COMPUTER AIDED DIAGNOSTIC SCHEME FOR BRAIN TUMOR DETECTION IN MR-IMAGES USING WAVELETS & NEURAL NETWORKS.

Magnetic Resonance Imaging (MRI) is commonly used for detecting brain diseases because they are the most cost-effective, routinely available, and effective diagnostic tool. Brain magnetic resonance (MR) segmentation algorithms are commonly used to analyze tissues and diagnose tumor present in MR-Images. The main objective is to develop a Computer Aided Detection (CAD) system for improving neurologist’s efficiency in the detection of tumor using image processing techniques & neural network classifier. Detection of tumor is very important for treatment planning.T1- weighted MR-Images are used as input to the CAD system. The first step is pre-processing, to reduce the noise & improve the quality of images. To preserve the edges & smoothing the interior parts, anisotropic diffusion filter is used for pre-processing. Normalize the range of image intensities to 0 & 1.Next step is skull stripping also called as whole brain segmentation, to remove the non-cerebral tissues such as skin, skull, muscles, fat etc. Input feature vector is obtained from stationary wavelet coefficients. In-order to increase the segmentation accuracy, extract additional features like shape, texture & statistical features from the input images. Finally the features are extracted and neural network classification & SVM classification is used to classify the tumor classes.