Segmentation and Detection of Brain Tumor in Magnetic Resonance Images

In medical image processing Segmentation of anatomical regions of brain is the fundamental problem. As the brain structure is very complex involving white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) this makes feature extraction of brain images as a basic work. Recently MR images are handled manually for the diagnosis of brain tumor which involves errors and consumed time as due to large variation of the various images indicating varied brain structure. Tumor segmentation from magnetic resonance (MR) images may aid in tumor treatment by tracking the progress of tumor growth and shrinkage. There are a number of techniques to segment an image into homogeneous regions. As the structure of MR image or any medical images is nonhomogeneous and complex, these techniques are not suitable for their analysis. In this report, a new approach for segmentation of MR images has been proposed by incorporating the advantages of the undecimated wavelet transform and Gabor wavelets. The proposed method worked on T1, T2 weighted images to produce an appreciative result though the image is noisy. Undecimated wavelet transform decomposed an image into four sub-bands (LL, LH, HL, HH).