Brain Tumor Detection by Using Moments and Transforms on Segemented Magnetic Resonance Images

Abstract

In this paper, we propose a novel approach to detect tumor in Magnetic Resonance (MR) brain images. The feature set is extracted by using 2D Continuous Wavelet Transform (2D-CWT) and segmentation is done using Improved Incremental Self Organize Mapping (I2SOM). Symmetry in the MR image is analyzed by using Zernike Moments (ZMs) or Polar Harmonic Transform (PHTs). The region of tumor is extracted by using PHTs. The effectiveness of proposed method is analyzed by experiments on 40 normal and noisy brain images. It is observed that tumor detection is successfully realized for the tumorous 20 MR brain images.