procession.With the help of Omnikinetics software,ADC maps that included both the entire tumor and peripheral edema were delineated in each slice.The ROIs were drawn without excluding the necrotic,hemorrhagic and cystic areas.Multiple regions of interest(ROI)images obtained from each slice of the ADC maps were merged into a single 3D image.The histogram curves of each case were also recorded for further reference.Subsequently,33 texture parameters namely: 15 First order histogram parameters(FOHP)that included minimum intensity(MIN),maximum intensity(MAX),median intensity(MI),skewness,kurtosis,mean deviation(MD),relative deviation(RD),standard deviation(SD),range,voxel value sum(VVS),root mean square(RMS),uniformity,variance,mean value(MV)and volume count(VC);8 gray-level co-occurrence matrix(GLCM)parameters that included energy,entropy,inertia,correlation,inverse difference moment(IDM),cluster shade(CS),cluster prominence(CP)and haralick correlation(HC)and 10 gray level run length maxia(GLRLM)parameters that included short run emphasis(SRE),long run emphasis(LRE),gray level non uniformity(GLN),run length non uniformity(RLN),low gray level run emphasis(LGLRE),high gray level run emphasis(HGLRE),short run low gray level emphasis(SRLGLE),short run high gray level emphasis(SRHGLE),long run low gray level emphasis(LRLGLE)and long run high gray level emphasis(LRHGLE)of these maps were listed.The entire data was analyzed using SPSS 20.0.Shapiro-Wilk test was used to determine normality of the parameters.The TA parametric values like VC,VVS and IDM were compared between HGG and METS using independent sample t test and non-parametric ones like MIN,MAX,MI,skewness,kurtosis,MD,RD,SD,range,RMS,uniformity,variance,MV,energy,entropy,inertia,correlation,IDM,CS,CP,HC,SRE,LRE,GLN,RLN,LGLRE,HGLRE,SRLGLE,SRHGLE,LRLGLE and LRHGLE were compared between HGG and METS using Mann-Whitney U test.Receiver Operating Characteristic(ROC)curve tests to calculate Area Under Curve(AUC)and p value and thereby determine the diagnostic ability of each independent risk factor were carried out.The efficacies of the parameters were assessed in terms of cutoff values,sensitivity and specificity.P < 0.05 was considered statistically significant.RESULTS1)VC(p=0.009),VVS(p=0.013),RMS(p=0.000),RD(p=0.042),correlation(p=0.016),energy(p=0.000),entropy(p=0.000),GLN(p=0.009),RLN(p=0.005),LGLRE(p=0.023),SRLGLE(p=0.027),LRLGLE(p=0.000)and LRHGLE(p=0.030)between HGG and METS are statistically significant.The Mean ± S.D.values of VVS,VC,RD,energy,entropy,correlation GLN,RLN and LRHGLE of HGG are found to be higher than those of METS while the Mean ± S.D.values of RMS,LGLRE,SRLGLE and LRLGLE are found to be higher in METS as compared to HGG.2)ROC curve analysis showed good differentiating diagnostic ability for HGG and METS with regards to entropy(AUC=0.955,P=0.009),RLN(AUC=0.684,P=0.014),VC(AUC=0.672,P=0.042),GLN(AUC=0.671,P=0.000),VVS(AUC=0.663,P =0.016),correlation(AUC=0.659,P =0.010)and RD(AUC=0.634,P =0.005).The biggest AUC(0.955)was observed in entropy with the sensitivity of 91.8% and the specificity of 100%.CONCLUSIONUsing MRTA based on whole tumor,13 parameters namely VVS,VC,RMS,RD,energy,entropy,correlation,GLN,RLN,LGLRE,SRLGLE,LRHGLE,LRLGLE were found to have significant difference between HGG and METS.ROC curve tests also proved 7 out of these 13 parameters to have strong diagnostic ability.Therefore,we can conclude that Texture Analysis of Quantitative Apparent Diffusion Coefficient maps based on entire region of tumor aids in analyzing the tumor heterogeneity of both HGG and METS and entropy can be considered as the best differentiating diagnostic index of HGG and METS.