Supervised learning based multimodal MRI brain tumour segmentation using texture features from supervoxels

•Supervoxel segmentation using multimodal MRI to produce boundaries across multiple image protocols.•Unified framework to classify each supervoxel using features calculated from multimodal MRI.•Improved performance for classification of brain tumour supervoxels by using texton descriptors.•Applying DTI with conventional MRI increases the segmentation accuracy for tumour structures.