Comparative study of supervised and unsupervised classification methods: Application to automatic MRI glioma brain tumors segmentation

MRI is a noninvasive neuro-imaging modality largely used in neurology explorations and provides more objective and valuable diagnostic information for High-grade gliomas (HGG). In this context, HGG Segmentation is challenging due to their heterogeneous nature. The present research investigates a comparative study of supervised and unsupervised classification methods for MRI glioma segmentation. These methods are tested with data sets defined in BRATS 2015. We noted that artificial neural networks (ANN) provide efficient segmentation results based on DICE and Jaccard evaluation metrics.