3D Segmentation Of Brain Tumors In MRI Based On Improved CMF And Multi-Modality

Brain tumor diseases are direct threats to the lives of patients.Brain tumor segmentation is a key step in the treatment of brain tumors.Because of its high resolution to soft tissues,magnetic resonance imaging(MRI)is considered as a preferred medical imaging tool for brain diseases.In order to solve the problem of insufficient segmentation of brain tumors in MRI caused by noise,poor contrasts and diffused boundaries of tumors,a 3D segmentation algorithm for brain tumor MRI images based on the improved continuous max-flow is proposed in this thesis.The proposed algorithm in this thesis is as follows.Firstly,pre-processed stage,in which three types of MIR images,Flair,T1 C and T2,are pre-processed with the median filtering and fast fuzzy C means clustering.The pre-processed images are linearly fused according to the ratio of Flair:T1C:T2=5:1:4,which is statistically observed from amount of experiments.Then,coarse segmentation stage,when the 3D fused images are clustered using fast fuzzy C-means algorithm to get 3D undersegmented images.Finally,fine segmentation stage,in which the proposed improved continuous max-flow algorithm is acted on the 3D under-segmented images.Then we obtain the final segmentation results with scattering points removed according to the analysis of the structural features and statistical characteristics of the 3D undersegmented images.The tumor images used in the experiments come from Brain tumor image segmentation benchmark 2015(BRATS2015).The BRATS2015 has four registered model images of MRI,Flair、T1、T1C and T2.In this thesis,45 groups of high grade gliomas(HGG)and 5 groups of low grade gliomas(LGG)are randomly selected and 50 groups of three-dimensional brain tumor images are experimented.The experimental results show that the average Dice,Precision and Recall of the proposed method relative to the gold standard are up to 0.90,0.94 and 0.86 respectively.The segmentation time is 0.3 minutes.The 3D segmentation of the target regions can be realized by the improved algorithm to meet the clinical need precisely,instantly and automatically.