Deep Learning for Automated Brain Tumor Segmentation in MRI Images [chapter]

Rupal R. Agravat, Mehul S. Raval
2018 Soft Computing Based Medical Image Analysis  
In this paper a simple strategy for the automatic segmentation of tissues in magnetic resonance images of multispectral classification based mainly on minimum Euclidean distance is presented From a set of 3D images in the T1, T1 modalities with gadolinium contrast, T2 and FLAIR and its segmentation reference descriptors for each tissue type are obtained through the centroid of each class, which are used to classify new input images. Magnetic Resonance Imaging (MRI) offers the possibility of
more » ... imodality image acquisition, different imaging studies can be performed on the patient in the same slice position with little movement artifacts so images are acquired virtually registered. Multimodality MRI studies are usually performed to diagnose and characterize brain tumors and typically a set of 4 or more modalities are acquired. In the present work, a method based on multidimensional mathematical morphology is used to classify brain tissues for multimodality MRI comprising 4 modalities, allowing for tumor image segmentation and characterization. The method also proposes a general view for image integration or fusion that allows for a targeted application of therapy.
doi:10.1016/b978-0-12-813087-2.00010-5 fatcat:l4mlyo7635cqtpgrdhsor6z4ty