GLISTRboost: Boosted GLioma Image SegmenTation and Registration

GLISTRboost describes our proposed method for segmenting low- and high-grade gliomas in multimodal magnetic resonance imaging (MRI) volumes. Note that this is the winning method of the Multimodal Brain Tumor Image Segmentation (BRATS) Challenge, held in conjunction with the Medical Image Computing and Computer Assisted Intervention (MICCAI) conference in Technische Universitaet Muenchen (TUM) in Munich (Germany) - October 2015.

The proposed approach is based on a hybrid generative-discriminative model. Firstly, a generative approach of a joint segmentation-registration scheme based on an Expectation-Maximization framework, that incorporates a glioma growth model, is used to segment the brain scans into tumor and healthy tissue labels (i.e. GLISTR). Secondly, a discriminative, gradient boosting multi-class classification, scheme is used to refine tumor labels based on information from multiple patients. Lastly, a probabilistic Bayesian strategy is employed to further refine and finalize the tumor segmentation based on patient-specific intensity statistics from the multiple modalities.

Assumptions: All input MRI volumes (i.e. T1, T1-Gad, T2, T2-FLAIR) must be skull-stripped, co-registered and in the LPS coordinate system, for the method to produce meaningful results.

Users should use the visual interface, Cancer and Phenomics Toolkit (CaPTk), to easily make the initializations required by GLISTRboost.