Abstract: Objective: The study aimed to investigate the comparison of different regions of interest (ROI) on reproducibility of semi-quantitative parameters in dynamic contrast-enhanced MRI (DCE-MRI) of epithelial ovarian cancer. Methods: DCE-MRI data of 60 epithelial ovarian cancers, confirmed by surgery and histopathology, were analyzed retrospectively. Four DCE-MRI semi-quantitative parameters, including the peak time (TTP), the maximum concentration (MAX), the area under the concentration time curves (AUC) and the maximal slope (MS), were measured with postprocessing software. The two physicians used the irregular shape, the circular shape and the rectangular method to delineate the ROI on the semi-quantitative AUC pseudo-color image. The data were calculated by two observers independently, observer 1 reevaluated all of the images, which were presented in a different order, one month after the first session. The differences of DCE-MRI semi-quantitative values with different ROI methods, which can be compared in One-Way ANOVA.Intra and interobserver agreement, were evaluated using the intraclass correlation coefficient (ICC), coefficient of variation (CV) and Bland-Altman analysis. Results: There was no significant difference among the three groups (P>0.05). Intra- and interobserver reproducibility were excellent by polygon ROI, The values of ICC in the intraobserver and the interobserver were (0.975~0.985) and (0.961~0.984), respectively; CV were (2.73~4.21) and (3.48~4.86), respectively. The results showed that polygon repeat the measurement of consistency in the range of small, good accuracy. Conclusions: Epithelial ovarian cancer with different ROI had no significant effect on the semiquantitative parameters, and showed a high degree of consistency.