Objective An analysis method of brain complexity based on resting state fMRI data was proposed for the lack of objective evaluation standard of cognitive impairment in cerebral tumor patients currently. Methods Firstly, we preprocessed BOLD fMRI images of 24 cerebral tumor patients and 18 healthy volunteers respectively, calculated the mean whole-brain sample entropy and the whole-brain voxel sample entropy maps, and then determined the tumor site for region of interest (ROI) area to analyze the correlation between function connection and cross-sample entropy. Results Whole brain entropy of healthy volunteers was significant higher than cerebral tumor patients (F=15.907，P=0.000), and negative correlation between function connection and cross sample entropy was observed. And there were significant differences in the left frontal, left temporal and cuneus brain functional areas between patients with brain tumor of the left frontal or left temporal and healthy volunteers. The results were consistent with clinical diagnosis results. Conclusions This analysis method of brain complexity based on sample entropy help to illustrate that tumor will affect the complexity of the brain, and the increase of tumor grade leads to the decrease of the brain complexity. The results in this method can be used as one of the quantitative parameters to evaluate the tumor grade.