An advanced algorithm combining SVM and ANN classifiers to categorize tumor with position from brain MRI images

Abstract

Brain tumor is such an abnormality of brain tissue that causes brain hemorrhage. Therefore, apposite detections of brain tumor, its size, and position are the foremost condition for the remedy. To obtain better performance in brain tumor and its stages detection as well as its position in MRI images, this research work proposes an advanced hybrid algorithm combining statistical procedures and machine learning based system Support Vector Machine (SVM) and Artificial Neural Network (ANN). This proposal is initiated with the enhancement of the brain MRI images which are obtained from oncology department of University of Maryland Medical Center. An improved version of conventional K-means with Fuzzy C-means algorithm and temper based K-means & modified Fuzzy C-means (TKFCM) clustering are used to segment the MRI images. The value of K in the proposed method is more than the conventional K-means. Automatically updated membership of FCM eradicates the contouring problem in detection of tumor region. The set of statistical features obtained from the segmented images are used to detect and isolate tumor from normal brain MRI images by SVM. There is a second set of region based features extracted from segmented images those are used to classify the tumors into benign and four stages of the malignant tumor by ANN. Besides, the classified tumor images provide a feature like orientation that ensures exact tumor position in brain lobe. The classifying accuracy of the proposed method is up to 97.37% with Bit Error Rate (BER) of 0.0294 within 2 minutes which proves the proposal better than the others.