AUTOMATED SEGMENTATION AND CLASSIFICATION OF MAGNETIC RESONANCE IMAGES FOR BRAIN TUMOR DETECTION

Simran Kaur, Amandeep Singh Monga

Abstract: In this paper an approach for automated segmentation and classification of magnetic resonance images for brain tumor detection has been developed. The input image is first pre-processed to reduce the noise present and then Segmentation is carried out to isolate the tumor region. Different approaches for segmentation are analyzed and compared. Otsu approach is considered for final segmentation. Texture features were extracted from GLCM. Other features computed using corner detection are fast features, Harris Corner Detection. Thus a total of six features are extracted for each suspicious region. A Neural Network is trained using back propagation method to classify the tumor according to its grade type. The proposed method gives reasonable results for the tested images.

Keywords: Classification; MRI Images; Preprocessing; Segmentation; and Tumor

DOI: https://doi.org/10.15623/ijret.2016.0512012