Value of MRI histogram analysis in differential diagnosis of brain high-grade glioblastomas,primary lymphoma and metastatic tumor / 实用放射学杂志

ABSTRACT

Objective To investigate the feasibility of histogram analysis in differentiating brain high-grade glioblastomas,primary lymphoma from metastatic tumor.Methods 26 cases of brain high-grade glioblastomas,22 cases of primary lymphoma and 18 cases of metastatic tumor confirmed by postoperative pathological were analyzed retrospectively.Delineation of ROI and extraction of texture parameters were performing by using Mazda software.The histogram parameters,including Mean,Variance,Skewness,Kurtosis,Perc0.1%,Perc10%,Perc50%, Perc90% and Perc99% were analyzed statistically,and the ROC was then established.Results Mean,Perc0.1%,Perc10%and Perc50% exhibited statistically significant differences (P＜0.05).The best diagnostic parameters for differentiation between brain high-grade glioblastomas and primary lymphoma,primary lymphoma and metastatic tumor,and brain high-grade glioblastomas and metastatic tumor were Perc0.1%,Perc0.1%and Kurtosis.The AUC for these preferred diagnostic parameters were 0.937,0.879 and 0.7 1 8,respectively,with optimal thresholds of 50,70 and －0.43,sensitivity and specificity of 90.9% and 88.5%,77.3% and 88.9%,and 61.5% and 77.8%.Conclusion The histogram analysis of MRI images contributes to differentiate quantitatively between brain high-grade glioblastomas,primary lymphoma and metastatic tumor.