Author:

Keywords:

PSI_MIC, Science & Technology, Technology, Life Sciences & Biomedicine, Engineering, Biomedical, Radiology, Nuclear Medicine & Medical Imaging, Engineering, brain tumor, segmentation, unsupervised, level-set, Expectation Maximization, IMAGES, PSI_4021

Abstract:

© 2016 IEEE. We present a fully-automated MRI brain tumor segmentation method that does not require any manually annotated training data. The method is independent of the scanner or acquisition protocol and is directly applicable to any individual patient image. An Expectation Maximization-approach is used to estimate intensity models for both normal and tumorous tissue. The segmentation is represented by a level-set that is iteratively updated to label voxels as normal or tumorous, based on which intensity model explains the voxels' intensity the best. The method is compared with the method by Menze et al. [1], which is considered to be a benchmark for unsupervised tumor segmentation. The performance of our method for segmenting the tumor volume is summarized by an average Dice score of 0.87 ± 0.06 on the training data set of the MICCAI BraTS Challenge 2012-2013.