Region-Dependent Mechanical Properties of Human Brain Tissue Under Large Deformations Using Inverse Finite Element Modeling.

This study aims to facilitate intracranial simulation of traumatic events by determining the mechanical properties of different anatomical structures of the brain. Our experimental indentation paradigm used fresh, post-operative human tissue, which is highly advantageous in determining mechanical properties without being affected by postmortem time. This study employed an inverse finite element approach coupled with experimental indentation data to characterize mechanical properties of the human hippocampus (CA1, CA3, dentate gyrus), cortex white matter, and cortex grey matter. We determined that an uncoupled viscoelastic Ogden constitutive formulation was most appropriate to represent the mechanical behavior of these different regions of brain. Anatomical regions were significantly different in their mechanical properties. The cortex white matter was stiffer than cortex grey matter, and the CA1 and dentate gyrus were both stiffer than cortex grey matter. Although no sex dependency was observed, there were trends indicating that male brain regions were generally stiffer than corresponding female regions. In addition, there were no statistically significant age dependent differences. This study provides a structure-specific description of fresh human brain tissue mechanical properties, which will be an important step toward explicitly modeling the heterogeneity of brain tissue deformation during TBI through finite element modeling.

Contaminant film thickness affects walkway friction measurements.

Walkway tribometers are used to measure available friction for evaluating walkway safety and pedestrian slip risk. Numerous variables can affect tribometer measurements, including the type and distribution of contaminants on the surface. Here, we quantified the effect of application method on contaminant film thickness, and the effect of film thickness on tribometer measurements on the four reference walkway surfaces used in ASTM F2508-16e. Distilled water, 0.05% sodium lauryl sulfate (SLS) solution, and 0.04% Triton X-100 solution were poured, squirted, and sprayed onto the surfaces to quantify their naturally occurring film thicknesses. These application methods had a significant effect on the resulting film thickness (p < 0.038), with the pour method consistently generating the thickest films and the spray method generating the thinnest films. We then quantified the effect of film thickness for the three contaminants (thickness range 0.3-3.3 mm) on the friction measurements of three common tribometers (Mark IIIB, English XL, and BOT 3000E) on each reference surface. A separate ANOVA was used for each of the 3 × 4 × 3 = 36 combinations of tribometer, surface, and contaminant. Friction measured with the Mark IIIB decreased with increasing film thickness on one surface across all three contaminants and on a second surface with the SLS contaminant. Friction measured with the BOT 3000E was sensitive to film thickness on two surfaces with water and one surface with Triton. The XL was unaffected by contaminant film thickness. Overall, despite significant differences in film thickness with contaminant application method, friction measurements were either insensitive to film thickness or varied only a small amount in all cases except for the Mark IIIB on the roughest surface. Film thickness did not alter the relative slip resistance of the four ASTM F2508 reference surfaces.

Region-Dependent Viscoelastic Properties of Human Brain Tissue Under Large Deformations.

This study characterizes the mechanical properties of human brain tissue resected during the course of surgery under multistep indentation loading up to 30% strain. The experimental characterization using fresh, post-operative, human brain tissue is highly advantageous since postmortem times can affect its biomechanical behavior. Although the quasilinear theory of viscoelasticity (QLV) approach has been widely used to model brain tissue mechanical properties, our analysis concluded that the linear viscoelastic approach provided a better fit to the experimental data overall. The only statistically significant regional difference in observed stiffness was between the cortex gray and dentate gyrus. There were no statistically significant age or sex dependent differences, although the data suggested that the cortex white matter in males was stiffer than that in females. Our results can help improve the accuracy of finite element models of brain tissue deformation to predict its response to traumatic brain injury.

Quantifying the uncertainty in tribometer measurements on walkway surfaces.

Properly estimating and reporting the uncertainty of walkway surface friction is key to ensuring pedestrian safety. Here we quantified the amount and sources of uncertainty in friction measurements by having four users of four units of each of two walkway tribometer models (Slip-Test Mark IIIB, English XL) perform 12 measurements on four samples of four different surfaces that ranged from slippery to slip-resistant. We found that 51-82% of the total variance in the measurements was explained by the user, unit, sample and a user-unit interaction, which means that the variance a single user calculates from their own data does not capture most of the uncertainty in their measurements. Based on these data, the minimum uncertainty associated with the mean of a user's measurements is ±0.064 (Mark IIIB) and ±0.072 (XL) to be 95% confident that their mean captures a surface's available friction. Practitioner Summary: Walkway surface friction measurements are less accurate than they appear. Based on an experiment quantifying the amount and sources of uncertainty in surface friction measurements using two common tribometers, we quantified and report the minimum uncertainty that users can assign to their walkway surface friction measurements. Abbreviations: ANOVA: analysis of variance; ANSI: American National Standards Institute; CI: confidence interval; E: east; ILS: interlaboratory study; ISO: International Standards Organization; JCGM: joint committee for guides in metrology; N: north; S: south; SBR: styrene-butadiene rubber; SD: standard deviation; TR: test result; W: west.

Brain tissue strains vary with head impact location: A possible explanation for increased concussion risk in struck versus striking football players.

BACKGROUND: On-field football helmet impacts over a large range of severities have caused concussions in some players but not in other players. One possible explanation for this variability is the struck player's helmet impact location. METHODS: We examined the effect of impact location on regional brain tissue strain when input energy was held constant. Laboratory impacts were performed at 12 locations distributed over the helmet and the resulting head kinematics were simulated in two finite element models of the brain: the Simulated Injury Monitor and the Global Human Body Model Consortium brain model. FINDINGS: Peak kinematics, injury metrics and brain strain varied significantly with impact location. Differences in impact location explained 33 to 37% of the total variance in brain strain for the whole brain and cerebrum, considerably more than the variance explained by impact location for the peak resultant head kinematics (8 to 23%) and slightly more than half of the variance explained by the difference in closing speed (57 to 61%). Both finite element models generated similar strain results, with minor variations for impacts that generated multi-axial rotations, larger variations in brainstem strains for some impact locations and a small bias for the cerebellum. INTERPRETATION: Based on this experimental and computational simulation study, impact location on the football helmet has a large effect on regional brain tissue strain. We also found that the lowest strains consistently occurred in impacts to the crown and forehead, helmet locations commonly associated with the striking player.

Regional mechanical properties of human brain tissue for computational models of traumatic brain injury.

To determine viscoelastic shear moduli, stress relaxation indentation tests were performed on samples of human brain tissue resected in the course of epilepsy surgery. Through the use of a 500µm diameter indenter, regional mechanical properties were measured in cortical grey and white matter and subregions of the hippocampus. All regions were highly viscoelastic. Cortical grey matter was significantly more compliant than the white matter or hippocampus which were similar in modulus. Although shear modulus was not correlated with the age of the donor, cortex from male donors was significantly stiffer than from female donors. The presented material properties will help to populate finite element models of the brain as they become more anatomically detailed. STATEMENT OF SIGNIFICANCE: We present the first mechanical characterization of fresh, post-operative human brain tissue using an indentation loading mode. Indentation generates highly localized data, allowing structure-specific mechanical properties to be determined from small tissue samples resected during surgery. It also avoids pitfalls of cadaveric tissue and allows data to be collected before degenerative processes alter mechanical properties. To correctly predict traumatic brain injury, finite element models must calculate intracranial deformation during head impact. The functional consequences of injury depend on the anatomical structures injured. Therefore, morbidity depends on the distribution of deformation across structures. Accurate prediction of structure-specific deformation requires structure-specific mechanical properties. This data will facilitate deeper understanding of the physical mechanisms that lead to traumatic brain injury.

Strong Correlation of Genome-Wide Expression after Traumatic Brain Injury In Vitro and In Vivo Implicates a Role for SORLA.

The utility of in vitro models of traumatic brain injury (TBI) depends on their ability to recapitulate the in vivo TBI cascade. In this study, we used a genome-wide approach to compare changes in gene expression at several time points post-injury in both an in vitro model and an in vivo model of TBI. We found a total of 2073 differentially expressed genes in our in vitro model and 877 differentially expressed genes in our in vivo model when compared to noninjured controls. We found a strong correlation in gene expression changes between the two models (r = 0.69), providing confidence that the in vitro model represented at least part of the in vivo injury cascade. From these data, we searched for genes with significant changes in expression over time (analysis of covariance) and identified sorting protein-related receptor with A-type repeats (SORLA). SORLA directs amyloid precursor protein to the recycling pathway by direct binding and away from amyloid-beta producing enzymes. Mutations of SORLA have been linked to Alzheimer's disease (AD). We confirmed downregulation of SORLA expression in organotypic hippocampal slice cultures by immunohistochemistry and Western blotting and present preliminary data from human tissue that is consistent with these experimental results. Together, these data suggest that the in vitro model of TBI used in this study strongly recapitulates the in vivo TBI pathobiology and is well suited for future mechanistic or therapeutic studies. The data also suggest the possible involvement of SORLA in the post-traumatic cascade linking TBI to AD.

Whiplash Injury or Concussion? A Possible Biomechanical Explanation for Concussion Symptoms in Some Individuals Following a Rear-End Collision.

Study Design Finite element modeling of experimental data. Background The clinical presentations of whiplash injury and concussion have considerable overlap. Both diagnoses are generally based on presenting signs and symptoms, and a history of neck or head trauma. With incomplete knowledge of the trauma, differentiating between whiplash injury and concussion can be clinically challenging. Objectives To estimate the brain strains that develop during rear-end car crashes, evaluate how these strains vary with different head kinematic parameters, and compare these strains to those generated during potentially concussive football helmet impacts. Methods Head kinematic data were analyzed from 2 prior studies, one that focused on head restraint impacts in rear-end crash tests and another that focused on football helmet impacts. These data were used as inputs to a finite element model of the human brain. Brain strains were calculated and compared to different peak kinematic parameters and between the 2 impact conditions. Results Brain strains correlated best with the head's angular velocity change for both impact conditions. The 4 crashes with head angular velocity changes greater than 30 rad/s (greater than 1719°/s) generated the highest brain stains. One crash, in which the head wrapped onto the top of the head restraint, generated brain strains similar to a 9.3-m/s rear football helmet impact, a level previously associated with concussion. Conclusion This work provides new insight into a potential biomechanical link between whiplash injury and concussion, and advances our understanding of how head restraint interaction during a rear-end crash may cause an injury more typically associated with sports-related head impacts. J Orthop Sports Phys Ther 2016;46(10):874-885. doi:10.2519/jospt.2016.7049.

Viscoelastic properties of the P17 and adult rat brain from indentation in the coronal plane.

This technical brief serves as an update to our previous work characterizing the region-dependence of viscoelastic mechanical properties of the P17 and adult rat brain in the coronal plane (Elkin et al., 2011, "A Detailed Viscoelastic Characterization of the P17 and Adult Rat Brain," J. Neurotrauma, 28, pp. 2235-2244.). Here, modifications to the microindentation device provided for the reliable measurement of load during the ramp portion of load relaxation microindentation tests. In addition, a correction factor for finite sample thickness was incorporated to more accurately assess the intrinsic mechanical properties of the tissue.The shear relaxation modulus was significantly dependent on the anatomic region and developmental age, with a general increase in stiffness with age and increased stiffness in the hippocampal and cortical regions compared with the white matter and cerebellar regions of the brain. The shear modulus ranged from ∼0.2 kPa to ∼2.6 kPa depending on region, age, and time scale. Best-fit Prony series parameters from least squares fitting to the indentation data from each region are reported, which describe the shear relaxation behavior for each anatomic region within each age group at both short (<10 ms) and long (∼20 s) time scales. These data will be useful for improving the biofidelity of finite element models of rat brain deformation at short time scales, such as models of traumatic brain injury.

Why is CA3 more vulnerable than CA1 in experimental models of controlled cortical impact-induced brain injury?

One interesting finding of controlled cortical impact (CCI) experiments is that the CA3 region of the hippocampus, which is positioned further from the impact than the CA1 region, is reported as being more injured. The current literature has suggested a positive correlation between brain tissue stretch and neuronal cell loss. However, it is counterintuitive to assume that CA3 is stretched more during CCI injury. Recent mechanical studies of the brain have reported on a level of spatial heterogeneity not previously appreciated-the finding that CA1 was significantly stiffer than all other regions tested and that CA3 was one of the most compliant. We hypothesized that mechanical heterogeneity of anatomical structures could underlie the proposed heterogeneous mechanical response and hence the pattern of cell death. As such, we developed a three-dimensional finite element (FE) rat brain model representing detailed hippocampal structures and simulated various CCI experiments. Four groups of material properties based on recent experiments were tested. In group 1, hyperelastic material properties were assigned to various hippocampal structures, with CA3 more compliant than CA1. In group 2, linear viscoelastic material properties were assigned to hippocampal structures, with CA3 more compliant than CA1. In group 3, the hippocampus was represented by homogenous linear viscoelastic material properties. In group 4, a homogeneous nonlinear hippocampus was adopted. Simulation results demonstrated that for CCI with a 5-mm diameter, flat shape impactor, CA3 experienced increased tensile strains over a larger area and to a greater magnitude than did CA1 for group 1, which best explained why CA3 is more sensitive to CCI injury. However, for groups 2-4, the total volume with high strain (>30%) in CA3 was smaller than that in CA1. The FE rat brain model, with detailed hippocampal structures presented here, will help to engineer desired experimental neurotrauma models by virtually characterizing brain biomechanics before testing.

Viscoelastic properties of the rat brain in the sagittal plane: effects of anatomical structure and age.

Rat is the most commonly used animal model for the study of traumatic brain injury. Recent advances in imaging and computational modeling technology offer the promise of biomechanical models capable of resolving individual brain structures and offering greater insight into the causes and consequences of brain injury. However, there is insufficient data on the mechanical properties of brain structures available to populate these models. In this study, we used microindentation to determine viscoelastic properties of different anatomical structures in sagittal slices of juvenile and adult rat brain. We find that the rat brain is spatially heterogeneous in this anatomical plane supporting previous results in the coronal plane. In addition, the brain becomes stiffer and more heterogeneous as the animal matures. This dynamic, region-specific data will support the development of more biofidelic computational models of brain injury biomechanics and the testing of hypotheses about the manner in which different anatomical structures are injured in a head impact.

Dynamic, regional mechanical properties of the porcine brain: indentation in the coronal plane.

Stress relaxation tests using a custom designed microindentation device were performed on ten anatomic regions of fresh porcine brain (postmortem time <3 h). Using linear viscoelastic theory, a Prony series representation was used to describe the shear relaxation modulus for each anatomic region tested. Prony series parameters fit to load data from indentations performed to ∼10% strain differed significantly by anatomic region. The gray and white matter of the cerebellum along with corpus callosum and brainstem were the softest regions measured. The cortex and hippocampal CA1/CA3 were found to be the stiffest. To examine the large strain behavior of the tissue, multistep indentations were performed in the corona radiata to strains of 10%, 20%, and 30%. Reduced relaxation functions were not significantly different for each step, suggesting that quasi-linear viscoelastic theory may be appropriate for representing the nonlinear behavior of this anatomic region of porcine brain tissue. These data, for the first time, describe the dynamic and short time scale behavior of multiple anatomic regions of the porcine brain which will be useful for understanding porcine brain injury biomechanics at a finer spatial resolution than previously possible.

In vitro models of traumatic brain injury.

In vitro models of traumatic brain injury (TBI) are helping elucidate the pathobiological mechanisms responsible for dysfunction and delayed cell death after mechanical stimulation of the brain. Researchers have identified compounds that have the potential to break the chain of molecular events set in motion by traumatic injury. Ultimately, the utility of in vitro models in identifying novel therapeutics will be determined by how closely the in vitro cascades recapitulate the sequence of cellular events that play out in vivo after TBI. Herein, the major in vitro models are reviewed, and a discussion of the physical injury mechanisms and culture preparations is employed. A comparison between the efficacy of compounds tested in vitro and in vivo is presented as a critical evaluation of the fidelity of in vitro models to the complex pathobiology that is TBI. We conclude that in vitro models were greater than 88% predictive of in vivo results.

Chondroitinase ABC reduces brain tissue swelling in vitro.

Increased intracranial pressure (ICP) caused by edema following severe traumatic brain injury (TBI) or stroke contributes to high rates of mortality and morbidity. The search continues for more effective treatments that target the edema that contributes to increased ICP. We previously described the effect of the fixed charge density (FCD) of brain on its swelling behavior according to the Donnan effect. Here we show that reduction of brain tissue FCD is an effective means of reducing brain tissue swelling and edema in rat and porcine cortical brain tissue in vitro. The effect of enzymes directed at digesting candidate contributors to cellular FCD such as chondroitin sulfate proteoglycans (CSPGs), heparin sulfate proteoglycans (HSPGs), and DNA was examined in slices of the adult rat cortex. All enzymes were capable of decreasing FCD in the tissue by ?20%, and reducing tissue swelling over a 24?h period following dissection from ?60% to ?30%. Chondroitinase ABC (ChABC) was most effective at reducing dead brain tissue swelling in response to changes in ionic osmotic environments. ChABC reduced swelling in live slices of tissue even within the first 2?h following dissection. It also significantly reduced the FCD, initial tissue swelling, and volume change in response to hypotonic bathing solution in porcine cortical brain tissue. The use of ChABC to reduce tissue FCD may be an effective method for reducing brain edema and controlling ICP following injury.

A detailed viscoelastic characterization of the P17 and adult rat brain.

Brain is a morphologically and mechanically heterogeneous organ. Although rat brain is commonly used as an experimental neurophysiological model for various in vivo biomechanical studies, little is known about its regional viscoelastic properties. To address this issue, we have generated viscoelastic mechanical property data for specific anatomical regions of the P17 and adult rat brain. These ages are commonly used in rat experimental models. We measured mechanical properties of both white and gray matter regions in coronal slices with a custom-designed microindentation device performing stress-relaxation indentations to 10% effective strain. Shear moduli calculated for short (100?ms), intermediate (1?sec), and long (20?sec) time points, ranged from ?1?kPa for short term moduli to ?0.4?kPa for long term moduli. Both age and anatomic region were significant factors affecting the time-dependent shear modulus. White matter regions and regions of the cerebellum were much more compliant than those of the hippocampus, cortex, and thalamus. Linear viscoelastic models (Prony series, continuous phase lag, and a power law model) were fit to the time-dependent shear modulus data. All models fit the data equally with no significant differences between them (F-test; p>0.05). The F-test was also used to statistically determine that a Prony series with three time-dependent parameters accurately fit the data with no added benefit from additional terms. The age- and region-dependent rat brain viscoelastic properties presented here will help inform future biomechanical models of the rat brain with specific and accurate regional mechanical property data.

Age-dependent regional mechanical properties of the rat hippocampus and cortex.

Age-dependent outcomes following traumatic brain injury motivate the study of brain injury biomechanics in experimental animal models at different stages of development. Finite element models of the rat brain are used to better understand the mechanical mechanisms behind these age-dependent outcomes; however, age- and region-specific rat brain tissue mechanical properties are required for biofidelity in modeling. Here, we have used the atomic force microscope (AFM) to measure region-dependent mechanical properties for subregions of the cortex and hippocampus in P10, P17, and adult rats. Apparent elastic modulus increased nonlinearly with indentation strain, and a nonlinear Ogden hyperelastic model was used to fit the force-deflection data. Subregional heterogeneous distributions of mechanical properties changed significantly with age. Apparent elastic modulus was also found to increase overall with age, increasing by >100% between P10 and adult rats. Unconfined compression tests (epsilon=-0.3) were performed on whole slices of the hippocampus and cortex of P10, P17, and adult rats to verify the mechanical properties measured with the AFM. Mean apparent elastic modulus at an indentation strain of 30% from AFM measurements for each region and age correlated well with the long-term elastic modulus measured from 30% unconfined compression tests (slope not significantly different from 1, p>0.05). Protein, lipid, and sulfated glycosaminoglycan content of the brain increased with age and were positively correlated with tissue stiffness, whereas water content decreased with age and was negatively correlated with tissue stiffness. These correlations can be used to hypothesize mechanistic models for describing the mechanical behavior of brain tissue as well as to predict relative differences between brain tissue mechanical properties of other species, at different ages, and for different regions based on differences in tissue composition.

Fixed negative charge and the Donnan effect: a description of the driving forces associated with brain tissue swelling and oedema.

Cerebral oedema or brain tissue swelling is a significant complication following traumatic brain injury or stroke that can increase the intracranial pressure (ICP) and impair blood flow. Here, we have identified a potential driver of oedema: the negatively charged molecules fixed within cells. This fixed charge density (FCD), once exposed, could increase ICP through the Donnan effect. We have shown that metabolic processes and membrane integrity are required for concealing this FCD as slices of rat cortex swelled immediately (within 30 min) following dissection if treated with 2 deoxyglucose + cyanide (2DG+CN) or Triton X-100. Slices given ample oxygen and glucose, however, did not swell significantly. We also found that dead brain tissue swells and shrinks in response to changes in ionic strength of the bathing medium, which suggests that the Donnan effect is capable of pressurizing and swelling brain tissue. As predicted, a non-ionic osmolyte, 1,2 propanediol, elicited no volume change at 2000 x 10(-3) osmoles l(-1) (Osm). Swelling data were well described by triphasic mixture theory with the calculated reference state FCD similar to that measured with a 1,9 dimethylmethylene blue assay. Taken together, these data suggest that intracellular fixed charges may contribute to the driving forces responsible for brain swelling.

Quantification of functional alterations after in vitro traumatic brain injury.

Traumatic brain injury (TBI) is caused by mechanical forces, producing tissue deformation at the moment of injury. Complex cellular, neurochemical and metabolic alterations are initiated by the deformation and result in delayed cell death and dysfunction. Using an in vitro model of TBI based on organotypic brain slice cultures, we have quantitatively studied the relationship between tissue deformation and functional outcome. Specifically, we studied the effects of low levels of tissue deformation on the functional outcomes as measured by electrophysiology recordings. In response to 5% and 10% biaxial Lagrangian strain, the maximal evoked response and the excitability of neural networks were found to be decreased. Additionally, the different anatomic subregions of the hippocampus displayed different levels of impairment to the injuries. These results suggest that the network function was affected by low levels of applied strain which induced minimal cell death in previous studies.

Mechanical heterogeneity of the rat hippocampus measured by atomic force microscope indentation.

Knowledge of brain tissue mechanical properties may be critical for formulating hypotheses about traumatic brain injury (TBI) mechanisms and for accurate TBI simulations. To determine the local mechanical properties of anatomical subregions within the rat hippocampus, the atomic force microscope (AFM) was adapted for use on living brain tissue. The AFM provided advantages over alternative methods for measuring local mechanical properties of brain because of its high spatial resolution, high sensitivity, and ability to measure live samples under physiologic conditions. From AFM indentations, a mean pointwise or depth-dependent apparent elastic modulus, E, was determined for the following hippocampal subregions: CA1 pyramidal cell layer (CA1P) and stratum radiatum (CA1SR), CA3 pyramidal cell layer (CA3P) and stratum radiatum (CA3SR), and the dentate gyrus (DG). For all regions, E was indentation-depth-dependent, reflecting the nonlinearity of brain tissue. At an indentation depth of 3microm, E was 234 +/- 152 Pa for CA3P, 308 +/- 184 Pa for CA3SR, 137 +/- 97 Pa for CA1P, 169 +/- 52 Pa for CA1SR, and 201 +/- 133 Pa for DG (mean +/- SD). Our results demonstrate for the first time that the hippocampus is mechanically heterogeneous. Based on our findings, we discuss hypotheses accounting for experimentally observed patterns of hippocampal cell death, which can be tested with biofidelic finite element models of TBI.

Region-specific tolerance criteria for the living brain.

Computational models of traumatic brain injury (TBI) can predict injury-induced brain deformation. However, predicting the biological consequences (i.e. cell death or dysfunction) of induced brain deformation requires tolerance criteria. Here, we present a tolerance criterion for the cortex which exhibits important differences from that of the hippocampus. Organotypic slice cultures of the rat cortex, which maintain tissue architecture and cell content consistent with that in vivo, were mechanically injured with an in vitro model described previously. Cultures were stretched equibiaxially up to 0.35 Lagrangian strain at strain rates up to 50 s(-1). Cell death was quantified at 1, 2, 3, and 4 days following injury. Statistical analysis (repeated measures ANOVA) showed that all three factors (Strain, Strain Rate, and Time post-injury) significantly affected cell death. An equation describing cell death as a function of the significant parameters was then fit to the data. Compared to the hippocampus, the cortex was less vulnerable to stretch-induced injury and demonstrated a strain threshold below 0.20. Strain rate was also a significant factor for cortical but not hippocampal cell death. Cortical cell death began at an earlier time point than in the hippocampus, with cell death evident at 1 day post-injury versus 3 days in the hippocampus. In conclusion, different regions of the brain respond differently to identical mechanical stimuli, and this difference should be incorporated into finite element models of TBI if they are to more accurately predict in vivo consequences of TBI.