Immune-based feature selection in rigid medical image registration using supervised neural network

Abstract

Different radiological images like computed tomography (CT) and magnetic resonance (MR) are increasingly being used in medical science research for diagnosisandtreatment.Thisarticle presentsanautomaticimageregistrationtechnique to registerMR-MRimages using gray-level co-occurrence matrix(GLCM) and neural networks. This technique identifies different features of a brain image and its transformational counterpart.GLCM-based image feature extraction is a co-occurrence-based method by which different feature parameters are obtained. These parameters are calculated from the co-occurrence matrix along four directions, namely 0°, 45°, 90°, and 135°. Six features are selected from a set of features using an artificial immune system-based optimized feature selection technique and these six parameters are fed into the proposed neural network. Based on the principle of backpropagation algorithm, transformation parametersbetweenthe referencedandthe sensedimages are estimated. To demonstrate the effectiveness of the proposed method, experiment is carried out on MRT1,T2datasets,andthe results arecomparedwithtwoother existing medical image registration techniques. The proposed method shows convincing results compared to others with respect to the estimation of underlying transformation parameters.