Multi Modal Medical Image Registration: A New Data Driven Approach

Abstract

Image registration is a challenging task in building computer-based diagnostic systems. One type of image modality will not be able to provide all information needed for better diagnostic. Hence data from multiple sources/image modalities should be combined. In this work canonical correlation analysis (CCA) based image registration approach has been proposed. CCA provides the framework to integrate information from multiple sources. In this work, the information contained in both images is used for image registration task. T1-weighted, T2- weighted and FLAIR MRI images has Multimodal registration done on it. The algorithm provided better results when compared with mutual information based image registration approach. The work has been carried out using the 3D rigid registration of CT and MRI images. The work is carried out using the public datasets, and later performance is evaluated with the work carried out by Research scholars previously. Our algorithm performs better with mutual information based image registration. Medical image registration of multimodality images like MRI, MRI-CT, and MRI-CT-PET. In this paper for MRI-CT Medical Image Registration CT image is used as a fixed image and MRI image as moving image and later compared results with some benchmark algorithm presented in literature such as correlation coefficient, correlation ratio, and mutual information and normalized mutual information methods.