Structural and functional biomarkers of PTSD

Structural and functional biomarkers of PTSD

Specific Aims

In order to address a critical gap in diagnosing PTSD, we are currently performing a 3T imaging study of Iraq and Afghanistan Veterans. We will enroll 140 cases with combat related PTSD and 140 combat exposed controls matched for age, gender and ethnicity. Participants are assessed with the Structured Interview for DSM IV,TR, Clinician Administered PTSD Scale and neurocognitive testing for attention, concentration, intelligence, learning, memory. Imaging is performed at the NYUSOM Center for Biomedical Imaging in collaboration with Dr. Sodickson and Dr. Mariana Lazar. In addition to imaging we ascertain genetic, epigenetic, metabolic, neuroendocrine, and proteomic markers. Structural and DTI image processing will be conducted by our collaborator Dr. Michael Weiner at UCSF and SF VAMC. DKI and resting state image processing will be conducted at NYU. Specific aims of the imaging component of the study are as follows:

  1. Compare hippocampal subfield volumes in subjects with PTSD versus controls. Hypothesis: PTSD selectively impacts the dentate/CA3 hippocampal subfield. It is predicted that the PTSD group will have smaller dentate gyrus/CA3 hippocampal subfield volumes than controls.
  2. Compare total hippocampal volumes in subjects with PTSD versus controls. Hypothesis: PTSD is associated with reduced anterior and posterior total hippocampal volumes, and medial prefrontal cortex volume.
  3. Exploratory aim: Explore new imaging metrics including additional subfield volumes, diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), and resting state fMRI as candidate diagnostic markers, and assess correlations with non-imaging-related biomarkers.

Significance

It is estimated that 10% to 20% of warfighters who have served in Iraq and Afghanistan have PTSD (1). An important limitation of these estimates is the reliance on self-report screening measures and clinical interviews to make the diagnosis of PTSD. These methods are subject to a number of biases, including underreporting of PTSD symptoms because of stigma of mental illness and concerns about adverse effects on careers, and exaggeration of symptoms in those seeking compensation for service-connected disability.

Innovation

Development of imaging biomarkers of PTSD is critical as objective indicators of PTSD for use in post-deployment medical screening, treatment selection, treatment outcome monitoring, disability evaluations, and for informing novel targets for treatment development. Additionally, biomarkers hold great potential for explaining and mitigating the associations between war zone-related PTSD and physical health problems, including cardiovascular and metabolic disorders. Imaging markers also hold promise for differentiating PTSD from TBI.

Approach

Image Acquisition Methods
Structural MRI – including hippocampal subfields: We are acquiring the following structural MRI sequences for quantitative brain volume measurements, including hippocampal subfields, as well as for image registration, spatial normalization and anatomic parcellation of the brain:

T1-weighted MRAGE with 1 x 1 x 1mm3 resolution, optimized for high gray-to-white matter contrast, T 2- weighted fast spin echo sequence (TR/TE: 3500/19 ms) with submillimeter resolution (0.5 x 0.5 mm2 inplane) for tracing hippocampal hippocampal subfields and T2-weighted, turbo spin-echo (TSE) sequence with TR/TE = 4000/30ms and the same resolution matrix and FOV than MPRAGE, providing high contrast between CSFand brain boundaries

Diffusion tensor and kurtosis imaging: Diffusion data are acquired using a twice refocused-spin-echo EPI sequence, which has been shown to significantly reduce the eddy-current-related distortions in the diffusion weighted images. Imaging parameters include: TR = 6700ms, TE = 97ms and a field of view of 220 x 220 mm2 with an acquisition matrix of 100 x 100. We employ parallel imaging with an acceleration factor of 2 to reduce image distortions. We acquire 50 contiguous slices with a slice thickness of 2.2 mm (voxel size=2.2x2.2x2.2 mm2) which give nearly full brain coverage. The DTI sequence is augmented by applying 64 uniformly distributed diffusion encoding and for two b values (b = 1000 s/mm2 and b=2000 s/mm2). We also acquire 10 sets of images with b=0 s/mm2.

Resting Slate fMRI: Amplitude of low frequency fluctuation (ALFF), which reflects the magnitude of spontaneous brain activity is obtained in a voxelwise fashion. ALFF maps of each subject are then normalized to the Talairach and Tournoux space for each subject using a non-linear transformation of the individual anatomical high-resolution images. Functional connectivity (FC), which reflects the temporal synchronization of the BOLD signal, is calculated following well established protocols (2,3). Average pre-processed BOLD signal is obtained from pre-defined regions of interest (ROIs). FC is calculated on a voxelwise basis, after which Fisher transformation is conducted to obtain a Z score on each voxel. Z maps is then be normalized to the Talairach and Tournoux space for each subject using a non-linear transformation of the individual anatomical high-resolution images.

Image Processing and Analysis Methods
Structural MRI:
Using anatomical landmarks, the manual procedure traces the CA1 region, CA2, a combination of CA3 and the dentate gyrus, the subiculum and the entorinal cortex Our primary analysis (1) characterizes differences in CA3 and Dentate Gyrus subfield volume between PTSD and controls, (2) relates subfield volume to severity of PTSD, (3) calculates area under the ROC curved to estimate sensitivity and specificity of CA3, other hippocampal subfield and exploratory ROI volumes for classifying cases and controls, and (4) identifies hippocampal subfield volumes as classifiers of PTSD in conjunction with, or independently of, other imaging modalities and blood and urine markers.

Diffusion Tensor Imaging: Our primary statistical analysis (1) characterizes the regional pattern of DTI-derived measures of fiber integrity (2) relates DTI changes to measures of PTSD; (3) identifies DTI classifiers of PTSD in conjunction with, or independently of, other imaging modalities.

Resting State fMRI: Amplitude of alternating low frequency fluctuation (ALFF), which reflects the magnitude of spontaneous brain activity is obtained in a voxelwise fashion. ALFF maps of each subject are then normalized to the Talairach and Tournoux space for each subject using a non-linear transformation of the individual anatomical high-resolution images. Functional connectivity (FC), which reflects the temporal synchronization of the BOLD signal, is calculated following well established protocols (2,3).

Data Analysis: The primary data analytic methods for testing the specific and exploratory aims are simple between group ANOVA’s to test differences between PTSD +/- groups on candidate imaging markers. Pearson correlations testing relationships between markers and area under the ROC curve are utilized for determining accuracy of classification of cases and controls. Pearson correlations and Spearman Rank correlations for non-parametric also test relationships between biomarkers and PTSD symptom severity in the PTSD positive group. Candidate biomarkers are combined to assess their joint ability to discriminate PTSD cases from non-cases. This is accomplished by entering them as predictors in a multiple logistic regression model predicting PTSD +/- group membership.

Preliminary findings: To date we have imaged 48 PTSD positive cases and 72 PTSD negative controls with current enrollment rate at 2 new cases imaged a week with the following preliminary findings:

  1. We performed manual tracing for hippocampal subfield volumes in 33 PTSD cases and 34 combat exposed controls matched for age gender and ethnicity. We found reduced CA1 subfield volume in cases 381.9 (42.4) vs 398.6 (37.9)*, controlling for childhood trauma exposure, history of loss of consciousness and current depression. Contrary to our prediction we did not find differences for CA3 & DG or other subfield volumes.
  2. Utilizing FS V 5.1 we performed whole brain parcellation in 37 cases and 51 combat exposed controls matched for age, gender and ethnicity. We found reduced rostral cingulate volume in cases 5.57 (0.42) vs. 5.72 (0.30) *, reduced caudal cingulate volume 4.99 (0.39) vs 5.11 (0.41) ** and reduced insula volume in cases 6.08 (0.33) vs. 6.19 (0.24) *, controlling for childhood trauma exposure, history of loss of consciousness and current depression (4).
  3. In resting state analyses with 40 cases and 40 controls matched for age, gender and ethnicity we found reduced ALFF at the thalamus in cases and selecting thalamus as a seed decreased functional connectivity between the thalamus and the superior occipital cortex, the precuneus and the insula (5).

Key Personnel: 
Charles Marmar, MD

Sponsors

Latest Updates

Philanthropic Support

We gratefully acknowledge generous support for radiology research at NYU Langone Medical Center from:
 
• The Big George Foundation
• Raymond and Beverly Sackler
• Bernard and Irene Schwartz

Go to top