Abstract: Brain tumor is a high-risk disease. The information about location and size is critical during treatment. In order to effectively detect and diagnose brain tumors, this paper presents an improved multi-modal brain tumor segmentation hydroid algorithm for multi-modal MRI brain tumor images: FLAIR, T2 and T1C. First, the preprocessing image is obtained by median filtering and fast fuzzy C-means clustering of MRI images of the three modals. And. the preprocessed image is linearly fused in the proportion of FLAIR:T2:T1C=5:4:1. Then, the fast fuzzy C-means algorithm is used to segment the preprocessed images, to produce the under-divided image of brain tumors. Finally, the under-divided images of brain tumors are processed by the mixed level set algorithm to obtain the segmented images of brain tumors. Compared with the gold standard, the average Dice of the result reached 0.90. Compared with other similar algorithms, the Dice is the best. the real-time and stability of ours algorithm are better, which could meet the clinical demand.