Decision Forests for Tissue-specific Segmentation of High-grade Gliomas in Multi-channel MR

We present an automatic method for segmentation of high-grade gliomas and their subregions from multi-channel MR images. Besides segmenting the gross tumor (GT), we also differentiate between active cells (AC), necrotic core (NC), and edema (E), which is important for various clinical applications.

We present a novel discriminative approach based on Decision Forests with context-aware spatial features, and integrate a generative model of tissue appearance. The integration is performed by using the probabilities obtained by a Gaussian mixture model (GMM) as additional input for the forest. Our approach allows simultaneous classification of the individual tissue types, which can simplify the modeling of intensity distributions of each class. The method is computationally efficient and of low model complexity.

The validation is performed on a labeled database of 40 multi-channel MR images, including DTI. We assess the effects of using DTI, and varying the amount of training data. Our segmentation results for the individual tissues are highly accurate, and compare favorably to state of the art results.