Allostatic Load, Genetics and Neurobiology in PTSD: Identifying Novel Biomarkers of Combat-Related PTSD

Methods for High Resolution Brain MRI

One overall goal of the Neuroimaging Core is first to achieve homogeneous acquisition of structural as well as perfusion and diffusion tensor images at high field at the James J. Peters VA Medical Center (JJPVAMC) in a fashion that harmonizes with the project ongoing in San Francisco. The second overall goal is image processing on the 40 pilot subjects (20 PTSD + and 20 PTSD -). The Neuroimaging Core will perform automated image processing of structural MRI including intensity normalization, segmentation of T-1 and T-2 weighted images and FLAIR, spatial normalization, and automated analysis of brain structure using Freesurfer software; image processing of the diffusion tensor images and high resolution hippocampal subfield MRIs, as well as analysis and hypothesis testing. Dr Weiner’s Neuroimaging Core will: provide an overall cohesion to the collection of imaging data, create a data repository for program wide access to shared information, support program wide methodological development, and will ensure uniform methods of acquisition, so that data collected at JJPVAMC will be analyzable by the proprietary and well-validated methods in his lab.

The PI of the Imaging Core and his group have considerable experience performing multi-site studies, including the Imaging Core of a Program Project Grant for more than a decade and the Alzheimer’s Disease Neuroimaging Initiative (ADNI). All data will continue to be organized and archived in a relational database in the Core (which is mirrored to the main database of the project) so that it is easily available to the other investigators in the project. This Neuroimaging project will perform the following tasks:

  • Establish harmonized acquisition of structural MRI (T1 and FLAIR), diffusion tensor MRI, and high resolution hippocampal MRI at James J. Peters VA Medical Center similar to the acquisition the VA Medical Center/ UC San Francisco.
  • Establish QA procedures and insuring continued quality control of MRI acquisition.
  • Archive and manage all MRI data acquired.
  • Perform intensity corrections (bias field corrections).
  • Image segmentation of structural MRI (GM, WM, CSF, white matter lesions (WMLs), using Expectation maximization likelihood software.
  • Perform spatial normalization of all data.
  • Perform Freesurfer analysis including quantification of cortical thickness measurements and parcellation of brain regions including total hippocampal volume.
  • Ensure rater reliability and continue rater testing for all manual steps in the above.
  • Maintain an image repository of raw and processed data and data base of all results. It is expected that as the project evolves, new hypotheses will be generated. These will be readily testable with the current data archiving structure.
  • Provide raw and processed image data to members of the project on a timely basis.
  • Provide raw and processed images and data to outside qualified investigators who request data, consistent with the data sharing plan described elsewhere.
  • Continue previous work in developing and implementing completely automated methods for: intensity inhomogeneity correction, tissue segmentation, voluming of hippocampus and major lobes. This development effort will utilize brain atlases and nonlinear transformation techniques.

We propose a harmonized structural (3D T-1 weighted MPRAGE using the ADNI sequences) and isotropic 3D FLAIR, DTI and high resolution MRI of hippocampus. The MRI sessions will include structural, diffusion (DTI), and perfusion MRI, totaling no more than one hour scan time. Our a priori hypotheses focus on the volume of the dentate/CA3 subfield, but other imaging modalities will be collected for exploratory analysis for examination of additional possible biomarkers. Our experience of many years is that that up to 1 hour scan time is usually tolerated by more than 90% of elderly and cognitively impaired subjects. Head padding will be used to reduce movement. Scanning will begin with a three-planar scout scan for head positioning and for accurate alignment of the imaging field of view to guarantee fully brain coverage. Three structural MRI sequences, 1x1x1 mm3, will be obtained for image registration, spatial normalization, brain parcellation, tissue segmentation, and quantitative volume measurements.

For DTI, we will use a multislice single-shot EPI sequence with TR/TE = 6000/90ms, base matrix size 112 x 112 for 340 x 340 mm2 FOV, yielding 3 x 3 mm2 resolution, 40 contiguous slices, 3 mm each, with large FOV to avoid aliasing of Nyquist ghosting from EPI, zero-padding to 128 x128 upon Fourier reconstruction. Before clinical studies begin, image resolution, SNR, and contrast will be harmonized between James J. Peters VA Medical Center and VA/UCSF using initial phantoms and eventually 3 traveling volunteers, who will be scanned back to back at each site.

The Freesurfer Software package will be used for cortical thickness measurements regions of interest analysis. Before importing the T1 images into the Freesurfer package, bias correction and skull stripping is performed by applying the bias map and tissue map derived from segmentation in EMS. The images are then imported into Freesurfer.

The first step of the Freesurfer processing is modified to suppress skull stripping and bias correction done by Freesurfer during this step. After a visual check of the output, the second step of FreeSurfer is started which produces cortical thickness measurements and labeling of the subcortical structures. Finally, the last part of Freesurfer is run, which produces a parcellation of different brain structures. The parcellations of Freesurfer are exported in other display software like Review ( and manipulated to derive hippocampal volumes or regions of interest (ROI) for structural (hippocampal volume or cortical and subcortical gray matter volumes for volumetric measurements) and functional measurements (perfusion) for verification of voxel-based methods and for analysis of images not suitable for voxel-based analysis due to cortical defects.