The Application of Machine Learning to Identify Large Animal Blast Exposure Thresholds.

INTRODUCTION: The Office of Naval Research sponsored the Blast Load Assessment Sense and Test program to develop a rapid, in-field solution that could be used by team leaders, commanders, and medical personnel to make science-based stand-down decisions for service members exposed to blast overpressure. Toward this goal, the authors propose an ensemble approach based on machine learning (ML) methods to derive a threshold surface for potential neurological deficits that encompasses the intensity of the blast events, the number of exposures, and the period over which the exposures occurred. Because of collection challenges presented by human subjects, the authors utilized data representing a comprehensive set of measures, including structural, behavioral, and cellular changes, from preclinical large animal studies on minipig models. This article describes the development process used to procure the resulting methodology from these studies. METHODS AND MATERIALS: Using an ensemble of ML methods applied to experimental data obtained from 71 Yucatan minipigs, the relationship between blast exposure and neurological deficits was delineated. Despite a relatively small sample size, ML methods with k-fold cross-validation (with k = 5) were justified because of the complexity of the dataset reflecting numerous nonlinear relationships between cellular, structural, and behavioral markers. Based on the physiological responses and environmental measures collected during the large animal study, two models were developed to investigate the relationship between multiple outcome measures and exposure to blast. The histological features model was trained on single-exposure animal data to predict a binary injury response (injured or not) using histological features. The environmental features model related the observed behavioral changes to the environmental parameters collected. RESULTS: The histological features model predicted a binary injury outcome from cellular and physiological measurements. Features identified in developing this classification model showed some level of correlation to observed behavioral changes, suggesting that glial activation inflammation and neurodegenerative responses occur even at the lowest levels of blast exposures tested. The results of the environmental features model, which estimated injury risk from environmental blast exposure characteristics, suggested that the observed changes are not just a function of impulse but an average dynamic impulse rate. Noticeable behavioral deficits were observed at loading rates of 100 kPa (impulse/positive duration) or peak pressures of 300-350 kPa, with an approximate positive phase duration of 3.4 ms for single exposure. Based on this analysis, a 3D threshold surface was developed to characterize the potential risk of neurological deficits. CONCLUSIONS: The ensemble approach facilitated the identification of a pattern of changes across multiple variables to predict the occurrence of changes in brain function. Many changes observed after blast exposure were subtle, making them difficult to measure in human subjects. ML methodologies applied to minipig data demonstrated the value of these techniques in analyzing complex datasets to complement human studies. Importantly, the threshold surface supports the development of science-based blast exposure guidelines.

Development of a Minipig Model of BINT From Blast Exposure Using a Repeatable Mobile Shock Expansion Tube.

INTRODUCTION: The Office of Naval Research (ONR) sponsored the Blast Load Assessment Sense and Test (BLAST) program to provide an approach to operationally relevant monitoring and analysis of blast exposure for optimization of service member performance and health. Of critical importance in this effort was the development of a standardized methodology for preclinical large animal studies that can reliably produce outcome measures that cannot be measured in human studies to support science-based guidelines. The primary advantage of this approach is that, because animal studies report physiological measures that correlate with human neuropathology, these data can be used to evaluate potential risks to service members by accounting for the anatomical and physiological differences between humans and large animal models. This article describes the methodology used to generate a comprehensive outcome measure dataset correlated with controlled blast exposure. METHODS AND MATERIALS: To quantify outcomes associated with a single exposure to blast, 23 age- and weight-matched Yucatan minipigs were exposed to a single blast event generated by a large-bore, compressed gas shock tube. The peak pressure ranged from 280 to 525 kPa. After a post-exposure 72-hour observation period, the physiological response was quantified using a comprehensive set of neurological outcome measures that included neuroimaging, histology, and behavioral measures. Responses of the blast-exposed animals were compared to the sham-treated cohort to identify statistically significant and physiologically relevant differences between the two groups. RESULTS: Following a single exposure, the minipigs were assessed for structural, behavioral, and cellular changes for 3 days after exposure. The following neurological changes were observed: Structural-Using Diffusion Tensor Imaging, a statistically significant decrement (P < .001) in Fractional Anisotropy across the entire volume of the brain was observed when comparing the exposed group to the sham group. This finding indicates that alterations in brain tissue following exposure are not focused at a single location but instead a diffuse brain volume that can only be observed through a systematic examination of the neurological tissue. Cellular-The histopathology results from several large white matter tract locations showed varied cellular responses from six different stains. Using standard statistical methods, results from stains such as Fluoro-Jade C and cluster of differentiation 68 in the hippocampus showed significantly higher levels of neurodegeneration and increased microglia/macrophage activation in blast-exposed subjects. However, other stains also indicated increased response, demonstrating the need for multivariate analysis with a larger dataset. Behavioral-The behavior changes observed were typically transient; the animals' behavior returned to near baseline levels after a relatively short recovery period. Despite behavioral recovery, the presence of active neurodegenerative and inflammatory responses remained. CONCLUSIONS: The results of this study demonstrate that (1) a shock tube provides an effective tool for generating repeatable exposures in large animals and (2) exposure to blast overpressure can be correlated using a combination of imaging, behavioral, and histological analyses. This research demonstrates the importance of using multiple physiological indicators to track blast-induced changes in minipigs. The methodology and findings from this effort were central to developing machine-learning models to inform the development of blast exposure guidelines.

Evaluation of a Field-Ready Neurofunctional Assessment Tool for Use in a Military Environment.

INTRODUCTION: The Office of Naval Research sponsored the Blast Load Assessment Sense and Test (BLAST) program to develop a rapid, in-field solution that could be used by team leaders, commanders, and medical personnel to provide a standardized approach to operationally relevant monitoring and analysis of service members exposed to single or repeated low-level blast. A critical piece of the BLAST team's solution was the development of the Brain Gauge technology which includes a cognitive assessment device that measures neurofunctional changes by testing sensory perceptions and a suite of mathematical algorithms that analyze the results of the test. The most recent versions of the technology are easily portable; the device is in the size and shape of a computer mouse. Tests can be administered in a matter of minutes and do not require oversight by a clinician, making Brain Gauge an excellent choice for field use. This paper describes the theoretical underpinnings and performance of a fieldable Brain Gauge technology for use with military populations. MATERIALS AND METHODS: The methods used by the Brain Gauge have been documented in over 80 peer-reviewed publications. These papers are reviewed, and the utility of the Brain Gauge is described in terms of those publications. RESULTS: The Brain Gauge has been demonstrated to be an effective tool for assessing blast-induced neurotrauma and tracking its recovery. Additionally, the method parallels neurophysiological findings of animal models which provide insight into the sensitivity of specific metrics to mechanisms of information processing. CONCLUSIONS: The overall objective of the work was to provide an efficient tool, or tools, that can be effectively used for (1) determining stand-down criteria when critical levels of blast exposure have been reached and (2) tracking the brain health history until return-to-duty status is achieved. Neurofunctional outcome measures will provide the scientific link between blast sensors and the impact of blast on biological health. This calibration process is strengthened with outcome measures that have a biological basis that are paralleled in animal models. The integrative approach that utilizes the Brain Gauge technology will provide a significant advance for assessing the impact of blast exposure and support rapid, science-based decision-making that will ensure mission success and promote the protection of brain health in service members.

Pre-Clinical Common Data Elements for Traumatic Brain Injury Research: Progress and Use Cases.

Traumatic brain injury (TBI) is an extremely complex condition due to heterogeneity in injury mechanism, underlying conditions, and secondary injury. Pre-clinical and clinical researchers face challenges with reproducibility that negatively impact translation and therapeutic development for improved TBI patient outcomes. To address this challenge, TBI Pre-clinical Working Groups expanded upon previous efforts and developed common data elements (CDEs) to describe the most frequently used experimental parameters. The working groups created 913 CDEs to describe study metadata, animal characteristics, animal history, injury models, and behavioral tests. Use cases applied a set of commonly used CDEs to address and evaluate the degree of missing data resulting from combining legacy data from different laboratories for two different outcome measures (Morris water maze [MWM]; RotorRod/Rotarod). Data were cleaned and harmonized to Form Structures containing the relevant CDEs and subjected to missing value analysis. For the MWM dataset (358 animals from five studies, 44 CDEs), 50% of the CDEs contained at least one missing value, while for the Rotarod dataset (97 animals from three studies, 48 CDEs), over 60% of CDEs contained at least one missing value. Overall, 35% of values were missing across the MWM dataset, and 33% of values were missing for the Rotarod dataset, demonstrating both the feasibility and the challenge of combining legacy datasets using CDEs. The CDEs and the associated forms created here are available to the broader pre-clinical research community to promote consistent and comprehensive data acquisition, as well as to facilitate data sharing and formation of data repositories. In addition to addressing the challenge of standardization in TBI pre-clinical studies, this effort is intended to bring attention to the discrepancies in assessment and outcome metrics among pre-clinical laboratories and ultimately accelerate translation to clinical research.

Methodological Problems With Online Concussion Testing.

Reaction time testing is widely used in online computerized concussion assessments, and most concussion studies utilizing the metric have demonstrated varying degrees of difference between concussed and non-concussed individuals. The problem with most of these online concussion assessments is that they predominantly rely on consumer grade technology. Typical administration of these reaction time tests involves presenting a visual stimulus on a computer monitor and prompting the test subject to respond as quickly as possible via keypad or computer mouse. However, inherent delays and variabilities are introduced to the reaction time measure by both computer and associated operating systems that the concussion assessment tool is installed on. The authors hypothesized systems that are typically used to collect concussion reaction time data would demonstrate significant errors in reaction time measurements. To remove human bias, a series of experiments was conducted robotically to assess timing errors introduced by reaction time tests under four different conditions. In the first condition, a visual reaction time test was conducted by flashing a visual stimulus on a computer monitor. Detection was via photodiode and mechanical response was delivered via computer mouse. The second condition employed a mobile device for the visual stimulus, and the mechanical response was delivered to the mobile device's touchscreen. The third condition simulated a tactile reaction time test, and mechanical response was delivered via computer mouse. The fourth condition also simulated a tactile reaction time test, but response was delivered to a dedicated device designed to store the interval between stimulus delivery and response, thus bypassing any problems hypothesized to be introduced by computer and/or computer software. There were significant differences in the range of responses recorded from the four different conditions with the reaction time collected from visual stimulus on a mobile device being the worst and the device with dedicated hardware designed for the task being the best. The results suggest that some of the commonly used visual tasks on consumer grade computers could be (and have been) introducing significant errors for reaction time testing and that dedicated hardware designed for the reaction time task is needed to minimize testing errors.

Noninvasive, In-pen Approach Test for Laboratory-housed Pigs.

Traumatic brain injury (TBI) incidences have increased in both civilian and military populations, and many researchers are adopting a porcine model for TBI. Unlike rodent models for TBI, there are few behavioral tests that have been standardized. A larger animal requires more invasive handling in test areas than rodents, which potentially adds stress and variation to the animals' responses. Here, the human approach test (HAT) is described, which was developed to be performed in front of laboratory pigs' home pen. It is noninvasive, but flexible enough that it allows for differences in housing set-ups. During the HAT, three behavioral ethograms were developed and then a formula was applied to create an approach index (AI). Results indicate that the HAT and its index, AI, are sensitive enough to detect mild and temporary alterations in pigs' behavior after a mild TBI (mTBI). In addition, although specific behavior outcomes are housing-dependent, the use of an AI reduces variation and allows for consistent measurements across laboratories. This test is reliable and valid; HAT can be used across many laboratories and for various types of porcine models of injury, sickness, and distress. This test was developed for an optimized manual timestamping method such that the observer consistently spends no more than 9 min on each sample.

Quantification of Mild Traumatic Brain Injury via Cortical Metrics: Analytical Methods.

Mild traumatic brain injuries are difficult to diagnose or assess with commonly used diagnostic methods. However, the functional state of cerebral cortical networks can be rapidly and effectively probed by measuring tactile-based sensory percepts (called cortical metrics), which are designed to exercise various components of cortical machinery. In this study, such cortical metrics were obtained from 52 college students before and after they experienced sports-related concussions by delivering vibrotactile stimuli to the index and middle fingertips. Performance on four of the sensory test protocols is described: reaction time, amplitude discrimination, temporal order judgment, and duration discrimination. The collected test performance data were analyzed using methods of uni- and multivariate statistics, receiver operated characteristic (ROC) curves, and discriminant analysis. While individual cortical metrics vary extensively in their ability to discriminate between control and concussed subjects, their combined discriminative performance greatly exceeds that of any individual metric, achieving cross-validated 93.0% sensitivity, 92.3% specificity, 93.0% positive predictive value, and 92.3% negative predictive value. The cortical metrics vector can be used to track an individual's recovery from concussion. The study thus establishes that cortical metrics can be used effectively as a quantitative indicator of central nervous system health status.

Inosine enhances axon sprouting and motor recovery after spinal cord injury.

Although corticospinal tract axons cannot regenerate long distances after spinal cord injury, they are able to sprout collateral branches rostral to an injury site that can help form compensatory circuits in cases of incomplete lesions. We show here that inosine enhances the formation of compensatory circuits after a dorsal hemisection of the thoracic spinal cord in mature rats and improves coordinated limb use. Inosine is a naturally occurring metabolite of adenosine that crosses the cell membrane and, in neurons, activates Mst3b, a protein kinase that is part of a signal transduction pathway that regulates axon outgrowth. Compared to saline-treated controls, rats with dorsal hemisections that were treated with inosine showed three times as many synaptic contacts between corticospinal tract collaterals and long propriospinal interneurons that project from the cervical cord to the lumbar level. Inosine-treated rats also showed stronger serotonergic reinnervation of the lumbar cord than saline-treated controls, and performed well above controls in both open-field testing and a horizontal ladder rung-walking test. Inosine was equally effective whether delivered intracranially or intravenously, and has been shown to be safe for other indications in humans. Thus, inosine might be a useful therapeutic for improving outcome after spinal cord injury.

GGF2 (Nrg1-ß3) treatment enhances NG2+ cell response and improves functional recovery after spinal cord injury.

The adult spinal cord contains a pool of endogenous glial precursor cells, which spontaneously respond to spinal cord injury (SCI) with increased proliferation. These include oligodendrocyte precursor cells that express the NG2 proteoglycan and can differentiate into mature oligodendrocytes. Thus, a potential approach for SCI treatment is to enhance the proliferation and differentiation of these cells to yield more functional mature glia and improve remyelination of surviving axons. We previously reported that soluble glial growth factor 2 (GGF2)- and basic fibroblast growth factor 2 (FGF2)-stimulated growth of NG2(+) cells purified from injured spinal cord in primary culture. This study examines the effects of systemic administration of GGF2 and/or FGF2 after standardized contusive SCI in vivo in both rat and mouse models. In Sprague-Dawley rats, 1 week of GGF2 administration, beginning 24 h after injury, enhanced NG2(+) cell proliferation, oligodendrogenesis, chronic white matter at the injury epicenter, and recovery of hind limb function. In 2',3'-cyclic-nucleotide 3'-phosphodiesterase-enhanced green fluorescent protein mice, GGF2 treatment resulted in increased oligodendrogenesis and improved functional recovery, as well as elevated expression of the stem cell transcription factor Sox2 by oligodendrocyte lineage cells. Although oligodendrocyte number was increased chronically after SCI in GGF2-treated mice, no evidence of increased white matter was detected. However, GGF2 treatment significantly increased levels of P0 protein-containing peripheral myelin, produced by Schwann cells that infiltrate the injured spinal cord. Our results suggest that GGF2 may have therapeutic potential for SCI by enhancing endogenous recovery processes in a clinically relevant time frame.

Inosine augments the effects of a Nogo receptor blocker and of environmental enrichment to restore skilled forelimb use after stroke.

Stroke is the leading cause of disability in much of the world, with few treatment options available. Following unilateral stroke in rats, inosine, a naturally occurring purine nucleoside, stimulates the growth of projections from the undamaged hemisphere into denervated areas of the spinal cord and improves skilled use of the impaired forelimb. Inosine augments neurons' intrinsic growth potential by activating Mst3b, a component of the signal transduction pathway through which trophic factors regulate axon outgrowth. The present study investigated whether inosine would complement the effects of treatments that promote plasticity through other mechanisms. Following unilateral stroke in the rat forelimb motor area, inosine combined with NEP1-40, a Nogo receptor antagonist, doubled the number of axon branches extending from neurons in the intact hemisphere into the denervated side of the spinal cord compared with either treatment alone, and restored rats' level of skilled reaching using the impaired forepaw to preoperative levels. Similar functional improvements were seen when inosine was combined with environmental enrichment (EE). The latter effect was associated with changes in gene expression in layer 5 pyramidal neurons of the undamaged cortex well beyond those seen with inosine or EE alone. Inosine is now in clinical trials for other indications, making it an attractive candidate for the treatment of stroke patients.

Inosine alters gene expression and axonal projections in neurons contralateral to a cortical infarct and improves skilled use of the impaired limb.

Recovery after stroke and other types of brain injury is restricted in part by the limited ability of undamaged neurons to form compensatory connections. Inosine, a naturally occurring purine nucleoside, stimulates neurons to extend axons in culture and, in vivo, enhances the ability of undamaged neurons to form axon collaterals after brain damage. The molecular changes induced by inosine are unknown, as is the ability of inosine to restore complex functions associated with a specific cortical area. Using a unilateral injury model limited to the sensorimotor cortex, we show that inosine triples the number of corticospinal tract axons that project from the unaffected hemisphere and form synaptic bouton-like structures in the denervated half of the spinal cord. These changes correlate with improved recovery in animals' ability to grasp and consume food pellets with the affected forepaw. Studies using laser-capture microdissection and microarray analysis show that inosine profoundly affects gene expression in corticospinal neurons contralateral to the injury. Inosine attenuates transcriptional changes caused by the stroke, while upregulating the expression of genes associated with axon growth and the complement cascade. Thus, inosine alters gene expression in neurons contralateral to a stroke, enhances the ability of these neurons to form connections on the denervated side of the spinal cord, and improves performance with the impaired limb.

Increased growth factor expression and cell proliferation after contusive spinal cord injury.

The damage caused by traumatic central nervous system (CNS) injury can be divided into two phases: primary and secondary. The initial injury destroys many of the local neurons and glia and triggers secondary mechanisms that result in further cell loss. Approximately 50% of the astrocytes and oligodendrocytes in the spared white matter of the epicenter die by 24 h after spinal cord injury (SCI), but their densities return to normal levels by 6 weeks. This repopulation is largely due to the proliferation of local progenitors that divide in response of CNS injury. Previous studies indicate that the secondary events that cause cell death after SCI also increase the local levels of several growth factors that stimulate the proliferation of these endogenous progenitors. We compared the spatial pattern of the post-injury up-regulation of the pro-mitotic growth factors with that of 5-bromodeoxyuridine (BrdU) incorporation to determine if each could play a role in proliferation. Three days after a standard contusive SCI or laminectomy, animals received intraperitoneal BrdU injections to label dividing cells and were perfused 2 h after the last injection. Immunohistochemistry for BrdU and basic fibroblast growth factor (FGF2) and in situ hybridization for ciliary neurotrophic factor (CNTF) and glial growth factor (GGF2) mRNA were used to compare the number of dividing cells with growth factor levels in sections 2 and 4 mm from the epicenter. All three growth factors are significantly up-regulated 3 days after SCI, when cell proliferation is maximal. The increase in GGF2 and FGF2 levels is highest in sections 2 mm rostral to the epicenter, mimicking BrdU incorporation. Addition of rhGGF2 to cultured cells isolated from the spinal cord 3 days after SCI increased the number of NG2+ glial progenitors. These data suggest that FGF2 and GGF2 may contribute to the spontaneous recovery observed after SCI by stimulating the proliferation of local progenitors that help repopulate the injured cord.

Cell proliferation and replacement following contusive spinal cord injury.

After spinal cord injury (SCI), about 50% of the oligodendrocytes and astrocytes in the residual white matter at the injury site are lost by 24 h. However, chronically after SCI, the density of oligodendrocytes is normal. Previous studies have shown that the adult rat spinal cord contains a pool of proliferating glial progenitors whose progeny could help restore cell density after injury. To study proliferation in response to injury, we performed SCI on adult female rats at the T8 level, using a standardized contusion model. Animals received bromodeoxyuridine (BrdU) injections during the first week after SCI, and were perfused within 2 h for acute studies, and at 6 weeks for chronic studies. The tissue was analyzed using immunohistochemical detection of BrdU and cell marker antigens. We demonstrate that cell proliferation in the residual white matter is increased at 1-7 days after SCI, peaking on day 3. Dividing cells include oligodendrocytes, astrocytes, microglia/macrophages, and a high proportion of NG2(+) glial precursors. By 6 weeks, some cells that had been labeled 2-4 days after SCI were still present. Double immunohistochemistry showed that while very few of these cells expressed NG2 or the microglia/macrophage marker OX42, about 50% expressed CC1 or glial fibrillary acidic protein (GFAP), markers of mature oligodendrocytes and astrocytes, respectively. The post-injury environment represented by residual white matter is thus permissive to the differentiation of glial precursors. Cells that are stimulated to divide during the first week after SCI develop chronically into mature phenotypes that replace macroglia lost after injury.

Chronic alterations in the cellular composition of spinal cord white matter following contusion injury.

Spinal cord injury (SCI) involves the loss of neurons and glia due to initial mechanical and secondary biochemical mechanisms. Treatment with the sodium channel blocker tetrodotoxin (TTX) reduces acute white matter pathology and increases both axon density and hindlimb function chronically at 6 weeks after injury. We investigated the cellular composition of residual white matter chronically to determine whether TTX also has a significant effect on the numbers and types of cells present. Rats received an incomplete thoracic contusion injury, in the presence or absence of TTX (0.15 nmole) injected focally, beginning at 15 min prior to injury. Six weeks later, cell density was significantly increased in the residual white matter of the dorsal, lateral, and ventral funiculi, both rostral and caudal to the injury site in both TTX-treated and injury control groups. Oligodendrocyte and astrocyte density was similar to normal but large numbers of cells expressing microglia/macrophage markers were present. Labeling with the progenitor markers nestin and NG2 showed that precursor cell density had also doubled or tripled as compared with uninjured controls. Some of these cells were also labeled for antigens that indicate their possible progression along an oligodendrocyte or astrocyte lineage. Our results support the hypothesis that the beneficial effect of TTX in SCI is related to its preservation of axons per se; no effect on chronic white matter cell composition was detected. They highlight the profound changes in cellular composition in preserved white matter chronically at 6 weeks after injury, including the accumulation of endogenous progenitor cells and the persistence of activated macrophages/microglia. The manipulation of these endogenous cells may be used in the future to enhance recovery after SCI.