Journal of Neurotrauma

41:942–956 (April 2024)

Mary Ann Liebert, Inc.
DOI: 10.1089/neu.2023.0309

Open camera or QR reader and

scan code to access this article
and other resources online.

ORIGINAL ARTICLE MILITARY

Neurological Effects of Repeated Blast Exposure

in Special Operations Personnel
James R. Stone,1,* Brian B. Avants,1 Nicholas J. Tustison,1 Jessica Gill,2 Elisabeth A. Wilde,3,4 Kiel D. Neumann,5
Leslie A. Gladney,1 Madison O. Kilgore,1 F. Bowling,6 Christopher M. Wilson,6 John F. Detro,6
Heather G. Belanger,7,8 Katryna Deary,6 Hans Linsenbardt,9 and Stephen T. Ahlers10

Abstract
Exposure to blast overpressure has been a pervasive feature of combat-related injuries. Studies exploring the neu-
rological correlates of repeated low-level blast exposure in career ‘‘breachers’’ demonstrated higher levels of
tumor necrosis factor alpha (TNFa) and interleukin (IL)-6 and decreases in IL-10 within brain-derived extracellular
vesicles (BDEVs). The current pilot study was initiated in partnership with the U.S. Special Operations Command
(USSOCOM) to explore whether neuroinflammation is seen within special operators with prior blast exposure.
Data were analyzed from 18 service members (SMs), inclusive of 9 blast-exposed special operators with an exten-
sive career history of repeated blast exposures and 9 controls matched by age and duration of service. Neuroin-
flammation was assessed utilizing positron emission tomography (PET) imaging with [18F]DPA-714. Serum was
acquired to assess inflammatory biomarkers within whole serum and BDEVs. The Blast Exposure Threshold Sur-
vey (BETS) was acquired to determine blast history. Both self-report and neurocognitive measures were ac-
quired to assess cognition. Similarity-driven Multi-view Linear Reconstruction (SiMLR) was used for joint
analysis of acquired data. Analysis of BDEVs indicated significant positive associations with a generalized
blast exposure value (GBEV) derived from the BETS. SiMLR-based analyses of neuroimaging demonstrated
exposure-related relationships between GBEV, PET-neuroinflammation, cortical thickness, and volume loss
within special operators. Affected brain networks included regions associated with memory retrieval and exec-
utive functioning, as well as visual and heteromodal processing. Post hoc assessments of cognitive measures
failed to demonstrate significant associations with GBEV. This emerging evidence suggests neuroinflammation
may be a key feature of the brain response to blast exposure over a career in operational personnel. The com-
mon thread of neuroinflammation observed in blast-exposed populations requires further study.
Keywords: military; neuroinflammation; neuroimaging; repeated blast exposure; serum biomarker

1
Department of Radiology and Medical Imaging, University of Virginia, Charlottesville, Virginia, USA.
2
School of Nursing, Johns Hopkins University, Baltimore, Maryland, USA.
3
Department of Neurology, University of Utah, Salt Lake City, Utah, USA.
4
George E. Wahlen VA, Salt Lake City Health Healthcare System, Salt Lake City, Utah, USA.
5
Molecular Imaging Research Hub, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA.
6
U.S. Special Operations Command, Tampa, Florida, USA.
7
Departments of Psychiatry and Behavioral Neurosciences, and Psychology, University of South Florida, Tampa, Florida, USA.
8
Cognitive Research Corporation, St. Petersburg, Florida, USA.
9
Naval Medical Research Command, Silver Spring, Maryland, USA.
10
Operational and Undersea Medicine Directorate, Naval Medical Research Command, Silver Spring, Maryland, USA.

*Address correspondence to: James R. Stone, MD, PhD, Department of Radiology and Medical Imaging, University of Virginia, 480 Ray C. Hunt Drive, Box 801339, Charlottesville,
VA 22903, USA E-mail: [email protected]

ª James R. Stone et al., 2024; Published by Mary Ann Liebert, Inc. This Open Access article is distributed under the terms of the Creative Commons License (CC-BY) (http://
creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

942
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 943

Introduction ations in cortical thickness, decreased white matter

Exposure to blast overpressure has been a pervasive integrity, and reductions in activation of the default
feature of combat-related injuries experienced by opera- mode network (DMN).8 Among the findings within expe-
tional personnel. In addition to combat-related blast- rienced breachers were elevated markers of neuroin-
induced traumatic brain injury (b-TBI), concern exists flammation within brain-derived extracellular vesicles
that repeated exposure to low-intensity blasts during (BDEVs) isolated from peripheral blood. Specifically,
day-to-day training and operations may be linked to increases in pro-inflammatory tumor necrosis factor
long-term neurological sequelae. This concern has fueled alpha (TNFa) and interleukin (IL)-6 and decreases in
research exploring potential cumulative injury mecha- anti-inflammatory IL-10 were noted, resulting in an ele-
nisms following repeated blast exposure. A major focus vation of the IL-6/IL-10 ratios, consistent with increased
of these efforts has been to assess the effects of repeated inflammation in the brains of experienced breachers.9
low-intensity blast exposure in military operational per- Additionally, a pilot study of whole-blood transcriptomic
sonnel with repeated exposures throughout their military analysis demonstrated a number of dysregulated genes
careers. Low-level blast has been previously described as related to chronic inflammation and immune response
referring to overpressure generated by outgoing muni- within experienced breachers compared with well mat-
tions, which tends to be lower in intensity as compared ched controls.10
with high-level blast, which has been described as over- The finding of elevated neuroinflammation within
pressure generated by incoming munitions.1 A number blast-exposed populations may have implications for
of studies have explored neurological sequelae in the long-term health of blast-exposed service members
‘‘breachers,’’ who employ explosives to breach hardened (SMs). Neuroinflammation plays a critical role in the
structures, such as walls or buildings. immune response following injury, where microglia and
Breachers are routinely exposed to repeated low-level astrocytes play a key role in mediating the overall neuroin-
blasts during training and operations.2 However, expo- flammatory response.11,12 Although the neuroinflamma-
sure to these individual low-level blasts is not associated tory response is potentially adaptive in early recovery
with an injury-level event. A study examining anony- intervals, there is growing evidence that injury may also
mous survey respondents from military and non-military initiate a prolonged and maladaptive pro-inflammatory
law enforcement populations reported an increased prev- response that can result in deleterious changes to brain
alence and severity of symptoms in chronically blast- structure12,13 and function.14 Many neurodegenerative
exposed populations, including descriptions of headaches, conditions, such as Alzheimer’s disease (AD), Parkinson’s
difficulty remembering, slowed thinking, loss of energy, disease (PD), and amyotrophic lateral sclerosis (ALS),
feelings of aggression, social withdrawal, and irritabil- share a common theme of chronic neuroinflammation15
ity.3 This work suggests subclinical effects associated that may be detectable in prodromal periods, and that
with repeated low-level blast exposure may accumulate may also share links to TBI.16 Accumulating evidence
over time and culminate in clinically significant chronic suggests that dysregulated neuroinflammation can persist
symptoms. More broadly, epidemiological studies exam- for years or decades following TBI, leading to secondary
ining the relationship between military occupational spe- tissue compromise, and this has been associated with
cialties (MOS) associated with high risk of blast poorer outcome.17
exposure and neurologically related medical outcomes Recent human studies have employed positron emis-
have demonstrated increased prevalence of neurosensory sion tomography (PET) radiopharmaceuticals to image
abnormalities, TBI, cognitive abnormalities, headaches, neuroinflammation. One of the more widely used class
behavioral health conditions, substance abuse, and symp- of radiopharmaceuticals in the study of the neuroinflam-
toms such as anxiety, fatigue, and migraines in high-risk matory response targets the translocator protein (TSPO).
MOS as compared with other specialties.4,5 Additionally, TSPO is an 18 kDa mitochondrial associated protein with
high-risk MOS were more likely to undergo evalua- a variety of functions throughout the body including cho-
tion for disability or be medically separated from the lesterol transport, bile acid synthesis, cardiac contractil-
military.4,5 ity, apoptosis, stress adaptation, and modulation of the
A study performed by our group that assessed experi- immune system. Within the brain, TSPO is found within
enced breachers with a career history of repeated blast microglia and demonstrates significantly increased exp-
exposure demonstrated several significant alterations ression within activated microglia. As such, radiophar-
in blast-exposed populations compared with controls. maceuticals consisting of a TSPO ligand coupled to a
These included symptoms of tinnitus, light and noise sen- PET radioisotope serve as effective imaging agents for
sitivity, irritability, and memory alterations along with neuroinflammation.
postural instability and hearing loss.6,7 Additionally, A study of former athletes demonstrated significant
magnetic resonance imaging (MRI) demonstrated alter- increase in uptake of the TSPO ligand [11C]DPA-713
ations in brain structure and function, including alter- within former National Football League (NFL) players
944 STONE ET AL.

as compared with healthy, age-matched controls.18 Sim- other medical condition that may affect cerebral metabo-
ilarly, the TSPO ligand [18F]DPA-714 was recently used lism were all criteria for exclusion from the study.
to study microglial activation in collegiate athletes fol-
lowing a sports concussion, compared with healthy, Demographics, clinical history,
age-matched controls.19 That study demonstrated persis- and neuropsychological assessment
tent elevated neuroinflammation in collegiate athletes Participant history was acquired, including demograph-
who were diagnosed with a sport concussion and cleared ics, military history, and medical history, along with a
for unrestricted return to play based on a clinical recov- head injury questionnaire. The head injury questionnaire
ery. Using the TSPO ligand [11C]PK11195, a recent included eliciting a history of cause of injury, age of
study showed increased neuroinflammation in the hippo- injury, loss of consciousness, presence of amnesia related
campus of athletes with sports-related concussion.20 to the event, changes in mood, sleep problems, or other
Additionally, [11C]PK11195 has been used to demon- associated clinical symptoms. Additionally, subjects
strate elevated neuroinflammation 6 months following were administered the Naval Medical Research Center
injury in patients with TBI as compared with age- (NMRC) Blast Exposure Threshold Survey (BETS).22
matched healthy controls.21 The BETS elicits information on exposure to weapons
Based on the above, a pilot study was initiated by our and explosives, injury history, auditory symptoms, vestib-
group in partnership with the U.S. Special Operations ular symptoms, mood issues, sleep issues, and cognitive
Command (USSOCOM) to explore whether neuroinflam- symptoms. Neuropsychological testing and administered
mation is seen within special operators with a history of inventories included the Automated Neuropsychological
prior blast exposure. This study was designed as an initial Assessment Metrics 4 TBI-MIL (ANAM4 TBI-MIL),
feasibility effort to recruit a small number of Special Combat Exposure Scale (CES), Pittsburgh Sleep Quality
Operations Forces (SOF) SMs with exposure to repeated Index (PSQI), PTSD Checklist for DSM-5 (PCL-5), Psy-
low-level blasts over their careers. The study is the first chological General Well-Being Index (PGWI), and Neu-
to explore neuroinflammation as a pivotal mechanism robehavioral Symptom Inventory (NSI).
through which blast-related changes impact brain struc-
ture and function. Further, this is the first work to explore Generalized blast exposure value (GBEV)
neuroimaging-derived critical generalized blast exposure The BETS with the accompanying GBEV was recently
value (GBEV) thresholds for blast exposure. developed specifically for military SMs to relate the type
and amount of blast exposure to reported symptomol-
Methods ogy.22 This value is analogous to the Cumulative Head
All procedures were reviewed and approved by the Insti- Impact Index (CHII) for former high school and college
tutional Review Board at the University of Virginia football players.23 The motivation for such measures is
(UVA) and the study protocol was approved by the to simultaneously characterize a specific population and
Naval Medical Research Center Institutional Review identify brain trauma exposure, relate that exposure to
Board in compliance with all applicable federal regula- outcomes, and determine a threshold for increased risk
tions governing the protection of human subjects. The of long-term problematic health outcomes. Analogously,
investigators have adhered to the policies for protection the GBEV is determined by the following formula:
of human subjects as prescribed in AR 70-25.
GBEV = 0:976ð1BEC Þ þ 288ð2BEC Þ þ 41ð3BECÞ
Informed consent was provided by all participants and
þ 77ð4BEC Þð4freqÞ þ 75ð5BEC Þð5freqÞ
procedures were performed during a 1-day evaluation at
UVA. A total of 21 SMs were recruited into experimental where BEC is the ‘‘blast exposure count’’ defined as
and control groups for the study. SMs were screened for the product of 1) years of experience with a weapon,
inclusion in the study by a SOCOM nurse practitioner 2) months of experience per year, 3) days of experience
located at MacDill Air Force Base in Tampa, Florida. per month, and 4) number of exposures per day. Catego-
SMs within the experimental group must have had a rization of Light arms, Artillery, Recoilless rifles, and
prior history of blast exposure in training or operations. Explosives (CLARE) are denoted for each term in paren-
Subjects in the control group must not have had a history theses: (1BEC) small arms; (2BEC) large arms (including
of previous exposure to explosives, including but not lim- shoulder-fired); (3BEC) artillery (or large weapons car-
ited to explosive entry (breacher) operations/training, ried by a vehicle); (4BEC) small explosives; and
heavy weapons use, and/or explosives ordinance disposal (5BEC) large explosives. freq refers to the daily frequ-
(EOD). A history of moderate or severe TBI as defined ency variable for each category.
by the American Congress of Rehabilitation, MRI contra-
indications, current severe medical condition, current Serum biomarkers
diagnosis of central nervous system (CNS) disorder All samples were measured in a fully blinded man-
other than mild TBI, and any cardiac, respiratory, or ner. Blood samples were collected into tubes with
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 945

ethylenediaminetetraacetic acid (EDTA) and centri- Neuroimaging

fuged before being aliquoted and placed in the freezer
PET-CT. Imaging was performed with a Siemens Biog-
at 80C for long-term storage. Extracellular vesicles
raph mCT (PET-CT) scanner (Siemens Healthineers,
(EVs) were isolated from thawed samples as previ-
Knoxville, TN, USA). Measurements of body height
ously published.24 Of thawed samples, 500 lL were
and weight were performed before scan acquisition.
defibrinated with thromboplastin D and incubated at
A computed tomography (CT) scan was performed with
room temperature for 30 min before centrifuging for
a 100 mAs tube current, 100 kV voltage, 64 · 0.6 mm col-
5 min at 10,000g at 4C. The supernatant was trans-
limation, pitch of 0.55, reconstruction at a 2-mm slice
ferred into fresh tubes with added ExoQuick exosome
thickness, and scanning time of 17.3 sec. Subjects recei-
solution, mixed, and incubated for 60 min at 4C be-
ved an intravenous, bolus injection of [18F]DPA-714
fore centrifuging. The resulting pellet containing all EVs
(296 – 17 MBq) at the onset of a 90-min dynamic list
was resuspended in 500 lL ultra-pure water and placed
mode PET acquisition. Genotyping was performed to
at 80C for long-term storage. EVs were precipitated
determine TSPO polymorphism status for each subject.
by using ExoQuick plasma prep and exosome precipitation
PET image reconstruction was performed at an image
kits (System Bioscience, Inc.). Five microliters of throm-
size of 400 with a 2.0 amplification factor and Gaussian
bin (5000 U/mL in phosphate-buffered saline [PBS])
filter with a full-width half maximum of 2.0. The
were added to 500 lL plasma to a final concentration of
TrueX + time of flight (TOF) method was used with
5 U/mL thrombin. Tubes were incubated at room temper-
five iterations and 21 subsets. Attenuation correction
ature (25C) for 5 min while gently mixing prior to centri-
was performed using the NeuroAC calculated attenuation
fugation at 10,000 rpm for 5 min. The resulting supernatant
correction method.26 Using MiM version 6.9.7, (MiM
was transferred to a new tube with ExoQuick to precipitate
Software, Inc., Cleveland, OH, USA), reconstructed
EVs for 30–60 min at 4C. Then, vials were centrifuged at
and attenuation corrected images underwent motion
1500g for 30 min. The resulting EV pellet at the bottom of
correction, and static images binned at 20–90 min were
the tube was resuspended in 500 lL of PBS.
generated using SumIP.
BDEVs were enriched in aqueous solution by precipi-
tating with neuronal-specific antibodies SNAP25, PSA-
NCAM, and CD171 from total circulating EVs from MRI. MRI neuroimaging was employed and the proto-
the plasma. Exosomal and plasma levels of tau, neurofila- col included three-dimensional (3D) T1-weighted imag-
ment light chain (NfL), and glial fibrillary acidic protein ing (isotropic 1.0-mm spatial resolution), diffusion
(GFAP) were analyzed using a high-sensitivity Simoa tensor imaging (DTI; (isotropic 1.5-mm spatial resolu-
HD-X analyzer (Quanterix, Lexington, MA, USA) and tion, multi-shell with b-values of 1500 and 3000 s/mm2;
a paramagnetic bead-based enzyme-linked immunosor- 98 directions), resting state blood oxygen dependent
bent assay (ELISA) according to the manufacturer’s pro- (BOLD) imaging (10-min resting state acquisition, iso-
tocol. Coefficient of variation (intra- and inter-plate) tropic 2-mm spatial resolution, and 800-msec temporal
values were below 15% for all analytes. Coreplex cyto- resolution), susceptibility weighted imaging (3D gradient
kines panel of IL-6, IL-10, and TNFa were analyzed echo), and 3D arterial spin labeling (ASL; gradient-and
using an SP-X Imaging and Analysis System (Quanterix). spin-echo [GRASE] acquisition module and pseudo-
Because the array volumes are approximately 2 billion continuous blood tagging). In addition, calibration/
times smaller than an ELISA, a rapid buildup of fluores- correction data were acquired to characterize spatial dis-
cent product is generated if labeled protein is present, and tortions in echo-planar-based acquisitions (BOLD and
it provides detection ability between 100 and 1000 times diffusion) and static field inhomogeneity. All data
that of ELISA methods. With diffusion defeated, this were collected on a state-of-the-art 3 Tesla MR scanner
high local concentration of product can be readily obs- (Prisma, Siemens Healthineers, Erlangen, Germany)
erved. Samples were distributed randomly across plates using a 64-channel head radiofrequency coil. The total
and all assays were run in duplicate. acquisition time was roughly 1.5 h.

Radiotracer synthesis Data processing

[18F]DPA-714 was synthesized in accordance with U.S. Processing and quantification of the previously de-
Pharmacopeia (USP)<823> guidelines by the University scribed imaging data employed multiple packages in-
of Virginia Radiochemistry Core, as previously descri- cluding MRtrix3 and the various libraries available
bed.25 [18F]DPA-714 was synthesized in a 13.5 – 3.04% within the ANTsX software ecosystem (Fig. 1).
radiochemical yield (EOB) with a molar activity of Derived imaging data used in this study included corti-
34.3 – 9.9 Ci/lmol (1258 – 366 GBq/lol; n = 3) in a cal thickness maps from T1-weighted MRI, diffusivity-
total synthesis time of 90 min. In all cases, radiochemical based scalar images (e.g., fractional anisotropy) from
purity was >99%. diffusion-weighted MRI, functional scalar activation
946 STONE ET AL.

FIG. 1. Neuroimage processing diagram for this study. Following image acquisition, computational
processing generates several derived images per subject, which are registered to the study template via
each subject’s T1-weighted MRI. These aligned images are organized per modality, which are reduced in
terms of dimensionality to representative bases using the SiMLR framework. CT, computed tomography; FA,
fractional anisotropy; fALFF, fractional amplitude of low-frequency fluctuation; MRI, magnetic resonance
imaging; PET, positron emission tomography; SiMLR, Similarity-driven Multi-view Linear Reconstruction.

summarized via amplitude of low-frequency fluctuation Cortical thickness. ANTsX employs a registration-
(ALFF), and network correlation maps. Summary per- based framework for estimating cortical thickness35 that
fusion and PET scalar images were also generated for has been tailored and evaluated for both cross-sectional
each subject. ANTs tools27 were used to register each and longitudinal36 MR data. More recently, a deep-
T1-weighted image to an average population template learning-based framework has been developed and eval-
as used in previous studies.28 Subsequently, each uated37 demonstrating both superior measurement quality
subject-specific scalar image was aligned to the corre- and increased computational efficiency. Briefly, process-
sponding T1-weighted image, thus providing a set of ing includes brain extraction, brain parcellation based on
transformations to warp each image to the common six tissue types (i.e., cerebrospinal fluid [CSF], gray mat-
template space.27 ter, white matter, deep gray matter, brainstem, and cere-
Details concerning modality-specific processing are as bellum), and application of the diffeomorphic registration
follows. based cortical thickness (DiReCT) cortical thickness
algorithm.35 This yields a scalar image with non-zero val-
Diffusion-weighted imaging (DWI). The standard ues in the cortical gray matter providing a voxelwise
MRtrix3 pipeline29 was used for diffusion tensor recon- estimate of thickness (in millimeters).
struction and subsequent generation of diffusivity maps,
specifically those corresponding to fractional anisotropy Resting-state fMRI. Resting-state functional MRI
(FA) and radial diffusivity (RD). Pre-processing steps (fMRI) processing, performed in the R-based ANTs
included denoising,30 whole-brain extraction,31 and ten- interface (ANTsR), has been described previously.8,38
sor reconstruction.32 The average diffusion-weighted Briefly, motion correction was applied to each time
MR image was normalized to the corresponding T1- series. Nuisance parameters included the transforma-
weighted MR image. Motion and eddy current correction tion parameters, the framewise displacement,39 and
were incorporated into the pipeline using the dwifslpre- component-based method (CompCor) contributions.40
proc MRtrix3 program, which interfaces with the under- Summary images from the corrected time series inclu-
lying FMRIB Software Library commands eddy, topup, ded fractional amplitude of low-frequency fluctuation
and applytopup.33,34 (fALFF).41,42
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 947

Cerebral blood flow (CBF) from first-pass perfusion clustering data prior to statistical testing and adjustment
imaging. Similar to the other image time series, the for multiple comparisons, analogous to principal compo-
mean spatial image was generated and used as the refer- nents regression.54
ence image for motion correction. CBF images were gen- SiMLR’s default setting identifies low-dimensional
erated using a singular value decomposition technique43 embeddings that optimize the joint predictive power
with the arterial input function generated automatically.44 between all modalities equally. We instead use a path
Visual inspection of the CBF images revealed non- modeling approach46,55 where SiMLR optimizes the pre-
biological artifacts that persisted after standard confound dictive power from PET to the other modalities and from
modeling and processing. Consequently, CBF was exclu- each MRI modality to PET. This strategy guarantees that
ded from the statistical analysis. the PET neuroinflammation modality drives the nature
of the learned features across all modalities.
PET. PET images were also motion-corrected using Formally, through SiMLR optimization, the ith modal-
ANTsX tools. The toolbox for partial volume correction ity, Xi, is represented by a pair of matrices ðUi , Vi Þ where
(PVC) in PET (PETPVC),45 which provides several pub- the columns of Ui are the low-dimensional bases span-
lished methods for performing PVC, was used for addi- ning Xi and the columns of Vi are the regularized, sparse
tional pre-processing. Finally, after inspecting several representation of each modality component. As men-
anatomical regions and the range of PET activity, values tioned previously, this permits principal component reg-
were normalized for each subject based on the corre- ressions of the form:
sponding PET values in the left and right caudate.
Ui, l ~d0 þ d1 þ . . . þ dn
These segmented regions were identified using the corre-
sponding T1-weighted image and the Desikan-Killiany- where the set of dj represents the explanatory variables
Tourville labeling protocol available in ANTsXNet.37 (e.g., age, gender, and brain volume) and l denotes the
specific basis vector of the ith solution.
Statistical analysis To explore the relationship between GBEV, control, or
The multifaceted neurological effects stemming from exposed status (denoted as armt below) and the imaging
blast exposure necessitate multiple observations/ modalities, the following quasi-Poisson regression model
measurements to characterize the associated structural, was used:56–58
functional, and metabolic sequelae. In this study, the g ðGBEV Þ~ba Age þ bc otherCon þ bt armt þ bml Umodality, l
battery of measurements included multi-modal neuroi-
maging, self-reported or clinical neuropsychological where each b is a coefficient, g is the log link function,56
assessment, and blood biomarkers. Such measurements Age is subject age, and otherCon represents the number
can be inherently noisy and incomplete while simul- of reported concussions from non-blast-related events.
taneously mutually overlapping and complementary. Quasi-Poisson regression models are appropriate for
A recently developed statistical framework, known as count or frequency data such as GBEV that is often
similarity-driven multi-view linear reconstruction non-Gaussian/overdispersed.58 The modality is one of
(SiMLR),46 has been employed for exploring and analyz- PET, CT cortical thickness from T1-weighted MRI, FA
ing such data. Successful application includes a recent (fractional anisotropy from DTI), fALFF (resting state),
investigation into a career breacher cohort characterized or Jacobian (a measure of localized volumetric change
by repetive low-level blast exposure8 where the use of relative to a study template).59
SiMLR permitted identification of significant group
effects spanning multiple modalities, including those Critical GBEV threshold analysis
mentioned previously. The above equation allows modeling the non-linear rela-
Briefly, SiMLR is a multi-view extension of earlier tionship between imaging variables and GBEV while
techniques (i.e., single-view eigenanatomy47,48 and dual- controlling for baseline exposure level (armt ), age, and
view sparse canonical correlation analysis [SCCAN]).49,50 concussions. The discrete control/exposed variable armt
Each of these techniques is rooted in principal component defines a split of the cohort based on a GBEV threshold
analysis (PCA) with additional spatial constraints specif- t where t varies between 9875.6 and 177,424 at discrete
ically tailored for neuroimaging data.46 Akin to PCA, points determined by the values of our cohort. The
these techniques are used for principled data dimension- threshold 65,309 corresponds to the original operator
ality reduction—an important consideration in the con- (n = 9) versus non-operator (n = 9) grouping. The regres-
text of neuroimaging where a single image can contain sion parameter bml on the term Umodality, l is our key out-
several orders of magnitude of statistically dependent come of interest for statistical testing; it represents the
data. In this way, and in contrast to standard mass voxel- degree to which the relationship between the modality’s
wise univariate51 or region of interest (ROI)-based lth feature embedding and the total GBEV score accel-
approaches,52,53 statistical power is conserved by erates in exposed subjects. In particular, we test for
948 STONE ET AL.

significant relationships between imaging variables (pri- Control: 0 – 0, p < 0.5). Additionally, non-blast-related
mary outcomes from PET with secondary outcomes from concussion history differed between the two groups
MRI) and GBEV after controlling for the other covariates. (Exposed: 3.9 – 1.6, Control: 1.1 – 1.4, p < 0.05). To con-
We assess significance of the bml coefficients via permuta- trol for non-blast-related concussion, this factor was
tion testing where the GBEV score is permuted 50,000 included in the quasi-Poisson analysis, the results of
times per model. This empirical approach to p-value calcu- which are outlined below.
lation is generally more conservative than model-based
methods and leads to a minimum p-value of 2e-05. Neuroimaging relationships to GBEV outcomes
The critical threshold analysis involved 200 tests display-
Analysis of biomarker ing the relationship between imaging and GBEV: five
and cognitive measurements components for each of five modalities were tested across
We adopt the same quasi-Poisson model as described eight threshold levels. We corrected the p-values associ-
above but instead use BDEVs or cognitive measures as ated with each bml using both Bonferroni (more conserva-
predictors instead of SiMLR imaging components. In tive) and false discovery rate correction but focused on
addition, we analyze these relationships only at a single Bonferroni corrected results at the p < 0.05 level. The
threshold of ‡65,309, which is consistent with the origi- results of this analysis are summarized in Figure 2 with
nal control versus exposed grouping. individual modality-specific features shown in Figure 3
for the most prominent PET feature, Figure 4 for the
Results most prominent cortical thickness feature, and Figure 5
Ten USSOCOM SOF SMs and 11 military control SMs for the most prominent Jacobian feature. Note that in
were recruited into the current study. One subject from Figures 3–6, The values are the feature weights on the
the exposed group was excluded due to demonstrating brain for a given component; regions with weights closer
low-affinity binding TSPO genotype. One subject from to 1 contribute more strongly to the derived weighted
the control group was also excluded based on a history average (e.g., of neuroinflammation). Regions without
of moderate to severe TBI revealed through the head in- color do not contribute at all. Overall, eight features sur-
jury questionnaire, the results of which met the criteria vived this stringent p-value threshold spanning corti-
for prior severe TBI. An additional control subject with- cal thickness, neuroinflammation, and volumetric loss;
drew prior to completing the study procedures. The final these model relationships are shown in Figure 7. Recall-
cohort included 9 controls and 9 exposed subjects with ing that p-values were computed with permutation tests,
complete data. No significant difference was found be- these surviving predictors should reflect the most reliable
tween groups both in terms of age (Exposed: 41.9 years GBEV-associated imaging patterns within this cohort.
–5.8 and Control: 41.6 years –5.6, p = 0.9) or duration
of service (Exposed: 21.3 years –4.3, Control: 19.9 – 6.6, Evidence for joint effects of exposure
p = 0.6). Subjects were recruited from multiple branches across modalities
of the military (Exposed: Army = 6, Navy = 1, Air Force = 1, SiMLR yields sparse patterns defining each feature vec-
Marine Corps = 1; Control: Army = 4, Navy = 3, Air tor. These patterns are unsigned weighted averages over
Force = 2) (Table 1). Blast-related concussion history the original voxelwise data matrix, that is, over the under-
differed between the two groups (Exposed: 1.7 – 1.3, lying neuroanatomy (Supplementary Fig. S1). Here, we
demonstrate post hoc results that summarize the common
Table 1. Demographic Summary of the Blast Exposed anatomy involved in features derived from PET, cortical
and Control Cohorts thickness, and volume loss. A visualization of the joint
Exposed (n = 9) Control (n = 9) feature set is shown in Figure 6 as well as in Supplemen-
tary Figures S2–S5.
Age (years) 41.9 (range: 35–53) 41.6 (range: 36–51)
Handedness Right = 7; left = 2 Right = 7; left = 1;
ambidextrous = 1 BDEV and cognitive measurements
Service Army = 6 Army = 4
Navy = 1 Navy = 3
The framework for analyzing cognitive and BDEV
Air Force = 1 Air Force = 2 data is the same as that for imaging. However, we treat
Marine Corps = 1 these predictors’ relationship with GBEV independently,
Duration of 21.3 (range: 8–27) 13.8 (range: 16–29)
service (years) which results in a total of 23 tests: six BDEVs in addition
Highest level Doctorate = 1 Doctorate = 1 to the NSI total score, the PGWI total score, the PCL-5
of education Master’s = 3 Master’s = 5 total score, the PSQI total score, and 12 measurements
Bachelor’s = 3 Bachelor’s = 3
Associate = 1 from ANAM 4 TBI-MIL. Of these, only two BDEV mea-
High School = 1 surements survive Bonferroni correction. Figure 8 dem-
GBEV 4,888,072 (112,393–16,534,930) 19,990 (0–65,309)
onstrates the model relationships between GBEV and
GBEV, generalized blast exposure value. NfL and GFAP values.
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 949

FIG. 2. Threshold analysis for imaging predicting GBEV. This figure displays the t-statistic for each feature
along with the permutation tested and Bonferroni-corrected significance level. At the 65,309 GBEV level
(n = 9 exposed) we see an increase in the number of imaging measurements that significantly relate to
exposure and these effects increase further at each successive level (CT3 and JAC4 excepted, see Discussion
section). GBEV thresholds beyond 177,424 result in too few subjects for current models. Corrected
significance denoted* for p < 0.05, p < 0.01, * for p < 0.005 and *** for p < 0.001. CT, cortical thickness; FA,
fractional anisotropy; fALFF, fractional amplitude of low-frequency fluctuation; GBEV, generalized blast
exposure value; JAC, Jacobian network; PET, positron emission tomography; WM, white matter.

Discussion mation relevant to history of blast exposure, neuropsy-

The current investigation was performed in partnership chological assessments, serum BDEV biomarkers, and
with USSOCOM as a pilot study to explore the neurolog- neuroimaging inclusive of MRI and PET-CT. Based on
ical sequelae of repeated low-level blast exposure in SOF previous evidence in repeatedly blast-exposed personnel,
SMs. Previous studies in other populations have demon- special emphasis was placed in the current study on an
strated structural and functional alterations in repeatedly examination of the presence and extent of neuroinflam-
blast-exposed personnel such as breachers.8 As SOF SMs mation through the visualization of an intravenously
may be repeatedly exposed to blast overpressure during a injected radioisotope targeting neuroinflammation with
career, it is important to better understand potential cumu- PET-CT as well as neuroinflammatory assessments
lative sequelae that may develop resulting from these within serum and BDEVs. These acquired measures gen-
exposures. The current study involved the acquisition erated a highly dimensional data set in a relatively small
of an array of measures, inclusive of self-reported infor- cohort.

FIG. 3. PET network 5 involves the salience network including the anterior insula and middle and superior
frontal gyri. There is additional involvement of frontal pole and anterior temporal lobe. Signal is also
contributed from precuneus, angular, and supramarginal gyri. Lastly, visual regions contribute including the
cuneus and occipital pole. PET, positron emission tomography.
950 STONE ET AL.

FIG. 4. CT network 5. Primarily involves anterior temporal lobe regions including temporal pole, entorhinal
cortex, and inferior temporal gyrus with additional involvement of the anterior cingulate, medial frontal
cortex, and supramarginal gyrus. CT, cortical thickness.

The statistical model SiMLR was employed to identify participants after controlling for potential confounds.
low-dimensional embeddings that optimize the joint pre- However, within the exposed cohort, PET neuroinflam-
dictive power between all modalities equally and it is mation measurements increase significantly with higher
well-suited to handle the unique statistical challenges exposure levels. Additionally, MRI measurements dem-
posed by such a data set. This model identifies specific onstrate that increased blast exposure relates to reduc-
networks or features (e.g., cortical thickness network 5 tions in volume and thickness of brain structures. These
[CT5] or Jacobian network 4 [JAC4]) that are highly cor- findings, together, establish that GBEV may provide an
related with predictive measures of interest, such as blast index into underlying effects of blast overpressure.
exposure. This is the first work to show neuroimaging- In addition to examining relationships between GBEV
derived critical GBEV thresholds for blast exposure. and neurological end-point assessments, a critical thresh-
The findings of the current study associate a history of old analysis was performed to identify GBEV levels at
blast exposure over a career with measures of increased which alterations in meaningful biomarker assessments
neuroinflammation as well as alterations in brain struc- may be observed. The GBEV critical threshold analysis
ture in SOF SMs. Specifically, GBEV increases in reveals two types of effects: those that increase with
exposed subjects as a function of both PET- and threshold and those that are maximized within the exp-
BDEV-related neuroinflammation measurements. Neuro- osed group membership level, that is, CT3 and JAC4.
inflammation levels, on their own, do not differ substan- It is possible that results that are maximized only at the
tially between more highly and less highly exposed exact operator/control threshold may be a consequence

FIG. 5. Jacobian/local volume network ( JAC) 4 is dominated by cingulo-opercular task control network
and ventral attention network regions including supramarginal gyri, temporal pole, and frontal pole.
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 951

FIG. 6. Spatially coincident features across all three modalities that reach Bonferroni significance.
Neuroinflammation, reductions in thickness, and volume loss in these regions relate to increased GBEV
scores. These cross-modality regions span the salience, task control, and default mode networks. GBEV,
generalized blast exposure value.

of other operator-related effects in conjunction with neu- tabolism in the medial temporal lobe.65 Also, veterans
roinflammation and/or blast exposure. If this is the case, with blast-related TBI demonstrated magnetic resonance
it may suggest that the patterns captured by PET1, spectroscopy (MRS) evidence of decreased N-acetyl
PET5, CT1, CT5, JAC3, and JAC5 are most specific to aspartate (NAA) to choline and NAA to creatine ratios
blast exposure effects. Of these, PET5 and JAC3 repre- within the hippocampus.69 Additionally, the hippocam-
sent the statistically most robust features. pus showed selective vulnerability to blast in an estab-
The neuroanatomical patterns revealed in this analy- lished mouse model.70 The consistent representation of
sis may provide insight into potentially sensitive brain anterior temporal lobe (ATL) in the results of the cur-
networks and/or mechanisms of injury. Both cerebellum rent study may be an indirect consequence of its net-
and medial temporal structures are central to these pat- work connectivity. ATL includes several densely
terns and these regions have been previously associated connected regions critical to semantic memory and the
with blast injury in pre-clinical models60–62 inclusive of hierarchical organization of category concepts.71,72 Dis-
swine63 and non-human primates64 and in military SMs ruption of these networks is associated with cogni-
as well.65,66 The middle frontal and superior frontal tive challenges of memory, language, and executive
gyrus (MFG, SFG) are also represented across modali- functioning.73
ties. Indeed, pattern FA4 shows large weights in the Volume reductions in cerebellum and brainstem are
SFG at the FDR-corrected significance level, suggesting among the most heavily weighted structures within the
a more marginal, but still notable, relationship of GBEV local volume patterns and are linearly associated with
with white matter integrity. Post-mortem studies of GBEV scores. The cerebellum is connected to many of
blast-exposed personnel have identified axonal injury the regions that are prominent in our analysis of GBEV
within the MFG and SFG brain regions.67 Additionally, including medial temporal anatomy and, through the thal-
in vivo MRI has revealed blast-related gray matter vol- amus, to cingulate cortex and sensorimotor cortex. These
ume loss in the SFG.68 These frontal regions play an cerebrocerebellar connections are important contributors
important role in executive functioning, specifically to executive function and emotional regulation74 in addi-
working memory. The current work suggests that neuro- tion to motor skills.75
inflammation may play a role in mediating structural It remains unclear whether cerebellar, temporal lobe,
alterations within these brain regions. or frontal lobe injury occur concurrently or via a cascade
Limbic and temporal lobe regions are also implicated of effects, or both. Future work is needed in larger cohorts
across three imaging modalities, inclusive of neuroin- to replicate and/or validate the sensitivity and/or prece-
flammation, cortical thickness, and local volume. These dence of blast effects on these circuitries. Although FA
findings are consistent with other studies involving exam- was used as the primary DTI-related outcome in this
ination of these brain regions, including a study of Iraq research, free water fraction may also be relevant to neu-
war veterans with repeated blast exposure and post- roinflammation76; this alternative approach may be pur-
concussive symptoms, demonstrating cerebral hypome- sued in future work.
952 STONE ET AL.

FIG. 7. Neuroimaging relationships with GBEV score for both controls and exposed defined at two
thresholds. Only predictors that survive Bonferroni correction at one of the two thresholds are shown.
Dotted red lines indicate 95% confidence intervals around the best fit solid line. CT, cortical thickness; GBEV,
generalized blast exposure value; JAC, Jacobian network; PET, positron emission tomography.

We also cannot, in this cohort, disentangle specific Cognitive measurements do not reveal direct rela-
effects of blast overpressure from those of concussion due tionships with GBEV. Larger sample sizes, different
to the high correlation between GBEV and blast-related statistical modeling approaches, or alternative neuro-
concussion experiences. However, we partially mitigate psychological testing paradigms may be needed. NfL
this issue by controlling for concussions due to non-blast- and GFAP BDEVs relate significantly (Bonferroni-
related events. We also cannot rule out the contribution corrected p < 0.05) to GBEV via quasi-Poisson model-
of behavioral or experiential confounds to the brain dif- ing. NfL is relatively non-specific and this is reflected
ferences observed here.67 Such confounds may explain in its moderate association with concussion (uncor-
the relatively minimal difference in neuroinflammation rected p < 0.001). GFAP has previously been associated
between the two groups in the original arms of the study. with blast exposure, although a recent study indicated
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 953

FIG. 8. BDEV measures of neuroinflammation related to GBEV. BDEV, brain-derived extracellular vesicle;
EV, extracellular vesicle; GBEV, generalized blast exposure value; GFAP, glial fibrillary acidic protein; NFL,
neurofilament light chain.

reduced levels with more acute exposure77—the oppo- Transparency, Rigor,

site of the association found here. and Reproducibility Summary
SiMLR improves detection power by clustering neuro- The analysis plan was not formally pre-registered, but the
anatomical data into spatial patterns before hypothesis lead author certifies that the analysis plan was pre-specified.
testing. The disadvantage of this procedure is that we The study was planned as an initial feasibility pilot study.
do not directly test the roles of specific anatomy within As such, a formal power analysis was not performed at
the pattern. Rather, we infer the relative contribution by the outset of the investigation. Twenty-one subjects were
investigating the weight values. This type of interpreta- recruited for the study, and data were successfully analyzed
tion must be approached with caution as the statistical in 18 subjects. Participants were told the results of clini-
tests only apply to the results of the whole pattern. cally significant or potentially clinically significant imag-
In summary, the current investigation demonstrated ing findings after the final clinical observations had been
that increased blast exposure was associated with mea- made. Imaging acquisition and analyses were performed
sures of increased neuroinflammation along with changes by team members blinded to relevant characteristics of
in brain structure in SOF SMs over a career. These obser- the participants, and clinical outcomes were assessed by
vations survive Bonferroni multiple-comparison correc- team members blinded to imaging results. MRI data
tion as well as control for potential confounds such as were acquired using a Siemens Prisma 3T scanner, soft-
non-blast-related TBI. Additionally, critical threshold ware version VE11C. All MRI data were acquired on
analyses demonstrate specific GBEV levels at which the same scanner. All equipment and software used to
an association between increased blast exposure and perform imaging and pre-processing are widely available
alterations in neurological end-point assessments are from commercial sources. The key inclusion criteria and
observed. outcome evaluations are established standards.
The current investigation is a limited pilot study and Neuroimaging processing and statistical analyses were
caution should be exercised in interpretation of the find- performed by Brian B. Avants and Nicholas J. Tustison
ings of the study. Future hypothesis-driven research using the open-source ANTsR package, of which they
should be performed to determine the validity and overall are the founders and principal developers. Corrections
reproducibility of these observed findings. Additionally, for multiple comparisons were performed using false
future studies with increased numbers of subjects should discovery rate available in the R statistical package. No
allow for a more complete critical threshold assessment. replication or external validation studies have been per-
This final point is of importance in identifying and refin- formed or are planned/ongoing at this time to our knowl-
ing important dose-response relationships between blast edge. Analytic code used to conduct the analyses are
overpressure exposure and potential injury metrics. available in (GitHub).
Such thresholds may be of key importance in operational
decision-making as well as the design of protective Acknowledgments
equipment to help mitigate the effects of blast exposure We would like to thank Col. (Dr.) Eveline F. Yao for
at the individual level. her support while at SOCOM for the study as well as
954 STONE ET AL.

Kelsey Radmanesh for her support with the regulatory Funding Information
aspects of the study. Additionally, we would like to This work was supported/funded by the Joint Pro-
thank Julie Matsumoto, MD, for her contributions per- gram Committee-5 Development of Exposure Stan-
forming screening neuroimaging evaluations for the dards to Repeated Blast Exposure program, work unit
study as well as Jamie Weathersbee, RT (R) (CT) #603115HP-3730-001-A1118.
(MR), and Jose Reyes, CNMT, RT (CT) (MR), for
their technical contributions in acquiring the imaging Author Disclosure Statement
for the study. No competing financial interests exist.
Preliminary results of this work were previously pub-
lished in Abstracts from the 39th Annual Symposium Supplementary Material
of the National Neurotrauma Society, including the Supplementary Figure S1
Supplementary Figure S2
AANS/CNS Joint Section on Neurotrauma and Critical Supplementary Figure S3
Care, June 26–29, 2022. Supplementary Figure S4
Supplementary Figure S5
The views expressed in this article reflect the results
of research conducted by the authors and do not neces- References
sarily reflect the official policy or position of the Depart- 1. Belding JN, Kolaja CA, Rull RP, et al. Single and repeated high-level
ment of the Navy, Department of Defense, nor the U.S. blast, low-level blast, and new-onset self-reported health condi-
tions in the U.S. Millennium Cohort Study: an exploratory inves-
government. tigation. Front Neurol 2023;14:1110717; doi: 10.3389/fneur.2023
Some of the authors are military SMs or federal/ .1110717
contracted employees of the U.S. government. This 2. Carr W, Stone JR, Walilko T, et al. Repeated low-level blast exposure: a
descriptive human subjects study. Mil Med 2016 181(5 Suppl):28–39;
work was prepared as part of their official duties. Title doi: 10.7205/MILMED-D-15-00137
17 U.S.C. 105 provides that ‘‘copyright protection 3. Carr W, Polejaeva E, Grome A, et al. Relation of repeated low-level blast
exposure with symptomology similar to concussion. J Head Trauma
under this title is not available for any work of the U.S. Rehabil 2015;30(1):47–55; doi: 10.1097/HTR.0000000000000064
government.’’ Title 17 U.S.C. 101 defines a U.S. gov- 4. Belding JN, Englert R, Bonkowski J, et al. Occupational risk of low-level
ernment work as work prepared by a military SM or blast exposure and TBI-related medical diagnoses: a population-based
epidemiological investigation (2005-2015). Int J Environ Res Public
employee of the U.S. government as part of that person’s Health 2021;18(24):12925; doi: 10.3390/ijerph182412925
official duties. 5. Carr W, Kelley AL, Toolin CF, et al. Association of MOS-based blast expo-
sure with medical outcomes. Front Neurol 2020;11:619; doi: 10.3389/
fneur.2020.00619
Authors’ Contributions 6. Modica CM, Johnson BR, Zalewski C, et al. hearing loss and irritability
The authors contributed as follows. James R. Stone: con- reporting without vestibular differences in explosive breaching profes-
sionals. Front Neurol 2020;11:588377; doi: 10.3389/fneur.2020.588377
ceptualization, methodology, formal analysis, investiga- 7. Haran FJ, Zampieri C, Wassermann EM, et al. Chronic effects of breaching
tion, data curation, writing–original draft, supervision, blast exposure on sensory organization and postural limits of stability.
J Occup Environ Med 2021;63(11):944–950; doi: 10.1097/JOM
project administration, funding acquisition; Brian B. .0000000000002266
Avants: conceptualization, methodology, software, for- 8. Stone JR, Avants BB, Tustison NJ, et al. Functional and structural neuroi-
mal analysis, data curation, writing–original draft, visu- maging correlates of repetitive low-level blast exposure in career
breachers. J Neurotrauma 2020;37(23):2468–2481; doi: 10.1089/neu
alization; Nicholas J. Tustison: conceptualization, .2020.7141
methodology, software, formal analysis, data curation, 9. Edwards KA, Leete JJ, Smith EG, et al. elevations in tumor necrosis factor
writing–original draft, visualization; Jessica Gill: con- alpha and interleukin 6 from neuronal-derived extracellular vesicles in
repeated low-level blast exposed personnel. Front Neurol 2022;13:
ceptualization, methodology, investigation, resources, 723923; doi: 10.3389/fneur.2022.723923
writing–review and editing; Elisabeth A. Wilde: concep- 10. Vorn R, Edwards KA, Hentig J, et al. A pilot study of whole-blood tran-
scriptomic analysis to identify genes associated with repetitive low-
tualization, methodology, supervision, writing–review level blast exposure in career breachers. Biomedicines 2022;10(3):690;
and editing; Kiel D. Neumann: conceptualization, meth- doi: 10.3390/biomedicines10030690
odology, investigation, writing–review and editing; 11. Ramlackhansingh AF, Brooks DJ, Greenwood RJ, et al. Inflammation after
trauma: microglial activation and traumatic brain injury. Ann Neurol
Leslie A. Gladney: investigation, writing–review and 2011;70(3):374–383; doi: 10.1002/ana.22455
editing; Madison O. Kilgore: formal analysis, writing– 12. DiSabato DJ, Quan N, Godbout JP. Neuroinflammation: the devil is in the
details. J Neurochem 2016;139 Suppl 2(Suppl 2):136–153; doi: 10.1111/
review and editing; F. Bowling: writing–review and jnc.13607
editing; Christopher M. Wilson: writing–review and 13. Witcher KG, Eiferman DS, Godbout JP. Priming the inflammatory pump of
editing; John F. Detro: writing–review and editing; the CNS after traumatic brain injury. Trends Neurosci 2015;38(10):609–
620; doi: 10.1016/j.tins.2015.08.002
Heather G. Belanger: writing–review and editing; 14. Bombardier CH, Fann JR, Temkin NR, et al. Rates of major depressive
Katryna Deary: investigation, writing–review and edit- disorder and clinical outcomes following traumatic brain injury. JAMA
2010;303(19):1938–1945; doi: 10.1001/jama.2010.599
ing; Hans Linsenbardt: formal analysis, writing–review 15. Kwon HS, Koh SH. Neuroinflammation in neurodegenerative disorders:
and editing; Stephen T. Ahlers: conceptualization, the roles of microglia and astrocytes. Transl Neurodegener 2020;9(1):42;
methodology, formal analysis, investigation, writing– doi: 10.1186/s40035-020-00221-2
16. Delic V, Beck KD, Pang KCH, et al. Biological links between traumatic brain
original draft, supervision, project administration, fund- injury and Parkinson’s disease. Acta Neuropathol Commun 2020;8(1):
ing acquisition. 45; doi: 10.1186/s40478-020-00924-7
REPEATED BLAST EXPOSURE IN SPECIAL OPERATORS 955

17. Schimmel SJ, Acosta S, Lozano D. Neuroinflammation in traumatic brain 38. Avants BB, Hutchison RM, Mikulskis A, et al. Amyloid beta-positive sub-
injury: a chronic response to an acute injury. Brain Circ 2017;3(3):135– jects exhibit longitudinal network-specific reductions in spontaneous
142; doi: 10.4103/bc.bc_18_17. Epub 2017 Oct 12. brain activity. Neurobiol Aging 2019;74:191–201; doi: 10.1016/j
18. Coughlin JM, Wang Y, Munro CA, et al. Neuroinflammation and brain .neurobiolaging.2018.10.002
atrophy in former NFL players: an in vivo multimodal imaging pilot 39. Power JD, Barnes KA, Snyder AZ, et al. Spurious but systematic correlations
study. Neurobiol Dis 2015;74:58–65; doi: 10.1016/j.nbd.2014.10.019 in functional connectivity MRI networks arise from subject motion. Neu-
19. Neumann KD, Seshadri V, Thompson XD, et al. Microglial activation per- roimage 2012;59(3):2142–2154; doi:10.1016/j.neuroimage.2011.10.018
sists beyond clinical recovery following sport concussion in collegiate 40. Behzadi Y, Restom K, Liau J, et al. A component based noise correction
athletes. Front Neurol 2023;14:1127708; doi: 10.3389/fneur.2023 method (CompCor) for BOLD and perfusion based fMRI. Neuroimage
.1127708 2007;37(1):90–101; doi: 10.1016/j.neuroimage.2007.04.042
20. Marklund N, Vedung F, Lubberink M, et al. Tau aggregation and increased 41. Zou QH, Zhu CZ, Yang Y, et al. An improved approach to detection of
neuroinflammation in athletes after sports-related concussions and in amplitude of low-frequency fluctuation (ALFF) for resting-state fMRI:
traumatic brain injury patients—a PET/MR study. Neuroimage Clin fractional ALFF. J Neurosci Methods 2008;172(1):137–141; doi: 10.1016/j
2021;30:102665; doi: 10.1016/j.nicl.2021.102665 .jneumeth.2008.04.012
21. Folkersma H, Boellaard R, Yaqub M, et al. Widespread and prolonged 42. Power JD, Cohen AL, Nelson SM, et al. Functional network organization of
increase in (R)-(11)C-PK11195 binding after traumatic brain injury. the human brain. Neuron 2011;72(4):665–678; doi: 10.1016/j.neuron
J Nucl Med 2011;52(8):1235–1239; doi: 10.2967/jnumed.110.084061 .2011.09.006
22. Modica LCM, Egnoto MJ, Statz JK et al. Development of a blast exposure 43. Ostergaard L, Sorensen AG, Kwong KK, et al. High resolution measure-
estimator from a Department of Defense-wide survey study on military ment of cerebral blood flow using intravascular tracer bolus passages.
service members. J Neurotrauma 2021;38(12):1654–1661; doi:10.1089/ Part II: experimental comparison and preliminary results. Magn Reson
neu.2020.7405 Med 1996;36(5):726–736; doi: 10.1002/mrm.1910360511
23. Montenigro PH, Alosco ML, Martin BM, et al. Cumulative head impact 44. Mlynash M, Eyngorn I, Bammer R, et al. Automated method for generat-
exposure predicts later-life depression, apathy, executive dysfunction, ing the arterial input function on perfusion-weighted MR imaging:
and cognitive impairment in former high school and college football validation in patients with stroke. AJNR Am J Neuroradiol 2005;26(6):
players. J Neurotrauma 2017;34(2):328–340; doi: 10.1089/neu.2016 1479–1486.
.4413 45. Thomas BA, Cuplov V, Bousse A, et al. PETPVC: a toolbox for performing
24. Kenney K, Qu BX, Lai C, et al. Higher exosomal phosphorylated tau and partial volume correction techniques in positron emission tomography.
total tau among veterans with combat-related repetitive chronic mild Phys Med Biol 2016;61(22):7975–7993; doi: 10.1088/0031-9155/61/22/7975
traumatic brain injury. Brain Inj 2018;32(10):1276–1284; doi: 10.1080/ 46. Avants BB, Tustison NJ, Stone JR. Similarity-driven multi-view embed-
02699052.2018.1483530 dings from high-dimensional biomedical data. Nat Comput Sci 2021;
25. McCauley KS, Wilde JH, Bufalino SM, et al. An automated radiosynthesis of 1(2):143–152; doi: 10.1038/s43588-021-00029-8
[18F]DPA-714 on a commercially available radiosynthesizer, Elixys 47. Dhillon PS, Wolk DA, Das SR, et al. Subject-specific functional parcellation
Flex/Chem. Appl Radiat Isot 2022;180:110032; doi: 10.1016/j.apradiso via prior based eigenanatomy. Neuroimage 2014;99:14–27; doi: 10
.2021.110032 .1016/j.neuroimage.2014.05.026
26. Peyrat J-M, Joshi A, Mintun M, et al. An automatic method for the 48. Cook PA, McMillan CT, Avants BB, et al. Relating brain anatomy and
quantification of uptake with florbetapir imaging. J Nucl Med 2012; cognitive ability using a multivariate multimodal framework. Neuro-
53:210–210. image 2014;99:477–486; doi: 10.1016/j.neuroimage.2014.05.008
27. Avants BB, Tustison NJ, Song G, et al. A reproducible evaluation of ANTs 49. Avants BB, Libon DJ, Rascovsky K, et al. Sparse canonical correlation
similarity metric performance in brain image registration. Neuroimage analysis relates network-level atrophy to multivariate cognitive mea-
2011;54(3):2033–2044; doi: 10.1016/j.neuroimage.2010.09.025 sures in a neurodegenerative population. Neuroimage 2014;84:698–
28. Tustison NJ, Cook PA, Klein A, et al. Large-scale evaluation of ANTs and 711; doi: 10.1016/j.neuroimage.2013.09.048
FreeSurfer cortical thickness measurements. Neuroimage 2014;99:166– 50. McMillan CT, Irwin DJ, Avants BB, et al. White matter imaging helps dis-
179; doi: 10.1016/j.neuroimage.2014.05.044 sociate tau from TDP-43 in frontotemporal lobar degeneration. J Neurol
29. Tournier JD, Smith R, Raffelt D, et al. MRtrix3: a fast, flexible and open Neurosurg Psychiatry 2013;84(9):949–955; doi: 10.1136/jnnp-2012-
software framework for medical image processing and visualisation. 304418
Neuroimage 2019;202:116137; doi: 10.1016/j.neuroimage.2019.116137 51. Ashburner J, Friston KJ. Voxel-based morphometry—the methods.
30. Veraart J, Novikov DS, Christiaens D, et al. Denoising of diffusion MRI Neuroimage 2000;11(6 Pt 1):805–821; doi: 10.1006/nimg.2000.0582
using random matrix theory. Neuroimage 2016;142:394–406; doi: 10 52. Mac Donald CL, Johnson AM, Cooper D, et al. Detection of blast-related
.1016/j.neuroimage.2016.08.016 traumatic brain injury in U.S. military personnel. N Engl J Med 2011;
31. Dhollander T, Raffelt D, Connelly A. Unsupervised 3-tissue response 364(22):2091–2100; doi: 10.1056/NEJMoa1008069
function estimation from single-shell or multi-shell diffusion MR data 53. Smith SM, Jenkinson M, Johansen-Berg H, et al. Tract-based spatial sta-
without a co-registered T1 image. Conference paper presented at the tistics: voxelwise analysis of multi-subject diffusion data. Neuroimage
ISMRM Workshop on Breaking the Barriers of Diffusion MRI, Lisbon, 2006;31(4):1487–1505; doi: 10.1016/j.neuroimage.2006.02.024
Portugal, 2016. 54. Jolliffe IT. A note on the use of principal components in regression. J Royal
32. Veraart J, Sijbers J, Sunaert S, et al. Weighted linear least squares esti- Stat Soc Series C (Applied Statistics) 1982;31:300–303.
mation of diffusion MRI parameters: strengths, limitations, and pitfalls. 55. Hanafi M. PLS path modelling: computation of latent variables with the
Neuroimage 2013;81:335–346; doi: 10.1016/j.neuroimage.2013.05.028 estimation mode b. Comput Stat 2007;22:275–292.
33. Andersson JL, Skare S, Ashburner J. How to correct susceptibility distor- 56. Ver Hoef JM, Boveng PL. Quasi-Poisson vs. negative binomial regression:
tions in spin-echo echo-planar images: application to diffusion tensor how should we model overdispersed count data? Ecology 2007;88(11):
imaging. Neuroimage 2003;2:870–888; doi: 10.1016/S1053- 2766–2772; doi: 10.1890/07-0043.1
8119(03)00336-7 57. O’Hara RB, Kotze DJ. Do not log-transform count data. Methods Ecol
34. Andersson JLR, Graham MS, Drobnjak I, et al. Towards a comprehensive Evolution 2010;1:118–122.
framework for movement and distortion correction of diffusion MR 58. Warton DI, Lyons M, Stoklosa J, et al. Three points to consider when
images: within volume movement. Neuroimage 2017;152:450–466; doi: choosing a LM or GLM test for count data. Methods Ecol Evolution
10.1016/j.neuroimage.2017.02.085 2016;7:882–890.
35. Das SR, Avants BB, Grossman M, et al. Registration based cortical thick- 59. Kim J, Avants B, Patel S, et al. Structural consequences of diffuse traumatic
ness measurement. Neuroimage 2009;45(3):867–879; doi: 10.1016/j brain injury: a large deformation tensor-based morphometry study.
.neuroimage.2008.12.016 Neuroimage 2008;39(3):1014–1026; doi: 10.1016/j.neuroimage.2007.10
36. Tustison NJ, Holbrook AJ, Avants BB, et al; Alzheimer’s Disease .005
Neuroimaging Initiative. Longitudinal mapping of cortical thickness 60. Kozlova Y, Kozlov S. Changes of trace elements in the cerebellum and
measurements: an alzheimer’s disease neuroimaging initiative-based their influence on the rats behavior in elevated plus maze in the acute
evaluation study. J Alzheimers Dis 2019;71:165–183; doi: 10.3233/JAD- period of mild blast-induced brain injury. J Trace Elem Med Biol
190283 2023;78:127189; doi: 10.1016/j.jtemb.2023.127189
37. Tustison NJ, Cook PA, Holbrook AJ, et al. The ANTsX ecosystem for 61. Bishop R, Won SJ, Irvine KA, et al. Blast-induced axonal degeneration in
quantitative biological and medical imaging. Sci Rep 2021;11(1):9068; the rat cerebellum in the absence of head movement. Sci Rep 2022;
doi: 10.1038/s41598-021-87564-6 12(1):143; doi: 10.1038/s41598-021-03744-4
956 STONE ET AL.

62. Ordek G, Asan AS, Cetinkaya E, et al. electrophysiological correlates of 70. Sajja VS, Ereifej ES, VandeVord PJ. Hippocampal vulnerability and sub-
blast-wave induced cerebellar injury. Sci Rep 2018;8(1):13633; doi: 10 acute response following varied blast magnitudes. Neurosci Lett 2014;
.1038/s41598-018-31728-4 570:33–7. doi: 10.1016/j.neulet.2014.03.072
63. Säljö A, Arrhén F, Bolouri H, et al. Neuropathology and pressure in the pig 71. Peelen MV, Caramazza A. Conceptual object representations in human
brain resulting from low-impulse noise exposure. J Neurotrauma 2008; anterior temporal cortex. J Neurosci 2012;32(45):15728–15736; doi: 10
25(12):1397–1406; doi: 10.1089/neu.2008.0602 .1523/JNEUROSCI.1953-12.2012
64. Lu J, Ng KC, Ling G, et al. Effect of blast exposure on the brain structure and 72. Mesulam MM. Temporopolar regions of the human brain. Brain 2023;
cognition in Macaca fascicularis. J Neurotrauma 2012;29(7):1434–1454. 146(1):20–41; doi: 10.1093/brain/awac339
65. Peskind ER, Petrie EC, Cross DJ, et al. Cerebrocerebellar hypometabolism 73. Argyropoulos GPD, van Dun K, Adamaszek M, et al. The cerebellar cog-
associated with repetitive blast exposure mild traumatic brain injury in 12 nitive affective/Schmahmann syndrome: a task force paper. Cerebellum
Iraq war veterans with persistent post-concussive symptoms. Neuroimage 2020;19(1):102–125; doi: 10.1007/s12311-019-01068-8
2011;54 Suppl 1(Suppl 1):S76–S82; doi: 10.1016/j.neuroimage.2010.04.008 74. Fastenrath M, Spalek K, Coynel D, et al. Human cerebellum and cortico-
66. Mac Donald C, Johnson A, Cooper D, et al. Cerebellar white matter cerebellar connections involved in emotional memory enhancement.
abnormalities following primary blast injury in U.S. military personnel. Proc Natl Acad Sci U S A 2022;119(41):e2204900119; doi: 10.1073/pnas
PLoS One 2013;8(2):e55823; doi: 10.1371/journal.pone.0055823 .2204900119
67. Ryu J, Horkayne-Szakaly I, Xu L, et al. The problem of axonal injury in the 75. Strick PL, Dum RP, Fiez JA. Cerebellum and nonmotor function. Annu Rev
brains of veterans with histories of blast exposure. Acta Neuropathol Neurosci 2009;32:413–434; doi: 10.1146/annurev.neuro.31.060407
Commun 2014;2:153; doi: 10.1186/s40478-014-0153-3 .125606
68. Vartanian O, Coady L, Blackler K, et al. Neuropsychological, neurocogni- 76. Huang C, Kritikos M, Sosa MS, et al. World Trade Center site exposure
tive, vestibular, and neuroimaging correlates of exposure to repetitive duration is associated with hippocampal and cerebral white matter
low-level blast waves: evidence from four nonoverlapping samples of neuroinflammation. Mol Neurobiol 2023;60(1):160–170; doi: 10.1007/
Canadian breachers. Mil Med 2021;186(3–4):e393–e400; doi: 10.1093/ s12035-022-03059-z
milmed/usaa332 77. Tschiffely AE, Statz JK, Edwards KA, et al. Assessing a blast-related bio-
69. Hetherington HP, Hamid H, Kulas J, et al. MRSI of the medial temporal marker in an operational community: glial fibrillary acidic protein in
lobe at 7 T in explosive blast mild traumatic brain injury. Magn Reson experienced breachers. J Neurotrauma 2020;37(8):1091–1096; doi: 10
Med 2014;71(4):1358–1367; doi: 10.1002/mrm.24814 .1089/neu.2019.6512