Learning to Segment Brain Tumors

摘要

Automatic segmentation of brain tumors is a challenging problem with many inherent difficulties, such as restrictedtraining data, great intra-class variance, and volumetric images with large computational requirementsin terms of processing. To overcome these difficulties, we propose Brain Tumor Parser (BTP), a novel convolutionalneural network that takes advantage of a refinement module and global 3D information to performsemantic segmentation of brain structures in volumetric images with multiple modalities. We draw inspirationfrom recent breakthroughs in edge detection and semantic segmentation in natural images, and we build anaccurate and effective three-dimensional network that segments small structures while refining large instances inmulti-modal Magnetic Resonance Imaging (MRI). We evaluate our approach on the data from the Brain Tumorsegmentation (BraTS) 2017 challenge, obtaining comparable results with the best performing algorithms, whileusing a single yet efficient architecture.