Brain tumor classification based on GLCM and moment invariant

Abstract

Brain tumor is a mass of tissue that grows inside the skull in random directions, attacking and destroying nerves or affecting any tissue. The lack of homogeneity in the brain tissue leads to inaccuracies in the identification and allocation of the tumor. In this research, it work on the images extracted from the MRI is performed in two stages, the first stage is the detection and extraction of the tumor. The second stage is the process of diagnosis of the tumor, which is performed by extraction tumor features and classifying them. This process passing through several steps, smoothing is an important step, which was used to remove the bones of the skull, which is an obstacle in the process of detection of tumor, density slicing and segmentation depending on YCbCr color transformation and then determine the tumor area accurately and remove noise and standardize image size. After detection of the tumors, the features of these tumors were extracted by extracting the texture features and moment invariants, where these features are input to the classification algorithm, which is the support vector machine classifier. The classification done between two types of brain tumors are glioma and meningioma tumors, 153 images (59 meningioma and 94 pictures of glioma tumors) were used. The brain tumor was diagnosed and the results obtained were complied with the classification by the specialist doctor manually. The accuracy of the classification in this work is 93.44%.