Feature extraction using image mining techniques to identify brain tumors

This paper is focused on the comparison of three different intensity based feature extraction method for the abnormal patterns in brain tumors. Physician's interpretation of brain tumors may lead to misclassification sometime. Hence an automated system is needed to solve our problem. The following major categories of brain tumor images are taken into our consideration. They are Metastatic bronchogenic carcinoma, Astrocytoma, Meningioma, sarcoma. The performance factor was evaluated against BRATS (Brain Tumor Segmentation) dataset. For the purpose of calculating and extracting various intensity related features MATLAB tool is used. The experimental results suggest that among the intensity based feature extraction methods GLCM (Gray Level Co-Occurance) method is showing better results than the other methods. WEKA tool classification algorithm J48 also shows close correlation with GLCM Features.