The goal of the proposed research is to develop retrospective methods that can correct motion artifacts in fMRI to improve capability of functional localization. Subject motion encountered in fMRI results in two main effects. Rigid head motion causes misoriented slices at each repetitive excitation inducing subsequent changes in susceptibility-induced non-linear image distortion. The other source of localized artifacts in slice image is physiological brain motion during cardiac cycle from compression of ventricles and brain stem displacement. The effects of these motion artifacts are three dimensional (3D) non- linear image deformations, which would not be completely removed by rigid-body transform. This study utilizes a retrospective motion correction method to map each slice into volume (MSV), which is applied to multislice single shot Echo-planar image (EPI) data set. Individual slices are spatially repositioned into an anatomical reference volume space to accommodate out-of-plane and in-plane subject motion which include rigid body and thin-plate-spline (TPS) warping. The MSV technique is based on our automated spatial registration system that uses mutual information as a cost function to drive the registration without the need for user-supplied information or preprocessing. Random permutation test is used for significance testing, which offers advantages with respect to the protection against the type I error for dealing with data sets whose true distribution of the test statistics is unknown. The project will focus on (1) extension of the rigid body MSV method to provide an automated and efficient motion correction using 3D warping and to enhance accuracy of statistical signal analysis in fMRI, and (2) evaluation and validation of accuracy of the MSV registration methods. The MSV approach using a rigid body transformation has shown potential for effective motion correction in multislice fMRI time series. The proposed full development of automatic MSV to include warping will provide capability to correct non-linear distortions induced by rigid head motion and local physiological brain motion. The accuracy of the signal detection will be assessed using phantom experiments which provide the ground truth and the efficacy of the method will be continuously evaluated for clinical fMRI data sets throughout the development.
