Meical Image Registration For Liver DCE-MRI And Resting-state FMRI

Medical image processing,analysis,and visualization are emerging fields of study that enable quantitative analysis and visualization of medical images of numer-ous modalities such as magnetic resonance image(MRI),computed tomography(CT),positron emission tomography(PET),and ultrasound.They have become significant components in many fields of bio-medical research and clinical practice.Image data sets convey highly detailed information as their dimensions’ increase,which needs to be inter-preted in a timely and accurate manner for analysis.To do this,registration of two images of the same target is essential for many areas in which the corresponding voxels between the two images convey the valuable information.Based on the systematic learning and in-depth understanding of image registration and the understanding and analysis of the limitations of current DCE-MRI and fMRI image registration methods,combined with the characteristics of the image itself,this paper is devoted to the study of anatomical deformation field The medical image analysis provides a more reliable guarantee.To sum up,this paper mainly includes the following aspects:(1)A technical challenge in the registration of dynamic contrast-enhanced magnetic resonance(DCE-MR)imaging in the liver is intensity variations caused by contrast agents.Such variations lead to the failure of the traditional intensity-based registration method.To address this problem,a manifold-based registration framework for liver DCE-MR time series is proposed.We assume that liver DCE-MR time series are located on a low-dimensional manifold and determine intrinsic similarities between frames.Based on the obtained manifold,the large deformation of two dissimilar images can be decomposed into a series of small deformations between adjacent images on the manifold through gradual deformation of each frame to the template image along the geodesic path.Furthermore,manifold construction is important in automating the selection of the template image,which is an approximation of the geodesic mean.Robust principal component analysis is performed to separate motion components from intensity changes induced by contrast agents;the components caused by motion are used to guide registration in eliminating the e ect of contrast enhancement.Visual inspection and quantitative assessment are further performed on clinical dataset registration.Experiments show that the proposed method e ectively reduces movements while preserving the topology of contrast-enhancing structures and provides improved registration performance.(2)Understanding brain function with resting-state functional magnetic resonance imaging(fMRI)relies on accurate inter-subject registration of brain functional regions.Typically,it is performed with the guidance of high-resolution structural images due to their rich spatial details and better tissue contrast.Recently,some studies have found that such strategy cannot have functional regions well aligned across subjects because functional areas are not necessarily consistent with the anatomical landmarks.Registration algorithms that directly use functional information were developed.However,only cortical grey matter(GM)was used to provide functional features for registration,which cannot be accurately achieved because the whole-brain deformation field estimated by features at such a thin,highly curved superficial layer is complex.Inspired by the fact that the functional information reflected by BOLD signals is not constrained in GM but also in white matter(WM)and utilizing a promising technique which could detect local functional connectivity anisotropy,we propose a method which directly and solely uses fMRI data,not only leveraging GM’s but also WM’s functional information,to conduct a functional registration.A novel strategy was used to achieve the aim,comprising of 1)extracting local functional connectivity anisotropy features from separately GM and WM with"Tissue-Specific patch-based Functional Correlation Tensors(ts-pFCTs),and 2)integrating different types of features from the different tissues using"Multi-channel Large Deformation Diffeomorphic Metric Mapping(mLDDMM)".Experimental results show that our method achieves superior functional registration compared with traditional anatomical information-guided registration and even with the state-of-the-art functional registration approaches.