Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jul 30.
Published in final edited form as: Psychiatry Res Neuroimaging. 2021 Apr 24;313:111297. doi: 10.1016/j.pscychresns.2021.111297

Gray Matter Volume Correlates of Adolescent Posttraumatic Stress Disorder: A Comparison of Manual Intervention and Automated Segmentation in FreeSurfer

Marisa C Ross a,b,*, Delaney Dvorak a, Anneliis Sartin-Tarm c, Chloe Botsford a, Ian Cogswell d, Ashley Hoffstetter a, Olivia Putnam e, Chloe Schomaker a, Penda Smith a, Anna Stalsberg f, Yunling Wang a, Megan Xiong a, Josh M Cisler a
PMCID: PMC8205994  NIHMSID: NIHMS1700526  PMID: 33962164

Abstract

Exposure to early life trauma is common and confers risk for psychological disorders in adolescence, including posttraumatic stress disorder (PTSD). Trauma exposure and PTSD are also consistently linked to alterations in gray matter volume (GMV). Despite the quantity of structural neuroimaging research in trauma-exposed populations, little consensus exists amongst research groups on best practices for image processing method and manual editing procedures. The purpose of this report is to evaluate the utility of manual editing of magnetic resonance (MR) images for detecting PTSD-related group differences in GMV. Here, T1-weighted MR images from adolescent girls aged 11–17 were obtained and analyzed. Two datasets were created from the FreeSurfer reconall pipeline, one of which was manually edited by trained research assistants. Gray matter regions of interest were selected and total volume estimates were entered into linear mixed effects models with method (manual edits or automated) as a within-subjects factor and group dummy-coded with PTSD as the reference group. Consistent with prior literature, individuals with PTSD demonstrated reduced GMV of the amygdala compared to trauma-exposed and non-trauma exposed controls, independent of editing method. Our results demonstrate that amygdala GMV reductions in PTSD are robust to certain methodological choices and do not suggest a benefit to the time-intensive manual editing pipeline in FreeSurfer for quantifying PTSD-related GMV.

1. INTRODUCTION

Exposure to early life trauma is common, with some studies reporting incidence of trauma before age 16 as high as 66% (Copeland et al., 2007). Trauma exposure can lead to lasting consequences such as posttraumatic stress disorder (PTSD), with one survey suggesting up to 8% of adolescent girls meet full clinical criteria, with an upward trend as age increases (Merikangas et al., 2010). The psychological burden of PTSD involves difficulties in emotional regulation and cognitive control, with studies showing alterations in memory and fear extinction, which can be especially debilitating in adolescent populations (Hayes et al., 2012). Given the burden this disorder can cause, it is important to understand the pathology of PTSD to create effective, targeted treatments.

One popular method of investigating neural differences in PTSD is examining gray matter volume (GMV), which is a hybrid measure of surface area and cortical thickness (Fischl and Dale, 2000). Structural neuroimaging research suggests broad variations in GMV in many adolescent psychiatric disorders including anxiety (Mueller et al., 2013) and bipolar disorder (Gold et al., 2016) as well as PTSD (Ahmed et al., 2012; Mutluer et al., 2018). Many studies on childhood maltreatment implicate GMV changes within the amygdala, hippocampus, and temporal lobe (Kribakaran et al., 2019) as well as within the anterior cingulate cortex (ACC; Hart and Rubia, 2012). GMV could be a particularly useful measure in adolescent populations, as there are well documented developmental decreases in GMV (Gennatas et al., 2017), leading to hypotheses about trauma as a developmental interrupter leading to lasting structural changes in GMV. Several studies have investigated this hypothesis and have supported smaller ventral medial prefrontal cortex (vmPFC) GMV that is unique to adolescent PTSD compared to trauma or maltreatment alone (Heyn et al., 2019; Morey et al., 2016). However, there has not been consensus in the specific GMV alterations that occur in various disorders, including PTSD. One possible explanation for discrepant results might include a lack of standardization in structural image processing methods.

Despite the quantity of structural neuroimaging research, there is little consensus between research groups on best practices for image processing method and manual editing procedures, leading to potentially inconsistent results. There are several structural processing programs that can be used, including FreeSurfer (http://surfer.nmr.mgh.harvard.edu/), Statistical Parametric Mapping (SPM; Penny et al., 2011), FMRIB Software Library (FSL; Smith et al., 2004), and in-house programs used by individual labs, each allowing for a variety of processing method choices. FreeSurfer is a commonly-used structural image processing pipeline that includes image preprocessing and estimations of regional and whole-brain GMV, cortical thickness, and surface area. FreeSurfer contains a default automated pipeline, which can be customized to fit the needs of each research group, and creates an automatically processed data set which can be used for analysis without researcher intervention. However, FreeSurfer also allows for research groups to individually create protocols for manual editing of the produced data set. The use and reporting of manual edits can vary from simple visual inspection to check for motion artifacts to any combination of manual skull stripping, adjusting white matter boundaries, and editing pial surface boundaries (Figure 1). Although FreeSurfer provides guidance for manual interventions, standardized error correction procedures are lacking, and groups can choose to forgo manual edits in favor of the fully-automated processing pipeline. A meta-analysis of 82 neuroimaging studies that utilized FreeSurfer found that 54% of studies that used a 1.5T scanner and 69% of studies that used a 3T scanner incorporated manual edits in some way (McCarthy et al., 2015). Given the option of performing manual edits and the lack of consistency in methods, image editing process may be an important source of variance in neuroimaging research that must be addressed to give consistency and support to neuroimaging findings.

Figure 1: Available Pipelines for Automated and Manually-Edited Datasets in FreeSurfer.

Figure 1:

Open in a new tab

Raw T1 images are processed using FreeSurfer’s reconall preprocessing, parcellation, and segmentation pipeline after visual inspection for image artifacts. Following initial reconall, trained editors can manually intervene to correct errors in white matter classification and pial surface boundaries, resulting in a manually-edited dataset. Alternatively, researchers can refrain from intervention and instead utilize the automated dataset.

Although manual editing could provide additional accuracy and control, it is a time-intensive process that requires highly trained editors and is susceptible to human error. In fact, recent evidence in pediatric (Beelen et al., 2020) and adult (McCarthy et al., 2015) samples suggests that manual intervention may slightly improve the validity of FreeSurfer segmentations but that improvement does not relate to neural differences between clinical groups. As structural neuroimaging moves toward larger sample sizes, the burden of performing manual inspection and correction potentially becomes impractical. One potential method to combat this issue uses a statistical approach to screen for potential outliers, thus decreasing the number of images that undergo manual editing and the labor required to edit a larger sample (Waters et al., 2019). One study compared statistically-driven manual edits on outliers of the mean, automatically targeting images to be manually edited, to manual editing of the entire sample (Waters et al., 2019). Comparison of these methods revealed that statistical outliers were 1.69 times as likely to have at least one error as compared to non-outliers, but the error rate was still 40.9% in the non-outlier group, demonstrating that the statistical outlier method of manual editing was ineffective. Given the ineffectiveness of this method and lack of well supported alternatives, researchers utilizing structural imaging data generally choose either the fully automated option with no intervention or the time-intensive process of manual inspection of the entire sample with peer and supervisor checks to ensure interrater reliability.

The goals of the following analysis are twofold: first, we aim to replicate existing literature that supports trauma- and PTSD-related alterations in GMV of regions of interest (ROIs) within a relatively large sample of adolescent females with trauma exposure. Second, we aim to assess the impact of manual intervention to correct errors in FreeSurfer-generated images on group-level differences in GMV of ROIs that are commonly detected as altered in trauma-exposed individuals. To achieve this aim, a manually edited dataset was created by trained research assistants in a multistep, peer checked process to correct for errors in classification of white matter, grey matter, and pial surface boundaries. The same set of raw images were also entered into the automated FreeSurfer pipeline without manual edits. GMV estimations of ROIs were compared between editing methods to determine if choice of editing method impacts detection of trauma- and PTSD-related differences in GMV. Discrepancies between methods may help explain variation in prior GMV differences found in adolescent PTSD and assess whether GMV alterations associated with PTSD are robust to certain methodological choices. Ultimately, this investigation may also contribute to a set of processing guidelines for databases of adolescent structural data by helping standardize procedures and reduce sources of error.

2. METHODS

2.1. Participant Recruitment.

Adolescent females aged 11–17 were recruited through four previously-reported studies at two sites: the University of Arkansas for Medical Sciences (UAMS; Little Rock, AR; n = 103) and the University of Wisconsin, Madison (UW; Madison, WI; n = 36). All studies were approved by the respective Institutional Review Boards, and details from these studies are reported elsewhere (Cisler et al., 2018b, 2018a, 2015, 2013; Sellnow et al., 2020). The present analyses do not repeat prior published reports from these studies. Two of the studies (n=98) recruited both a non-trauma exposed healthy adolescent sample and a sample of adolescents with interpersonal violence (IPV) exposure. Inclusion criteria for healthy comparisons included female sex, as well as absence of trauma exposure, current psychiatric disorders, and psychiatric medication use. Inclusion criteria for the IPV-exposed sample was female sex and a positive history of IPV-exposure; categorization as IPV-exposed was not dependent on the type or severity of psychiatric symptoms. The third study recruited specifically for IPV-exposed, treatment-seeking adolescents with a current diagnosis of posttraumatic stress disorder (PTSD, n = 31). The fourth study recruited only non-trauma exposed healthy adolescents, with inclusion criteria of female sex, absence of trauma exposure and current psychiatric medication use or psychiatric disorders (n=10). For each study, participant’s legal guardians provided informed consent and all participants younger than 18 provided informed assent.

2.2. Assessments.

Trauma histories were assessed with the National Survey of Adolescents (NSA; Kilpatrick et al., 2000) trauma section, which is a structured interview with behaviorally-specific questions to assess physical abuse by a caregiver, physical assault, sexual assault, witnessed domestic violence, witnessed community violence, and a range of other stressful life events. Two studies (n = 87) assessed the presence of mental health disorders, including PTSD, with the Mini-International Neuropsychiatric Interview for Children and Adolescents (MINI-Kid; Sheehan et al., 2010), while the two studies (n=52) utilized the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; Kaufman et al., 1997) to assess for mental health disorders. Participants also provided self-report measures of childhood maltreatment on the Childhood Trauma Questionnaire (CTQ; Bernstein and Fink, 1998) and PTSD and depression symptoms using the UCLA-PTSD Reaction Index (Steinberg et al., 2013). Verbal IQ was assessed with the Receptive One Word Picture Vocabulary test (Brownell, 2000).

2.3. Data Acquisition.

At the UAMS site, T1-weighted anatomic images were acquired with an MP-RAGE sequence (matrix = 192×192, 160 sagittal slices, TR/TE/FA = 7.5/3.7/9°, FOV = 256, 256, 160, final resolution = 1×1×1 mm) on a Philips Achieva 3T X-Series scanner with either an 8-channel (n = 33) or 32-channel (n = 70) headcoil. At the UW site, T1-weighted anatomic images were acquired with a similar MP-RAGE sequence (matrix = 256×256, 156 axial slices, TR/TE/FA = 8.2ms/3.2ms/12°, FOV = 25.6cm, final resolution = 1×1×1mm) on a GE MR750 3T scanner with an 8-channel headcoil.

2.4. Quality Control and Attrition.

All T1-weighted images (n = 139) were visually inspected to ensure that the resulting dataset contained only high-quality images. Following inspection, ten images were excluded due to poor quality T1s, largely due to participant head movement. The remaining 129 raw T1-images were retained, and the same raw images were processed through both manual edits and the automated pipeline.

2.5. Image Preprocessing.

Preprocessing of the raw T1-weighted images was completed in FreeSurfer Version 6.0.0 (Fischl et al., 2002; http://surfer.nmr.mgh.harvard.edu) on a Linux platform. FreeSurfer’s reconall command was used to preprocess all images through standard steps, including motion correction and averaging (Reuter et al., 2010), removal of non-brain tissue and skull-stripping (Segonne et al., 2007), automated Talairach transformation, segmentation of subcortical white and gray matter (Fischl et al., 2004, 2002), transformation to MNI305 atlas space, intensity normalization (Sled et al., 1998), and volumetric registration. Technical details for FreeSurfer are described in detail elsewhere (Dale et al., 1999; Dale and Sereno, 1993; Fischl et al., 2004, 2002, 2001; Fischl and Dale, 2000; Han et al., 2006; Jovicich et al., 2006; Reuter et al., 2010; Segonne et al., 2004).

2.6. Image Analysis.

For the automated dataset, reconall continued without intervention through the processing stream following initial preprocessing steps. This included steps for another intensity normalization, white matter segmentation and editing, tessellation of white and gray matter boundaries and automated topology correction (Fischl et al., 2001; Segonne et al., 2007), smoothing and surface inflation, spherical mapping and registration for each hemisphere (Dale et al., 1999; Dale and Sereno, 1993; Fischl and Dale, 2000), and parcellation and labelling of cortical and subcortical structures. Cortical structures were labelled according to the Desikan-Killiany-Tourville parcellation (DKT; Klein and Tourville, 2012) and mapped to the spherical surface, while subcortical and white matter structures are labelled according to FreeSurfer’s automated segmentation tool (aseg; Fischl et al., 2002) and mapped to the MNI305 common space. Final statistics for GMV of each region in both the DKT and subcortical aseg parcellations were generated, along with cortical thickness and surface area estimates for regions in the DKT parcellation. For the automated dataset, the run time for reconall was 20 – 40 hours per T1 image and processing was conducted in parallel scripts such that 10 images were processed simultaneously. Ultimately, processing time for the full automated dataset was 11.5 – 23.2 days.

For the manually-edited dataset, ten trained research assistants manually inspected each three-dimensional image in the dataset for points of misclassification of white matter, gray matter, and pial surface boundaries in three separate editing steps. Details on the research assistant training and the editing process are included in the Supplemental Materials. Following initial labeling, images were inspected for white matter omissions, and control points were added, extending the boundaries of the white matter segmentation to accurately incorporate white matter tracts. Next, images were inspected for white matter extensions, and edits were made to the segmentation to redraw white matter boundaries. Finally, corrections were made to any extensions of the pial surface to ensure that gray matter boundaries excluded dura and skull. All images were double-checked at each step by peer editors and trained supervisors to ensure that edits were implemented correctly. Before analysis, the first author inspected all edited images in the dataset for accuracy and image quality. On average, 9.5 hours of labor was spent for manually-editing each T1 image, excluding the reconall run time in between each edit, which ranged from 16–24 hours per edit; consequently, run time plus labor time for each T1 image in the manually-edited dataset was approximately 206 – 286 hours per image. The full manually-edited dataset was complete 18 months after beginning the editing process.

2.7. Region of Interest Selection.

Gray matter ROIs were selected a priori based on a recent meta-analysis of gray matter volume alterations in pediatric PTSD (Kribakaran et al., 2019). These brain volumes include total GMV, total cerebral volume, left and right hippocampal volume, and left and right amygdala volume. The DKT atlas was used to define these ROIs, with the temporal lobe ROI consisting of three separate ROIs for each hemisphere (inferior, middle, and superior temporal gyri) whose volumes were added together to generate a total volume estimate for the temporal lobe. Additionally, because the volume of the rostral anterior cingulate cortex (rACC) is also thought to be altered in PTSD (Meng et al., 2016), the rACC was also chosen as an ROI for this analysis.

2.8. Data Analysis.

Following processing in Freesurfer, gray matter ROIs from the DKT parcellation and total volume estimates were entered into linear mixed effects models with method (manual edits or automated process) as a within-subjects factor. Because each subject has two GMV estimates for each ROI, linear mixed effects models with by-subjects random intercepts and slopes were used for their ability to produce unbiased parameter estimates in datasets that violate the independence assumptions of ANOVA (Brauer and Curtin, 2018). Two sets of these models were created for each hemisphere; the primary model examined the effects of editing method, diagnostic group, and their interaction on GMV estimates. Based on a priori hypotheses regarding GMV differences related to PTSD diagnosis, “group” was dummy-coded with PTSD as the reference group, creating two contrasts: NTC vs. PTSD and TEC vs. PTSD. The secondary model, designed to replicate the group-level model with another commonly-used assessment of childhood trauma exposure, predicted GMV of each ROI from CTQ, editing method, and their interaction. Covariates for all models included participant age, verbal IQ, scanner site, estimated total intercranial volume (eTIV), and random effect of subject for each method. Models were estimated using the lmer function in R (lme4 package version 1.1–26).

3. RESULTS

3.1. Participants.

Following exclusion of images for poor quality, the remaining 129 participants were assigned to three different groups based on trauma exposure and diagnosis of PTSD. The non-trauma exposed control group (NTC; n = 59) did not differ in race or age from the trauma-exposed, no PTSD group (TEC; n = 31) or the trauma-exposed, PTSD group (PTSD; n = 39; Table 1). The NTC and PTSD groups differed from the TEC group in percentage of participants taking hormonal contraceptives (ps = .010 and .020, respectively), with a higher proportion of TEC participants taking some form of hormonal contraception compared to the other two groups. Additionally, both the TEC and PTSD groups differed from NTC in Verbal IQ (ps = .010 and .010, respectively). As expected, both trauma-exposed groups differed from NTC in CTQ total score, assault exposure, PTSD severity, and above-threshold major depression and anxiety disorder symptoms (all ps < .001). Interestingly, the PTSD group only differed from the TEC group in total number of sexual assault exposures (p = .010), percent of the sample with sexual assault exposure (p = .001), and above-threshold major depression and anxiety disorder symptoms (p < .001 and p = .020, respectively). See Table 1 for specific sample characteristics.

Table 1: Sample Demographic and Clinical Information.

Differences in proportions between groups were assessed with Chi-Square tests and differences in means between groups were assessed with ANOVA F-tests and followed-up with Tukey HSD.

Overall (n = 129) NTC (n = 59) TEC (n = 31) PTSD (n = 39) p-value group difference
NTC v. TEC NTC v. PTSD TEC v. PTSD
Mean Age (sd) 14.7 (1.9) 14.4 (2.0) 15.4 (1.5) 14.5 (1.8) .06 .99 .10
Mean Verbal IQ (sd) 105.1 (19.5) 111.6 (19.7) 98.4 (16.1) 100.7 (19.0) .01 .01 .87
Race
 Black (%) 27.1 23.7 25.8 33.3 .97 .42 .78
 White (%) 58.9 64.4 54.8 53.8 .41 .31 .95
 Latina (%) 3.1 1.7 3.2 5.1 1 .71 .54
 Other (%) 10.9 10.2 16.1 7.7 .63 .95 .42
Acquisition Site
 UAMS (n) 96 42 23 30
 UW (n) 33 17 8 9
Headcoil Channels
 8 (n) 55 28 17 10
 32 (n) 74 31 14 29
Mean CTQ Total (sd) 40.6 (17.7) 29.2 (4.8) 49.6 (17.0) 50.7 (20.6) < .01 < .01 .95
Mean Direct Assault Exposures (sd) 1.9 (2.6) 0 3.1 (2.2) 3.9 (2.9) < .01 < .01 .22
Mean Age at First Assault (sd) 7.7 (4.2) 0 7.6 (4.4) 7.6 (4.2) < .01 < .01 1
Mean Age at Last Assault (sd) 12.9 (3.1) 0 13.4 (3.5) 13 (2.7) < .01 < .01 .90
Mean Sexual Assault Exposures (sd) 1.0 (1.5) 0 1.3 (1.6) 2.3 (1.5) < .01 < .01 <.01
Sexual Assault Exposed (%) 43.4 0 58.1 94.9 < .01 < .01 <.01
Mean Physical Assault Exposures (sd) 0.6 (1.0) 0 1.1 (1.2) 0.9 (1.1) < .01 < .01 .15
Physical Assault Exposed (%) 32.6 0 61.3 51.3 < .01 < .01 .55
Mean Physical Abuse Exposures (sd) 0.4 (0.9) 0 0.7 (0.8) 0.8 (1.2) < .01 < .01 .95
Physical Abuse Exposed (%) 26.4 0 51.6 46.2 < .01 < .01 .91
Anxiety Disorder (%) 32.3 0 41.9 71.8 < .01 < .01 .02
PTSD (%) 30,2 0 0 100 - - -
Major Depression (%) 18.6 0 9.7 53.9 .07 < .01 < .01
UCLA-RI Sum (sd) 20.3 (23.1) 3.2 (6.7) 23.2 (18.3) 44 (20.1) < .01 < .01 < .01
Psychotropic Medication (%) 26.4 0 45.2 43.59 < .01 < .01 .89
Contraceptive Use (%) 14.0 5.1 35.5 10.3 .01 .62 .02
Open in a new tab

NTC = non-trauma exposed controls, TEC = trauma exposed controls, PTSD = posttraumatic stress disorder, UAMS = University of Arkansas for Medical Sciences, UW = University of Wisconsin, CTQ = Childhood Trauma Questionnaire, UCLA-RI = UCLA PTSD Reaction Index.

3.2. Effect of Editing Method on Gray Matter Volume Estimates.

Linear mixed effects models revealed a significant main effect of editing method on GMV of the left rostral anterior cingulate cortex, such that manual editing of brain volumes produced larger GMV estimates of this region in all clinical groups (t(250) = 11.61, p < .001; Figure S1). No other ROI displayed differential GMV estimates depending on editing method across all groups (Table 2).

Table 2:

Gray Matter Volume Estimates for All Regions of Interest.

ROI Gray Matter Volume Automated (mm3)
Mean (sd)		 Gray Matter Volume Manual Edits (mm3)
Mean (sd)
NTC TEC PTSD NTC TEC PTSD
Left Hemisphere
Amygdala 1631.0 (239.5) 1584.6 (177.1) 1518.4 (167.4) 1632.8 (239.6) 1584.6 (177.9) 1518.9 (170.9)
Hippocampus 3868.2 (385.6) 3818.0 (449.7) 3716.9 (342.3) 3870.0 (384.4) 3816.2 (414.2) 3717.2 (343.0)
rACC 3677.6 (617.1) 3498 (515.4) 3620.4 (506.7) 3694.6 (593.9) 3537.1 (561.7) 3650.6 (505.9)
Temporal Lobe 43223.9 (4798.8) 42555.5 (4756.8) 43259.1 (4934.2) 43258.3 (4766.9) 42642.9 (4962.2) 43259.2 (4807.1)
Right Hemisphere
Amygdala 1792.0 (244.3) 1774.6 (230.2) 1730.7 (240.1) 1794.4 (241.3) 1774.7 (231.9) 1730.8 (242.7)
Hippocampus 4031.1 (415.2) 3942.5 (449.7) 3850.1 (371.5) 4038.3 (412.4) 3942.0 (448.3) 3851.0 (372.7)
rACC 2709.8 (483.4) 2452.2 (493.0) 2728.4 (514.7) 2710.5 (466.1) 2464.0 (496.9) 2762.2 (534.7)
Temporal Lobe 42621.8 (4770.3) 41729.7 (4177.6) 42450.9 (4880.8) 42747.1 (4573.1) 41743.0 (4188.5) 42578.2 (4691.9)
Whole Brain
Total Cortical Volume 500915.2 (50822.7) 485011.2 (44003.3) 503715.0 (48281.5) 499849.0 (50313.5) 483297.8 (44339.3) 502927.5 (47677.8)
Total Gray Matter Volume 664369.2 (59927.5) 645429.2 (51692.0) 663604.0 (52508.0) 663342.9 (59472.5) 643715.7 (52206.3) 662816.4 (52002.7)
Open in a new tab

NTC = non-trauma exposed controls, TEC = trauma-exposed controls, PTSD = posttraumatic stress disorder, rACC = rostral anterior cingulate.

3.3. Gray Matter Volume of ROIs in PTSD.

Consistent with prior literature, linear mixed models revealed a significant main effect of PTSD diagnosis on GMV of the left amygdala, such that adolescents with a PTSD diagnosis demonstrated reduced GMV in this region compared to the NTC group (t(250) = 2.68, p = .008; Figure 2). This was the only ROI in this analysis to demonstrate altered GMV with PTSD diagnosis (Table 3) after single-step family-wise Bonferroni correction for multiple comparisons. While other ROIs failed to demonstrate significant differences in GMV between clinical groups, the effect sizes for the left and right hippocampus and right amygdala were within the small-to-moderate range (Table 3), which is consistent with prior work (Kribakaran et al., 2019).

Figure 2: Gray Matter Volume of the Left Amygdala is Reduced in Adolescents with PTSD.

Figure 2:

Open in a new tab

Adolescents with PTSD demonstrated reduced gray matter volume (GMV) in the left amygdala that was robust to editing method. GMV of the left amygdala is significantly reduced in PTSD subjects compared to non-trauma exposed controls (t(250) = 2.68, p = .008) and marginally reduced compared to trauma-exposed controls (t(250) = 2.10, p = .037). NTC = non-trauma exposed controls, TEC = trauma exposed controls, PTSD = posttraumatic stress disorder

Table 3:

Group-level statistics from linear mixed effects models predicting gray matter volume of ROIs from clinical group and editing method.

ROI Group Main Effect (NTC vs. PTSD) Group Main Effect (TEC vs. PTSD) Method Main Effect Group*Met hod Interaction (NTC vs. PTSD) Group*Met hod Interaction (TEC vs. PTSD)
t(250) p-value Cohe n’s d t(250) p-value Cohe n’s d t(250) p-value Cohe n’s d t(250) p-value t(250) p-value
Left Hemisphere
Amygdala 2.68 .008 0.483 2.10 .037 0.386 0.43 .670 0.076 0.82 .416 −0.29 .771
Hippocampus 1.55 .122 0.280 1.60 .111 0.290 0.17 .868 0.030 0.60 .550 −0.70 .488
rACC −0.01 .992 0.009 −0.02 .981 0.101 11.61 <.001 0.262 0.29 .774 −0.99 .324
Temporal Lobe −0.97 .333 0.175 −0.01 .997 0.002 <0.01 .998 <.001 0.33 .741 0.70 .483
Right Hemisphere
Amygdala 1.13 .261 0.204 1.17 .242 0.212 0.06 .953 0.010 0.79 .429 .001 .995
Hippocampus 1.93 .055 0.347 1.50 .134 0.273 0.18 .86 0.031 0.92 .359 −0.19 .853
rACC −0.62 .533 0.122 −2.26 .025 0.408 1.83 .068 0.326 −1.39 .165 −0.98 .426
Temporal Lobe −0.77 .445 0.129 −0.32 .750 0.059 0.87 .445 0.160 −0.01 .991 −0.53 .594
Whole Brain
Total Cortical Volume −1.82 .070 0.311 −1.17 .243 0.219 −1.62 .106 0.289 −0.45 .657 −1.27 .206
Total Gray Matter Volume −1.45 .148 0.248 −0.84 .402 0.157 −1.55 .123 0.276 −0.36 .716 −1.21 .227
Open in a new tab

rACC = rostral anterior cingulate cortex. Bolded values in table indicate significant differences after Bonferroni correction for multiple comparisons (for family-wise Bonferroni correction, the amygdala, hippocampus, rACC, and temporal lobe were each considered a family with four tests (right and left hemisphere for each group and CTQ (Table S1) as predictors of interest); therefore, the p-value threshold for significance was set at 0.0125).

3.4. Differences in Gray Matter Volume with Manual Editing in PTSD.

No volumes or regions demonstrated a group × method interaction, thereby failing to provide evidence that editing method influences GMV differences between PTSD groups (Table 3).

3.5. Relationship between Gray Matter Volume and Childhood Trauma Questionnaire.

Linear mixed effects models examining the relationship between gray matter volume of ROIs and CTQ responses revealed no significant correlations between severity of childhood trauma and gray matter volume of the amygdala, rACC, temporal lobe, total cortical volume, or total gray matter volume. A weak correlation was found between the CTQ and hippocampal volume such that individuals with heightened CTQ scores demonstrated reduced GMV of the right hippocampus (t(250) = −2.24, p = 0.027; Table S1); however, this relationship does not survive correction for multiple comparisons. Additionally, a CTQ by method interaction in the rACC was also found (t(250 = 2.50, p = 0.014) but fails to survive correction for multiple comparisons and thus was not interpreted.

4. DISCUSSION

The present study investigated critical methodological questions regarding the robustness of GMV alterations in PTSD and the utility of manual intervention in FreeSurfer-based image analysis in adolescent girls with trauma exposure and PTSD. Consistent with prior literature (Lim et al., 2014; Mutluer et al., 2018), we provide further evidence for reduced GMV of the left amygdala in adolescents with a PTSD diagnosis compared to both non-trauma exposed and trauma exposed control groups. Also consistent with prior literature (Kribakaran et al., 2019), we report small-to-moderate effect sizes for the difference in GMV of the left and right hippocampus and right amygdala between groups. Importantly, this analysis also examined the utility of manual intervention editing strategies in this clinical population, and ultimately, findings fail to provide support for the use of manual edits in investigations of the GMV alterations commonly associated with PTSD. Due to the time- and labor-intensive process of manual intervention with FreeSurfer, the similarity between the predictions generated from the manually-edited and automated datasets suggest that manual intervention is unnecessary in GMV-based investigations of group differences in adolescent trauma exposure and PTSD.

While analyses in this report provide additional evidence for reduced GMV of the amygdala in adolescent PTSD, investigations of subcortical GMV alterations in trauma-related clinical groups often produce conflicting results. Because a hyperactive amygdala during fear- and emotion-regulation tasks is well-characterized in PTSD (Admon et al., 2013; Patel et al., 2012), the amygdala is often a target of investigation for structural studies in PTSD as well. In fact, some evidence from studies in adults suggests that reduced GMV of the amygdala may be related to childhood trauma exposure (Paquola et al., 2016) and PTSD (O’Doherty et al., 2017). However, the pediatric PTSD literature is less clear, with trending (Kribakaran et al., 2019; Milani et al., 2017) or non-existent (De Bellis et al., 2002; Woon and Hedges, 2008) effects of PTSD on amygdala GMV. Though the size of the effect of PTSD on left amygdala GMV reported in this analysis is larger than the trending effects noted in the meta-analysis from Kribakaran and colleagues, suggesting stronger evidence for reduced GMV of the amygdala in adolescents with PTSD, it is likely that PTSD-related GMV alterations present differently between adolescents and adults. Nonetheless, the current study adds to this literature and demonstrates amygdala GMV reductions in PTSD among a well-characterized and relatively large sample.

FreeSurfer is a commonly used and well-validated tool for structural neuroimaging analysis, but no common standard exists for how researchers should choose whether or not to conduct manual interventions on structural datasets. Of the five studies to our knowledge that used FreeSurfer to examine gray matter volume abnormalities in pediatric trauma exposure and PTSD, two disclose the use of manual intervention (Tottenham et al., 2010; Weems et al., 2015), two fail to mention if manual intervention was used (Lim et al., 2018; Morey et al., 2016) and one explicitly noted that manual edits were not conducted (Ahmed et al., 2012). This differing methodology may contribute to the ambiguity regarding the effects of pediatric trauma exposure on GMV. The reported results from the two manually-edited datasets (Tottenham et al., 2010; Weems et al., 2015) indicate no difference in regional GMV between trauma-exposed and control groups while the other three studies that either did not use manual edits or did not explicitly mention the use of manual edits (Ahmed et al., 2012; Lim et al., 2018; Morey et al., 2016) provide evidence for reduced regional GMV throughout the brain in clinical groups. Similarly, our finding of a main effect of editing method on left rACC volumes (Figure S1) suggests that the choice to utilize manual interventions on the FreeSurfer reconall process may produce systematic differences between studies of GMV in pediatric subjects and could possibly contribute to spurious inconsistencies in models of structural gray matter differences between groups. While we failed to observe a method × group interaction in the left rACC, which indicates that detection of trauma- and PTSD-related differences in GMV is unaffected by choice of editing method, future studies should investigate which ROIs are the most susceptible to method-related differences in GMV estimates.

The analyses presented in this report are the first to provide a comprehensive comparison of manual intervention and automated segmentation in FreeSurfer on a large dataset of adolescents with and without trauma exposure. Our results suggest that manual editing of structural images in FreeSurfer is unnecessary for detecting differences in gray matter volume related to PTSD in this clinical sample. Because our sample is limited to female adolescents with trauma exposure, future work should address these same questions in a more sex- and gender-heterogenous sample of adolescents with a variety of trauma exposure types to test the generalizability of our conclusions. Additionally, we did not exclude for psychiatric comorbidity within our trauma-exposed sample, and thus our analyses are limited by potential variance introduced by comorbid psychopathology. Future studies of structural gray matter alterations in PTSD should standardize the choice of whether or not to conduct manual edits of the FreeSurfer automated segmentation to build a clearer picture of the structural brain differences in PTSD and control groups. Due to the time- and labor-intensive process of conducting manual edits with FreeSurfer and similarity of findings generated from both datasets, we suggest that researchers investigating gray matter alterations in PTSD should refrain from the manual intervention process in favor of the fully-automated FreeSurfer pipeline.

Supplementary Material

1

Highlights.

  • FreeSurfer is used to create automated and manually-edited structural datasets from adolescent girls with and without trauma exposure

  • Gray matter volumes are compared between adolescent girls with and without trauma exposure and PTSD

  • Girls with PTSD demonstrate reduced gray matter volume of the left amygdala compared to controls, which is robust to editing method

  • Manual edits with FreeSurfer to not provide additional benefit in this adolescent PTSD sample.

Funding and Disclosures

The authors have no disclosures or conflicts of interest to report.

Grant Support Acknowledgements:

Research in this publication was supported by the National Institute of Mental Health and the National Institutes of Health under awards T32MH018931-31, F31MH122047, T32GM007507, MH119132, MH108753, MH106860, and MH097784

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  1. Admon R, Milad MR, Hendler T, 2013. A causal model of post-traumatic stress disorder: disentangling predisposed from acquired neural abnormalities. Trends Cogn. Sci 17, 337–347. 10.1016/j.tics.2013.05.005 [DOI] [PubMed] [Google Scholar]
  2. Ahmed F, Ahmed Fatima, Spottiswoode BS, Carey PD, Stein DJ, Seedat S, 2012. Relationship between Neurocognition and Regional Brain Volumes in Traumatized Adolescents with and without Posttraumatic Stress Disorder. Neuropsychobiology 66, 174–184. 10.1159/000339558 [DOI] [PubMed] [Google Scholar]
  3. Beelen C, Phan TV, Wouters J, Ghesquière P, Vandermosten M, 2020. Investigating the Added Value of FreeSurfer’s Manual Editing Procedure for the Study of the Reading Network in a Pediatric Population. Front. Hum. Neurosci 14. 10.3389/fnhum.2020.00143 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bernstein DP, Fink L, 1998. Childhood trauma questionnaire: A retrospective self-report: Manual. Harcourt Brace & Company. [Google Scholar]
  5. Brauer M, Curtin JJ, 2018. Linear mixed-effects models and the analysis of nonindependent data: A unified framework to analyze categorical and continuous independent variables that vary within-subjects and/or within-items. Psychol. Methods 23, 389–411. 10.1037/met0000159 [DOI] [PubMed] [Google Scholar]
  6. Brownell R, 2000. Receptive one-word picture vocabulary test: Manual. Academic Therapy Publications. [Google Scholar]
  7. Cisler JM, Esbensen K, Sellnow K, Ross M, Weaver S, Sartin-Tarm A, Herringa RJ, Kilts CD, 2018a. Differential Roles of the Salience Network During Prediction Error Encoding and Facial Emotion Processing Among Female Adolescent Assault Victims. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 10.1016/j.bpsc.2018.08.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cisler JM, Privratsky A, Smitherman S, Herringa RJ, Kilts CD, 2018b. Large-scale brain organization during facial emotion processing as a function of early life trauma among adolescent girls. NeuroImage Clin. 17, 778–785. 10.1016/j.nicl.2017.12.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cisler JM, Scott Steele J, Smitherman S, Lenow JK, Kilts CD, 2013. Neural processing correlates of assaultive violence exposure and PTSD symptoms during implicit threat processing: A network-level analysis among adolescent girls. Psychiatry Res. Neuroimaging 214, 238–246. 10.1016/j.pscychresns.2013.06.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cisler JM, Sigel BA, Kramer TL, Smitherman S, Vanderzee K, Pemberton J, Kilts CD, 2015. Amygdala response predicts trajectory of symptom reduction during Trauma-Focused Cognitive-Behavioral Therapy among adolescent girls with PTSD. J. Psychiatr. Res 71, 33–40. 10.1016/j.jpsychires.2015.09.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dale A, Fischl B, Sereno MI, 1999. Cortical Surface-Based Analysis: I. Segmentation and Surface Reconstruction. NeuroImage 9, 179–194. [DOI] [PubMed] [Google Scholar]
  12. Dale AM, Sereno MI, 1993. Improved Localizadon of Cortical Activity by Combining EEG and MEG with MRI Cortical Surface Reconstruction: A Linear Approach. J. Cogn. Neurosci 5, 162–176. 10.1162/jocn.1993.5.2.162 [DOI] [PubMed] [Google Scholar]
  13. De Bellis MD, Keshavan MS, Shifflett H, Iyengar S, Beers SR, Hall J, Moritz G, 2002. Brain structures in pediatric maltreatment-related posttraumatic stress disorder: a sociodemographically matched study. Biol. Psychiatry 52, 1066–1078. 10.1016/S0006-3223(02)01459-2 [DOI] [PubMed] [Google Scholar]
  14. Fischl B, Dale AM, 2000. Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proc. Natl. Acad. Sci. U. S. A 97, 11050–11055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fischl B, Liu A, Dale AM, 2001. Automated manifold surgery: constructing geometrically accurate and topologically correct models of the human cerebral cortex. IEEE Med. Imaging 20, 70–80. [DOI] [PubMed] [Google Scholar]
  16. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, van der Kouwe A, Killiany R, Kennedy D, Klaveness S, Montillo A, Makris N, Rosen B, Dale AM, 2002. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 33, 341–355. [DOI] [PubMed] [Google Scholar]
  17. Fischl B, Salat DH, Kouwe A.J.W. van der, Makris N, Ségonne F, Quinn BT, Dale AM, 2004. Sequence-independent segmentation of magnetic resonance images. NeuroImage 23, S69–S84. https://doi.org/DOI: 10.1016/j.neuroimage.2004.07.016 [DOI] [PubMed] [Google Scholar]
  18. Han X, Jovicich J, Salat D, van der Kouwe A, Quinn B, Czanner S, Busa E, Pacheco J, Albert M, Killiany R, Maguire P, Rosas D, Makris N, Dale A, Dickerson B, Fischl B, 2006. Reliability of MRI-derived measurements of human cerebral cortical thickness: The effects of field strength, scanner upgrade and manufacturer. NeuroImage 32, 180–194. [DOI] [PubMed] [Google Scholar]
  19. Hart H, Rubia K, 2012. Neuroimaging of child abuse: a critical review. Front. Hum. Neurosci 6. 10.3389/fnhum.2012.00052 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hayes JP, VanElzakker MB, Shin LM, 2012. Emotion and cognition interactions in PTSD: a review of neurocognitive and neuroimaging studies. Front. Integr. Neurosci 6. 10.3389/fnint.2012.00089 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Jovicich J, Czanner S, Greve D, Haley E, Kouwe A. van der, Gollub R, Kennedy D, Schmitt F, Brown G, MacFall J, Fischl B, Dale A, 2006. Reliability in multi-site structural MRI studies: Effects of gradient non-linearity correction on phantom and human data. NeuroImage 30, 436–443. https://doi.org/DOI: 10.1016/j.neuroimage.2005.09.046 [DOI] [PubMed] [Google Scholar]
  22. Kaufman J, Birmhauer B, Brent D, Rao U, Flynn C, Moreci P, Williamson D, Ryan N, 1997. Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): Initial Reliability and Validity Data. J. Am. Acad. Child Adolesc. Psychiatry 36, 980–988. 10.1097/00004583-199707000-00021 [DOI] [PubMed] [Google Scholar]
  23. Kilpatrick DG, Acierno R, Saunders B, Resnick HS, Best CL, Schnurr PP, 2000. Risk factors for adolescent substance abuse and dependence: data from a national sample. J. Consult. Clin. Psychol 68, 19–30. [DOI] [PubMed] [Google Scholar]
  24. Klein A, Tourville J, 2012. 101 Labeled Brain Images and a Consistent Human Cortical Labeling Protocol. Front. Neurosci 6. 10.3389/fnins.2012.00171 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kribakaran S, Danese A, Bromis K, Kempton MJ, Gee DG, 2019. Meta-Analysis of Structural MRI Studies in Pediatric PTSD and Comparison with Related Conditions. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 10.1016/j.bpsc.2019.08.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lim L, Hart H, Mehta M, Worker A, Simmons A, Mirza K, Rubia K, 2018. Grey matter volume and thickness abnormalities in young people with a history of childhood abuse. Psychol. Med 48, 1034–1046. 10.1017/S0033291717002392 [DOI] [PubMed] [Google Scholar]
  27. Lim L, Radua J, Rubia K, 2014. Gray Matter Abnormalities in Childhood Maltreatment: A Voxel-Wise Meta-Analysis. Am. J. Psychiatry 171, 854–863. 10.1176/appi.ajp.2014.13101427 [DOI] [PubMed] [Google Scholar]
  28. McCarthy CS, Ramprashad A, Thompson C, Botti J-A, Coman IL, Kates WR, 2015. A comparison of FreeSurfer-generated data with and without manual intervention. Front. Neurosci 9. 10.3389/fnins.2015.00379 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Meng L, Jiang J, Jin C, Liu J, Zhao Y, Wang W, Li K, Gong Q, 2016. Trauma-specific Grey Matter Alterations in PTSD. Sci. Rep 6, 33748. 10.1038/srep33748 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Merikangas KR, He J, Burstein M, Swanson SA, Avenevoli S, Cui L, Benjet C, Georgiades K, Swendsen J, 2010. Lifetime Prevalence of Mental Disorders in US Adolescents: Results from the National Comorbidity Study-Adolescent Supplement (NCS-A). J. Am. Acad. Child Adolesc. Psychiatry 49, 980–989. 10.1016/j.jaac.2010.05.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Milani ACC, Hoffmann EV, Fossaluza V, Jackowski AP, Mello MF, 2017. Does pediatric post-traumatic stress disorder alter the brain? Systematic review and meta-analysis of structural and functional magnetic resonance imaging studies. Psychiatry Clin. Neurosci 71, 154–169. 10.1111/pcn.12473 [DOI] [PubMed] [Google Scholar]
  32. Morey RA, Haswell CC, Hooper SR, Bellis MDD, 2016. Amygdala, Hippocampus, and Ventral Medial Prefrontal Cortex Volumes Differ in Maltreated Youth with and without Chronic Posttraumatic Stress Disorder. Neuropsychopharmacology 41, 791–801. 10.1038/npp.2015.205 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Mutluer T, Şar V, Kose-Demiray Ç, Arslan H, Tamer S, Inal S, Kaçar AŞ, 2018. Lateralization of Neurobiological Response in Adolescents with Post-Traumatic Stress Disorder Related to Severe Childhood Sexual Abuse: the Tri-Modal Reaction (T-MR) Model of Protection. J. Trauma Dissociation 19, 108–125. 10.1080/15299732.2017.1304489 [DOI] [PubMed] [Google Scholar]
  34. O’Doherty DCM, Tickell A, Ryder W, Chan C, Hermens DF, Bennett MR, Lagopoulos J, 2017. Frontal and subcortical grey matter reductions in PTSD. Psychiatry Res. Neuroimaging 266, 1–9. 10.1016/j.pscychresns.2017.05.008 [DOI] [PubMed] [Google Scholar]
  35. Paquola C, Bennett MR, Lagopoulos J, 2016. Understanding heterogeneity in grey matter research of adults with childhood maltreatment—A meta-analysis and review. Neurosci. Biobehav. Rev 69, 299–312. 10.1016/j.neubiorev.2016.08.011 [DOI] [PubMed] [Google Scholar]
  36. Patel R, Spreng RN, Shin LM, Girard TA, 2012. Neurocircuitry models of posttraumatic stress disorder and beyond: A meta-analysis of functional neuroimaging studies. Neurosci. Biobehav. Rev 36, 2130–2142. 10.1016/j.neubiorev.2012.06.003 [DOI] [PubMed] [Google Scholar]
  37. Reuter M, Rosas HD, Fischl B, 2010. Highly Accurate Inverse Consistent Registration: A Robust Approach. NeuroImage 53, 1181–1196. 10.1016/j.neuroimage.2010.07.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Segonne F, Dale AM, Busa E, Glessner M, Salat D, Hahn HK, Fischl B, 2004. A hybrid approach to the skull stripping problem in MRI. NeuroImage 22, 1060–1075. https://doi.org/DOI: 10.1016/j.neuroimage.2004.03.032 [DOI] [PubMed] [Google Scholar]
  39. Segonne F, Pacheco J, Fischl B, 2007. Geometrically accurate topology-correction of cortical surfaces using nonseparating loops. IEEE Trans Med Imaging 26, 518–529. [DOI] [PubMed] [Google Scholar]
  40. Sellnow K, Sartin-Tarm A, Ross MC, Weaver S, Cisler JM, 2020. Biotypes of functional brain engagement during emotion processing differentiate heterogeneity in internalizing symptoms and interpersonal violence histories among adolescent girls. J. Psychiatr. Res 121, 197–206. 10.1016/j.jpsychires.2019.12.002 [DOI] [PubMed] [Google Scholar]
  41. Sheehan DV, Sheehan KH, Shytle RD, Janavs J, Bannon Y, Rogers JE, Milo KM, Stock SL, Wilkinson B, 2010. Reliability and Validity of the Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-KID). J. Clin. Psychiatry 71, 313–326. 10.4088/JCP.09m05305whi [DOI] [PubMed] [Google Scholar]
  42. Sled JG, Zijdenbos AP, Evans AC, 1998. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 17, 87–97. [DOI] [PubMed] [Google Scholar]
  43. Steinberg AM, Brymer MJ, Kim S, Briggs EC, Ippen CG, Ostrowski SA, Gully KJ, Pynoos RS, 2013. Psychometric Properties of the UCLA PTSD Reaction Index: Part I. J. Trauma. Stress 26, 1–9. 10.1002/jts.21780 [DOI] [PubMed] [Google Scholar]
  44. Tottenham N, Hare TA, Quinn BT, McCarry TW, Nurse M, Gilhooly T, Millner A, Galvan A, Davidson MC, Eigsti I-M, Thomas KM, Freed PJ, Booma ES, Gunnar MR, Altemus M, Aronson J, Casey BJ, 2010. Prolonged institutional rearing is associated with atypically large amygdala volume and difficulties in emotion regulation. Dev. Sci 13, 46–61. 10.1111/j.1467-7687.2009.00852.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Weems CF, Klabunde M, Russell JD, Reiss AL, Carrión VG, 2015. Post-traumatic stress and age variation in amygdala volumes among youth exposed to trauma. Soc. Cogn. Affect. Neurosci 10, 1661–1667. 10.1093/scan/nsv053 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Woon FL, Hedges DW, 2008. Hippocampal and amygdala volumes in children and adults with childhood maltreatment-related posttraumatic stress disorder: A meta-analysis. Hippocampus 18, 729–736. 10.1002/hipo.20437 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1700526-supplement-1.docx (1.1MB, docx)

RESOURCES