Design and development of 3d brain MRI image analysis system using soft computing techniques

Abstract

The common method for differential diagnostics of tumor type newlineis Magnetic Resonance Imaging (MRI). However, it is susceptible to newlinehuman subjectivity, and large amounts of data are difficult for human newlineobservation. In order to obtain precise diagnostics, and to avoid surgery, newlineIt is important to develop an effective diagnostics tool for tumor newlinesegmentation and classification from MRI images. In this research work, newlineMRI brain tumor recognition system image has been developed with newlinefeature extraction and without feature extraction. In this research study, newlinethe available brain tumor public datasets in figshare are used to analyze newlineand evaluate the types of brain tumor architecture. The MRI dataset newlinecontains 3064 MRI slices of brain images which acquired from 234 newlinepatients. newlineFirst method, MRI brain tumor images has applied with feature newlineextraction and machine learning algorithms. The numbers of experiments newlineare conducted in the 5-fold, 10-fold, 15-fold, and 20-fold cross validation newlinetechniques. Our proposed method RBF based SVM has achieved newlinemaximum accuracy of 94%, sensitivity 95% and specificity 90%. newlineWhile comparing with other classifiers, it was reflected that linear support newlinevector machine and feed forward neural network performs relatively newlinebetter in terms of classification accuracy, sensitivity and specificity. From newlinethe results, it can be concluded that MRI brain tumor recognition system newlineusing RBF based SVM suits as best classifier. newline