Modeling of inflicted head injury by shaking trauma in children: what can we learn?

In the majority of papers describing physical models, a doll was shaken by one or more volunteers. Instructions given to the volunteers vary, but often come down to: “shake as violently as possible”. In some studies, an explicit distinction was made between shaking in the sagittal plane while keeping the infant more or less upright and “gravity assisted shaking” where shaking is accompanied by a forceful up- and downward acceleration of the infant (i.e. Cory & Jones [4] and Lloyd et al. [7]). Accelerations at both the hands of the volunteers and at the head of the doll were only measured directly in Yamazaki et al. [10], thus taking into account Steps 1 to 3 in Fig. 2. Typically, shaking frequencies and durations attained during human shaking experiments were 2-5 Hz for 3-5 seconds. In Cheng et al. [6] and Koizumi et al. [9] machine-based shaking was applied, using actuators to apply shaking frequencies of about 3.5 Hz and 1.5 – 3.5 Hz, respectively.

In RBM studies on IHI-ST, researchers use experimental data on torso kinematics to drive their models, either by using these as a kinematic constraint (e.g. prescribing the movement of the modelled torso) or as an external force acting on the torso. Lintern et al. [17] experimentally measured torso motion while shaking a lamb is used as input to their lamb RBM. In Jones et al. [16] a sinusoidal shake of 3Hz and amplitude of 65mm is applied to their model, based on human shaking data from Cory & Jones [4]. Wolfson et al. [14] used experimental data on linear torso acceleration from shaking a doll as input to their model. Bondy et al. [15] used the same experimental data to drive their model.

FEMs generally model only part of the body (typically the head or eye and their internal structures) during IHI-ST events. Therefore, torso dynamics is not part of the modeling process.

In doll studies, head kinematics is strongly dependent on the coupling between the torso and the head: the neck. In studies that included effects of neck properties on the torso-skull transfer, either proprietary test dolls or commercial crash-test dummies were used for experimentation. In studies using proprietary dolls [2‐5, 8, 10, 13] different neck types were tested, ranging from elastic tubing to frictionless hinge joints. In these studies, biofidelity of the neck models was either not discussed or stated to be unknown. Prange et al. [5] argued that their hinge type neck might be appropriate for new-borns, who cannot support the weight of their heads, and that these experiments provided an upper bound for head kinematics during IHI-ST. Jenny et al. [3, 13] are the only ones providing quantitative data of the neck stiffness of their proprietary doll. Overall, there seems to be a consensus that quantitative data on biomechanical properties of the infant’s neck is lacking. In studies using commercial crash-test dummies [7‐9, 11], biofidelity of the dolls was not discussed in detail.

Two RBM studies investigated the role of passive neck stiffness on skull dynamics [14, 16]. In Jones et al. [16] it is reported that peak linear accelerations at the vertex during simulated IHI-ST events can increase sevenfold depending on the necks stiffness. Wolfson et al. [14] varied the neck stiffness profile in their model and reported that resulting head dynamics for all shaking simulations and all tested neck stiffnesses were below the injury thresholds for head injury against which the results were compared (see Step 6. Injury thresholds). Only when end-point constraints were encountered in the simulations, causing the head to impact on the chest and back during shaking, were head dynamics found to exceed injury thresholds. Bandak [34] calculated neck distraction forces from angular velocities and accelerations taken from literature. On the basis of these calculations, it was argued that the forces required to sustain angular velocities and angular accelerations during IHI-ST exceed the limits of structural failure of the infant’s spine, implying that IHI-ST caused by shaking alone would be unlikely to exist without neck injuries. However, Margulies et al. [35] recalculated neck distraction forces using data from Bandak [34] and criticized the forces reported in [34] as being at least ten times too large.

As in Step 1, none of the selected FEM studies addresses torso-head transfer: shaking inputs were applied directly to the head or eye.

In doll studies, head dynamics and kinematics are often measured using accelerometers embedded on different locations in the doll’s head. In many of these studies, overall head kinematics are the main results reported (Table 3). Three studies suggested that the largest angular accelerations in IHI-ST occurred during chin-to-chest or occiput-to-back contact [4, 7, 13].

RBM studies are similar to physical modeling studies in the sense that an IHI-ST event is applied to the model, often shaking applied to the torso, and the resulting head dynamics are calculated using the RBM (Table 4). Both Wolfson et al. [14] and Lintern et al. [17] reported that the largest accelerations in their models occur during chin-to-chest or occiput-to-back endpoint contacts.

Only a few studies used dolls with detailed internal head anatomy [6, 9‐12]. These dolls contained, for example, bone structures obtained from CT-scans or models for cerebro-spinal fluid (CSF), brain, or eyes. In these studies, water was used to model CSF, gelatin or silicone gel to simulate brain tissue, and agar gel to represent internal eye structures.

RBMs described in the literature focus on gross body dynamics and do not address internal transfer within the skull. FEMs, however, explicitly focus on modeling the mechanical behavior of micro-structures, such as stresses and strains in the brain and bridging veins or local pressures in the retina. For this purpose, attention has been devoted to modeling techniques for CSF [6, 24‐26, 28]. In some studies, the influence of anatomical features such as the fontanelle [6, 21, 26, 29] and the size of the subarachnoid space [22] on the amount of movement and deformation of the brain was qualitatively investigated. For all FEM studies, having representative material properties and detailed geometrical data for all anatomical structures in the head is quintessential. For this purpose, material properties from the literature are generally used in such studies. Batterbee et al. [29] applied another approach and did a sensitivity analysis to identify parameters to which FEM output was most sensitive.

Several studies investigated stretching of bridging veins during IHI-ST by placing markers onto a surrogate brain and on the skull [6, 9, 11]. By tracking the relative positions of these markers during IHI-ST experiments, the stretching of bridging veins was estimated. In Cheng et al. [6], marker excursions in the order of 2.5-3mm were reported. Peak values for bridging vein stretch ratios (stretched length divided by original length, see Appendix A) found by Miyazaki [11] ranged from 4 to 5 and the ones found by Koizumi et al. [9] ranged from 1.5 to 3.

Stresses within the eye during IHI-ST were measured using a pressure sensor inside an eye model in Yamazaki et al. [10], showing peak compressive stresses up to 0.85kPa and peak tensile stresses up to 0.62kPa, both at the posterior pole of the eye and averaged across their subjects shaking the doll. Using a 2D representation of a sagittal slice of the head, internal stress distributions during machine based shaking was visualized using a photo-elastic material in Tomlinson & Taylor [12]. They reported peak shear stresses at the brainstem of 1.15kPa.

In Batterbee et al. [26] it was qualitatively stated that the presence of a fontanelle could increase the likelihood of rupture of bridging veins. Couper & Albermani [28] showed data with peak stretch ratios of bridging veins up to approximately 2.15 and Morison [18] found bridging veins stretch ratios up to 1.25 during a simulated IHI-ST event. Raul et al. [22] investigated the effect of the size of the subarachnoid space on bridging vein stretch ratios, reporting peak ratios in the vertex area of 1.9 for normal and 1.22 for larger sized subarachnoid spaces. In sensitivity analyses performed by Batterbee et al. [29], bridging vein stretch ratios were found to range between 1.2 for pure translational motion up to 3 for combined translational and rotational movement. Values depended on choice of model parameters such as material properties and CSF layer thickness. Roth et al. [20] reported maximum bridging vein stretch ratios of 1.9 in their simulations and a maximum pressure of 22kPa at the frontal area of the brain. Using a two-step approach, Coats et al. [31] first validated their FEM for piglet brain by comparing model outputs to experimental data on piglet brain deformation and brain-skull displacement. They reported that for a good fit the model required adjusting the brain stiffness, which was derived from in vivo experiments, by a factor 1.5-2.25, suggesting that porcine brain is stiffer in situ than in vivo. Next, threshold values for injury were determined for their FEM by comparing model outputs to experimental data on porcine intracranial hemorrhaging during IHI-ST experiments.

In Cheng et al. [21] it was reported that the highest stresses (Von Mises stresses about 1MPa) in the brain appear near the fontanelle in their FEM. Couper & Albermani [28] provide detailed analyses of brain strains in their 3D FEM during IHI-ST. High strains were developed in contact points between frontal lobe and cranium, occipital lobe and sinus confluence, and in the inferior sagittal sinus. Besides this, high strains manifested in corpus callosum and in connections between brainstem and cerebrum. Peak values for strains reported in [28] were 0.7 (Tresca strains). Morison [18] observed brain stresses to increase from brainstem to vertex, with maximum values of 0.8 kPa. Ponce & Ponce [30] found the largest displacements in the central areas of the brain and a more or less uniform distribution of stresses across the brain.

In exploring the response of their eye FEM, Cirovic et al. [19] reported displacements accumulating up to 0.8 mm in the center of the eye and a maximum stress of 12MPa at the orbit when the model was driven by its characteristic frequency of 200Hz. Hans et al. [23] reported mean retinal nodal forces (forces at key points of the FEM) of 0.05-0.08N with peaks up to 0.45N. In the eye FEM from Nadarasa et al. [33] pressures of 1.5-2kPa at the posterior pole of the eye, extending to the mid retina were found. Rangarajan et al. [27] reported stresses in the eye of 14-120kPa, depending on modeling techniques, with the largest stresses found at the interface between the retina and vitreous body. Yoshida et al. [32] reported that the location of maximum stress in their IHI-ST simulations was at the posterior pole of the eye. In their data, the maximum normal component of stress was approximately 1kPa.

In many IHI-ST studies, injury thresholds are mainly referred to in order to estimate whether the measured head accelerations could be hazardous. For this purpose, many studies use plots of (peak) angular acceleration vs (peak) angular velocities (henceforth called α-ω plots), an injury criterion originally used in head impact studies. In order to be applied to IHI-ST events, injury thresholds from different sources in animal impact literature have been scaled down for use on infants (Table 5). The use of α-ω plots or scaled data from animal experiments has been criticized in several studies and it has been suggested that injury mechanisms, and hence injury thresholds, might be different for IHI-ST as compared to events with impact [4, 10, 13‐15, 18, 25, 28]. Alternatively, Cory & Jones [4] and Lloyd et al. [7] used the head-injury-criterion (HIC) using threshold values by Van Ee et al. [36] and Stürtz [37]. HIC is a measure to assess the likelihood of injury to occur as a consequence of impact originating from studies in automotive safety research. It is computed by integrating measured accelerations over short time periods (typically a few tens of milliseconds).

In the few physical model studies addressing internal head dynamics, threshold values for bridging vein rupture were taken from in situ experiments on cadaveric human bridging veins by Lee & Haut [38] and a cadaver impact study by Depreitere et al. [39] (Table 5).

RBM studies focusing on gross head dynamics rely on similar sources for injury thresholds as the physical modeling studies described before (Table 6). For FEM studies, however, much more detailed injury thresholds are required: detailed limits for the stresses and strains that the biomaterials can withstand.

FEM studies investigating the possibility of SDH in IHI-ST used data from Lee & Haut [38] and Depreitere et al. [39] for bridging vein stretch ratio limits. Morison [18] was one of the very few specifically measuring these stretch ratio limits for infants in a small sample of pediatric bridging veins and suggests that tolerance levels for these veins might be lower than values reported for larger and higher age ranges in literature. Threshold values for tissue strains leading to DAI in Couper & Albermani [28] were derived from in vitro animal experiments by Morrison et al. [40], while Ponce & Ponce [30] used brain shear stress thresholds that were not measured, but estimated by Meyer et al. [41] through numerical reconstruction of real-life cases. In their FEM used to study the likelihood of RH during IHI-ST events, Hans et al. [23] used threshold values for the retinal adhesion force from monkey eyes experiments by Kita & Marmor [42, 43]. Alternatively, Coats et al. [31] determined threshold values for intracranial hemorrhaging and principal stress at the surface of the brain in a piglet brain FEM using an optimizing approach. Reported injury threshold values of 1.31 for pia-arachnoid complex stretch ratio and 45.4kPa for principal stress at brain surface yielded the best match between FEM predictions and experimental data on piglet intracranial hemorrhaging due to IHI-ST events.

Bandak [34] estimated the limit strength of an infant’s neck based on several animal studies [44‐46] and an old study in which the force required to dissever the head was measured in a number of fetuses [47].

Besides forces and deformations exceeding injury thresholds, several studies suggest that other effects might play a role in causing injury in IHI-ST. Most notably, it has been suggested that repetitive IHI-ST events can lead to an accumulation of displacements and stresses in the eye [19, 23, 27] and in the head [6, 9, 14]. Besides this, some studies have suggested that, due to the anatomy of the head, rotational motion might be more damaging than translational motion of comparable intensity [18, 29, 33]. This is supposedly because the CSF protects the brain from sudden accelerations against the skull by providing a cushioning effect in translational motion, while the same CSF provides lubrication and hence promotes free brain movement in rotational motion. Finally, both Couper& Albermani [28] and Miyazaki [11] have pointed to the reverse rotational motion between brain and skull following chin-to-chest or occiput-to-back endpoint contacts. In Miyazaki [11] this is indicated as the most important mechanism for brain injuries.