Texture-Based brain tumor segmentation in MR images
Borja Rodríguez-Gálvez · RECERCAT (Consorci de Serveis Universitaris de Catalunya) · 2017
In this thesis, we explored the idea of texture as a tumor segmentation approach in MR images. To deal with this, we extended the definition of the texture descriptors provided by Kovalev et al., co-occurrence matrices [1]. Then, we defined a rotation-robust feature space derived from the natural feature space that our definition of co-occurence matrices defined. We used these features to train a random forest classifier [2] and see if tumors could be detected from MR images, and if the different tissues within a tumor could be distinguished.