Brain tumor pathology is one of the most common mortality issues considered as an essential priority for health care societies. Accurate diagnosis of the type of disorder is crucial to make a plan for remedy that can minimize the deadly results. The main purpose of segmentation and detection is to make distinction between different regions of the brain. Besides accuracy, these techniques should be implemented quickly. In this paper an automatic method for brain tumor detection in 3D images has been proposed. In the first step, the bias field correction and histogram matching are used for preprocessing of the images. In the next step, the region of interest is identified and separated from the background of the Flair image. Local binary pattern in three orthogonal planes (LBP-TOP) and histogram of orientation gradients (HOG-TOP) are used as the learning features. Since 3D images are used in this research we use the idea of in local binary pattern in three orthogonal planes in order to extend histogram orientation gradients for 3D images. The random forest is then used to segment tumorous regions. We evaluate the performance of our algorithm on glioma images from BRATS 2013. Our experimental results and analyses indicate that our proposed framework is superior in detecting brain tumors in comparison with other techniques.
