METHOD FOR BRAIN TUMOR SEGMENTATION IN MULTIMODAL THREE-DIMENSIONAL MAGNETIC RESONANCE IMAGES BASED ON ROBUST STATISTICAL INFORMATION PROPAGATION

摘要

A method for automatically segmenting brain tumor in a multimodality three-dimensional magnetic resonance image includes following steps: pre-processing the input multimodality three-dimensional magnetic resonance image; classifying each voxel of the pre-processed image and calculating the probability of each voxel belonging to the brain tumor, and determining initial label information of image segmenting depending on the classification probability; establishing image representation of the pre-processed image; seeking segmenting result of tumor region depending on the initial label information and the image representation. A method for automatic brain tumor segmentation in multimodal three-dimensional magnetic resonance images includes the following steps: pre-processing the input multimodal three-dimensional magnetic resonance images; classifying each voxel of the pre-processed images, calculating the probability of each voxel belonging to the brain tumor, and determining initial label information of image segmentation according to the class probability; establishing image representation of the pre-processed images; and attaining segmentation results of tumor regions according to the initial label information and the image representation.