Textural Analysis of MR Images for Brain Tumor Classification

Automatic recognition system is widely used in the field of medical image processing but its use is challenging one. There are other image modalities such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), etc. which are used for the computer-aided diagnosis system. In the medical field, it is more important to have brain tumor classification before further treatment. Automatic recognition system is more objective than that human beings recognition; it may misclassify the normal and abnormal medical images. In this research work, we proposed a efficient method for brain tumor classification which consisting of four main steps. In the first step we have collected and normalized the five different classes of brain tumors. In the second step we have extracted the LBP and GLCM textural features. Third step belongs to feature selection using principle component analysis (PCA). Then in the final step features are combined and trained the feed-forward neural network using Levenberg-Marquardt back-propagation training algorithm. In the experimental work, it is observed that when only LBP textural features are used we got the recognition rate of 96.00% and when GLCM textural features are used we got 97.00% but by combining LBP and GLCM we got the better recognition rate of 99.00%.