Brain tumor segmentation and classification using Deep Belief Network

Brain image segmentation and classification is the significant area of research to differentiate the tumor region from the non-tumor region, for which the segmentation is an effective step that assures the effective classification. The n9eed for the accurate classification is initiated with the extract9ion of the relevant features, for which the segmentation p9lays a major role. In this paper, the segmentation is progress9ed using the Probabilistic Fuzzy C-means algorithm that distinguishes the significant regions from the MRI brain image and offers a platform to reduce the dimensional reduction. The segments are further processed using the Local Directional pattern (LDP), for extracting the texture features of the significant regions from the segmentation method. Then, the features are presented to the Deep Belief Network (DBN) classifier that classifies the images as normal or abnormal indicating the presence or absence of tumors in MRI. Experimentation is performed using BRATS database and the proposed method is analyzed based on accuracy that acquired the greater percentage of 95.78%.