An Efficient Method for Brain Tumor Detection Using Texture Features and SVM Classifier in MR Images

Objective: Detection and classification of abnormalities in Magnetic Resonance (MR) brain images in medical field is very much needed. The proposed brain tumor classification system composed of denoising, feature extraction and classification. Noise is one of the major problems in the medical image and due to that retrieval of useful information from the image is difficult. The proposed method for denoising an image is PURE-LET transform. Methods: This method preserves the diagnostic property of the images. In feature extraction, combination of Modified Multi-Texton Histogram (MMTH) and Multi-Texton Microstructure Descriptor (MTMD) is used and then Gray Level Co-occurrence Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM)are used to extract the feature from the image to compare performance. In classification, classifiers like Support Vector Machine (SVM), K Nearest Neighbors (KNN) and Extreme Learning Machine (ELM)are trained by the extracted features and are used to classify the images. Result: The performance of feature extraction methods with three different classifiers are compared in terms of the performance metrics like sensitivity, specificity, and accuracy. Conclusion: The result shows that the combination of MMTH and MTMD with SVM shows the highest accuracy of 95%.