An Approach for classification in detecting tumor in Brain MRI images using GMM and Neural Network classifier

Image classification is a process of classifying an image based upon the training given to a classifier. There are various purposes of classification but in this work a Brain MRI image is taken as input and is mainly classified into three class’s malignant tumor and benign tumor and non tumor by using Neural Network classifier. Here a Gaussian Mixture model is used for the purpose of segmentation. First the input MRI image is taken and is pre-processed using weiner filter and it is then segmented into four regions namely the white matter, Gray matter, Cerebrospinal fluid and the high intensity tumor region. After segmentation is done from the tumor region statistical and shape based features are extracted from the tumor region. Based on these extracted features the neural network classifier will be trained to classify a image into tumor affected (Benign or Malignant) or not affected. During the testing phase a unknown image is considered and the pre-processing steps will be applied and then it will be segmented and Features will be extracted by using Gray Level Co-Occurrence matrix (GLCM) and based on the extracted features it will be compared and the Neural network classifier classifies it into either tumor affected or not affected and thus helps in the computer aided diagnosis of Brain MRI images into tumor affected (benign, Malignant) or not affected. Key wordsClassification, Segmentation, Pre-processing, Magnetic Resonance Imaging, Neural network classifier, Gray level Co-occurrence Matrix (GLCM), Gaussian Mixture Model