Classification of MRI Brain Tumor and Mammogram Images using Adaboost and Learning Vector Quantization Neural Network

Classification and accurate detection of brain tumor using MRI is essential for purpose of treatment and diagnosis of tumor. In this paper we propose and developed system using four stages namely image normalization, Image Binarization with morphological operation, Anisotropic Diffusion filtering and feature extraction using GLCM. The system evaluated on two types of database, Clinical Brain MRI Images and Digital Database for Screening Mammogram (DDSM). Normalization is process of contrast stretching which changes value of pixel intensity and Image Binarization is processing of Grey scale image into black and white image by fixing threshold level of pixel. If value of pixel above the threshold level is white either Black followed by steps of morphological operation i.e. Erosion and Dilation by processing MRI images. Apart from that anisotropic diffusion (ADF) is applied for detection and sharpen the edge detection. Features taken or extracted by using GLCM from filtered MR images. In the stage of classification, two Neural Networks have been implemented. The first Neural Network is Adaboost NN is based on boosting method which yields classification accurately and the second neural network, LVQ is feed forward network which uses Quantization machine learning algorithm and Lossy compression techniques. The extracted features hence given to train Neural Network for classification. Accuracy with success has been obtain 95% and 80.6% for Clinical Brain MRI images with 79.3% and 69.9% for DDSM.