Medical image segmentation using fast discrete curvelet transform and classification methods for MRI brain images

Abstract

Listen

Translate article

Image segmentation and classification are indispensable steps in the therapeutic image processing. Clustering based methods plays an inevitable role in the segmentation stage of the biomedical images especially for distinguishingcerebrum tumors from MRI brain images which is a very challenging task. This paper introduces a curvelet transform to extract features for the computerized detection of abnormalities in the MR brain images. Statistical and texture feature of the brain image using MRI are extracted in the curvelet domain. As manual segmentation is time consuming, an automatic support system for segmentation and classification of tumor stages is preferred. In this paper, the variants of Fuzzy CMeans (FCM) clustering segmentation such as Adaptive FCM, AMS modified FCM and other methods like genetic algorithm and snake algorithms are dealt for which the features extracted using the Curvelet transform serves as the input. The segmented image serves as the input for classifiers like Support Vector machine, Probabilistic neural network, and ANFIS which predict the inputs into the two possible classes being normal and abnormal.