Brain injuries exhibit high incidence and high mortality in vehicle collision accidents. An investigation from Centers for Disease Control and Prevention in American showed that there were about 2.87 million Traumatic Brain Injury (TBI) related emergency department (ED) visits, hospitalizations, and deaths occurred in the United States in 2014, with an average of 155 people died each day from injuries that include TBI (Peterson et al. 2014). Falls (52%) and motor vehicle crashes (20%) were the first and second leading causes of all TBI-related hospitalizations (Peterson et al. 2014). Based on NASS-CDS analyses of frontal crashes (Eigen and Martin, 2005), fatalities attributable to head injuries, with societal costs exceeding $6 Billion, are second only to fatalities attributable to thoracic region. TBI in vehicular accidents is one of the major causes of mortality and permanent disability in many countries (Faul et al. 2010). For instance, in the USA, such accidents account for the greatest fraction (31.8%) of the TBI-related deaths (Faul et al. 2010). Comprehensive understanding to brain injury may improve restraint system and traffic safety.
Brain injuries have different forms and severities, with different mechanisms and physiological characterizations. Understanding injury mechanisms may enhance the injury characterization in brain models. Clinically brain injuries can be classified in two broad categories: diffuse injuries and focal injuries. Statistics showed that 3/4 automobile/pedestrian brain injuries were of diffuse type, while assault and fall victims had 3/4 focal injuries (Gennarelli et al. 1982). Acute subdural hematoma (ASDH) and diffuse axonal injury (DAI) were the two most important causes of death, which accounted for more head injury deaths than all other lesions combined (Gennarelli et al. 1982). DAI is regarded as one of the most severe brain diffuse injuries, caused by axon damage in brain and the degrees of morbidity depend on the number of axons that are damaged (Povlishock and Christman, 1995; Blumbergs et al., 1995). As strain level applied to axons grows in intensity, a series of increasingly severe pathological changes occur. At strain levels of 10–15%, axonal swelling and a total loss of axonal transport begin (2–6 hours after the injury). Hematoma is a typical brain focal injury, classified by injury region and includes epidural hematoma (EDH), subdural hematoma (SDH), and intracerebral hematoma (ICH). Subdural hematoma is caused by penetrating wounds, large-contusion bleeding into the subdural, or more commonly tearing of veins that bridge the subdural (called bridging veins, BVs). According to Gennarelli and Thibault (1982), ASDH is the most important cause of death in severely head-injured patients due to high incidence (30%), high mortality (60%), and head injury severity (2/3 with Glasgow Coma scores of 3–5).
Finite element (FE) model of human head is suitable tool for studying the complex, tissue-level mechanical response of brain injury in head impact. From 1975, after the first head and brain FE model was proposed by Shugar and Katona (1975), many FE brain models have been developed and used for impact induced injury analysis. A few widely recognized brain models include WSUBIM (Wayne State University Brain Injury Model) (Zhang et al. 2001a), SIMon (by University of Western Australia) (Takhounts et al. 2003), KTH (by Royal Institute of Technology Sweden) (Kleiven 2007), UCDBTM (by Department of Mechanical Engineering, University College Dublin, Ireland) (Horgan and Gilchrist, 2003, 2004), SUFEHM (Strasbourg University Finite Element Head Model) (Willinger et al. 1995; Deck and Willinger, 2008; Sahoo et al. 2014), and the head part from whole body FE models such as THUMS (Total Human Model for Safety) (Kimpara et al. 2006) and GHBMC (Global Human Body Models Consortium) (Mao et al. 2013), etc. Most brain tissues are reflected in those models to characterize specific forms of injury. For example, bridging vein (BV) is a key component for research on ASDH. On human brain, one end of the bridging vein is connected to the cerebrum and the other end is to the superior sagittal sinus (SSS). The BVs in some FE brain models were simplified to spring or beam elements. To represent DAI, in recent years, axonal fibers were modeled by Fiber Tractograghy in FE brain models (Garimella et al. 2016, Wu et al. 2019).
Using brain models, some efforts were put on developing brain injury metrics. As summarized in Fig. 1, there are two representative research approaches. The counterclockwise flow shows a routine and logical way. It first studies physiological characterization of common brain injury types based on medical knowledge, then builds related brain structures in FE models, and lastly, uses physical parameters or physical phenomenon such as structural fracture to characterize the corresponding injury. The modeling of BV and axonal fibers are representatives of such an approach. However, the research progress using this approach is relatively limited and slow, as using these structures to represent brain injury requires advanced imaging technology and detailed biological material information. Takhounts et al. (2003) pointed out that while explicit BVs were modeled in SIMon, they were not used in injury prediction. In contrast, more researches have been done in direct and more effective way. They establish correlations between brain injury and corresponding injury metrics (Takhounts et al. 2003, Zhu et al. 2015, Al-Bsharat et al. 1999, Zhou et al. 1994, Zhang et al. 2004, Gabler et al. 2016) or between different metrics (Zhang et al. 2001b, Gabler et al. 2016, Takhounts et al. 2013). Each line in Fig. 1 indicates a piece of medical knowledge or some existing researches. For example, “line 1” indicates that bleeding happens when suffering a hemorrhage, while “line 2” represents that Takhounts et al. have studied the relation between DAI and Cumulative Strain Damage Measure (CSDM) (2003).
As the brain models were developed by various organizations, there were differences in modeling approaches. The two main aspects are material property and structure definition. THUMS, SIMon and GHBMC use linear viscoelastic material models for brain tissue, while KTH and SUFEHM use hyperelastic-viscoelastic models. The properties assigned to these materials vary from the stiffest with G0 = 49 kPa and G1 = 16 kPa (SUFEHM) to the softest with G0 = 1.6 kPa and G1 = 0.9 kPa (SIMon). The contact definitions between the parts are different as well. Wang et al. (2018) studied four kinds of interface between skull and brain, from the strongest (directly attaching to each other) to the weakest (friction-less sliding contact). There have been debates in the academic circle about the definitions of the skull-brain interface. Different FE modeling approaches affect mechanical responses (Claessens et al. 1997, Zhang et al. 2001a, Kleiven and Hardy 2002, Wang et al. 2018), and therefore, validation of model is important in developing human body model (HBM) to ensure the biofidelity. To validate the FE brain models, researchers usually used brain motion or intracranial pressure from cadaver tests (Hardy et al. 2001, 2007, Trosseille et al. 1992). The head models in THUMS and GHBMC, as two of the commonly used HBMs in our field, have been validated under a variety of loading conditions in their respective development processes, including using the same loading conditions such as the mechanical response of skull (Yoganandan et al. 1995), intracranial pressure (Nahum et al. 1977), and displacement of internal brain markers (Hardy et al. 2001). Those validations have ensured similar mechanical responses of the two models.
Although most brain models have been validated under several similar load cases, different modeling approaches can lead to different kinematic and kinetic response in application, even different injury metric values. The influence of modeling approaches on injury metric were less considered in the literature. Takhounts et al. published two papers in 2003 and 2008, respectively, studying the correlations of biomechanical injury metrics with their corresponding injuries. Surprisingly, they found that the correlation between RMDM (relative motion damage measure) and ASDH was totally different using different versions of SIMon models. It means that output of injury metrics are highly dependent on the models. Even differences between model versions could matter. These issues have increased the complexity and unknowingness of the problem. In this study, we compared the kinematic and mechanical responses of four brain models with different modeling approaches, concentrating on brain-skull interface and brain material property. The loading condition was based on animal tests from Abel et al (1978).The influence of brain modeling approaches on the application of injury metrics was also discussed.