Objective To evaluate three different tracer-kinetic models used for dynamic contrast-enhanced MRI (DCE-MRI) data processing in the prediction of the histopathologic grade of cerebral gliomas. Methods Forty-one patients with histopathologically graded gliomas (grade Ⅱ 13, grade Ⅲ 14, grade Ⅳ 14) were imaged with DCE-MRI from March,2013 to November,2014. The values of Ktrans of gliomas were obtained by three tracer-kinetic models, which were Patlak model, Tofts model and Extended Tofts Linear model. All data were analyzed statistically by a Graphpad 6.0 statistical software. Comparison of the differences of Ktrans among different grades of gliomas was conducted using the Kruskal-Wallis test and Dunn's multiple comparisons test for the data not conform to normal distribution. Correlations of Ktrans values among those three models were analyzed using linear regression analysis, The differences of Ktrans between low grades and high grades of gliomas was conducted using the Mann-Whitney U test. Receiver operating characteristic (ROC) curve analyses were performed to determine the cut-off values for Ktrans to distinguish different low grades and high grades of gliomas. Results The Ktrans values obtained by Patlak model was 0.008 (0.004-0.043) min-1for grade Ⅱ, 0.021 (0.005-0.088) min-1 for grade Ⅲ, and 0.035 (0.017-0.061) min-1 for grade Ⅳ. The Ktrans values obtained by Tofts model was 0.085 (0.041-0.158) min-1 for grade Ⅱ, 0.140 (0.063-0.315) min-1 for grade Ⅲ, 0.229 (0.126-0.419) min-1 for grade Ⅳ. The Ktrans values obtained by Extended Tofts Linear model was 0.012 (0.004-0.092) min-1 for grade Ⅱ, 0.048 (0.010-0.188) min-1 for grade Ⅲ, 0.094 (0.036-0.215) min-1 for grade Ⅳ. All the Ktrans values obtained by three models increased when the histological grades increased, with statistical significance between grade Ⅱ and Ⅲ (H=18.31, 18.09,20.18,P<0.05). Ktrans values among the three models had good linear correlations. The Ktrans obtained by Extended Tofts Linear model had good linear correlations with both Patlak model and Tofts model (r=0.933, 0.893, P<0.05) , and the Ktrans obtained by Patlak model had less linear correlation with Tofts model (r=0.822, P<0.05). The Ktrans values were statistically different between LGG and HGG (P<0.01). The cut-off value of Ktrans provided good combination of sensitivity and specificity in the differentiation between LGG and HGG, and the Ktrans obtained by Extended Tofts Linear model had the best sensitivity and specificity among the three models (sensitivity 92.3%, specificity 85.7%). Conclusions Ktrans from DCE-MRI has a high performance in predicting the histopathologic grade of brain glioma, and the Extended Tofts Linear model is more suitable for the evaluation of cerebral glioma. Key words: Glioma; Magnetic resonance imaging; Models, theoretical