Diffusion tensor imaging distortion correction with T 1

Introduction Diffusion weighted single shot spin-echo planer imaging (DW-EPI) is commonly used for diffusion tensor imaging (DTI). DW-EPI is very sensitive to static magnetic field (B0) inhomogeneities (such as that near air-tissue boundary) that produce geometric distortion, primarily along the phase-encoding direction. As a result, artifacts are severe in the frontal and temporal lobes due to the sinuses. These artifacts degrade the ability of using diffusion measures (fractional anisotropy (FA), mean diffusivity (MD)) and diffusion tractography to compare patient and control population. Several different techniques (field map, phase reversal, and image based linear and non-linear registration with T2) were suggested for correcting susceptibility distortion. However, to apply these correction techniques, additional images (field map, T2) are required. This is undesirable, especially when scan time is limited. In this work, we examine susceptibility distortion correction using image based nonlinear registration along with inverting the intensity of the T1 image. This method exhibits better performance for geometrical distortion correction in frontal region and an ability to analyze clinical diffusion data without additional collection of images. Methods A. Subjects and image acquisition: Twelve healthy subjects (mean age: 36.5±10.7, gender (F/M: 7/5) participated in accordance with Institutional Review Board policies at Emory University. Data were acquired on a 3T Siemens TimTrio scanner using a TX/RX birdcage head coil. DTI data were collected using a DW-EPI sequence with the following parameters: b=1000 sec/mm2; voxel resolution=2×2×2 mm; number of slices=64; matrix=128*128; 64 directions with 2 averages for each direction; TR/TE=9800/95ms. T1 images were collected in 128 sagittal slices using an MPRAGE sequence with following parameters: TR/TE=2600/3ms; voxel resolution=1×1×1 mm; number of slices=176; matrix=224×256. In addition to T1 images, T2 image was acquired from one subject for method comparison. B. Preprocessing of T1, T2, and DTI: FSL (http://www.fmrib.ox.ac.uk/fsl) and TrackVis (http://trackvis.org) were used. T1 and T2 data were skull stripped to remove non-brain regions. The intensity of T1 data was inverted (i.e., the voxel with minimum intensity converted to maximum intensity and vice versa). Diffusion data underwent brain extraction, eddy current correction, and local DTI fitting to generate FA images. C. Susceptibility distortion correction and validation: Inverse intensity T1 (and T2) images were pre-registered to the b0 image using rigid body registration (FLIRT: degree of freedom 7). These pre-registered images were then used as a reference image for distortion correction using a non-linear registration. The b0 image was then registered to the inverse intensity T1 and T2 image using large diffeomorphic registration method (FNIRT). In addition to correction with the T2 image for one subject, diffusion data of all twelve subjects were registered to the individual T1 image using normalized mutual information linear registration. Comparison to currently available image based distortion correction methods (non-linear with T2 and linear using mutual information) to non-linear registration method using the inverse intensity T1 was performed. To validate the results, the derived FA map (non-corrected, linear corrected, and non-linear corrected) of each individual subject was normalized to the MNI template using a combination of linear (diffusion to T1 data) and non-linear (T1 to MNI template) registration, and then the FA variance map of each correction method was calculated. The effects of distortion correction were also evaluated on deterministic fiber tractography. More specifically, TrackVis software was used to analyze tracks in the frontal lobe, one of the regions that suffer from severe susceptibility artifacts.