Article Text

The challenges in developing a finite element injury model of the neck to predict the penetration of explosively propelled projectiles
  1. Johno Breeze1,2,
  2. T Newbery3,
  3. D Pope4 and
  4. M J Midwinter1
  1. 1Academic Department of Military Surgery and Trauma, Royal Centre for Defence Medicine, Birmingham, UK
  2. 2Biomedical Sciences Department, Defence Science Technology Laboratory, Salisbury, Wiltshire, UK
  3. 3Land Battlespace Systems Department, Defence Science Technology Laboratory, Sevenoaks, Kent, UK
  4. 4Physical Sciences Department, Defence Science Technology Laboratory, Salisbury, Wiltshire, UK
  1. Correspondence to Maj Johno Breeze, Academic Department of Military Surgery and Trauma, Royal Centre for Defence Medicine, Birmingham Research Park, Vincent Drive, Birmingham B15 2SQ, UK; johno.breeze{at}


Introduction Neck injuries sustained by UK service personnel serving on current operations from explosively propelled fragments result in significant mortality and long-term morbidity. Many of these injuries could potentially have been prevented had the soldiers been wearing their issued neck collars at the time of injury. The aim of this research is to develop an accurate method of predicting the resultant damage to cervical neurovascular structures from explosively propelled fragments.

Current status A finite element numerical model has been developed based on an anatomically accurate, anthropometrically representative 3D mathematical mesh of cervical neurovascular structures. Currently, the model simulates the passage of a fragment simulating projectile through all anatomical components of the neck using material models based upon 20% ballistic gelatin on the simplification that all tissue types act like homogenous muscle.

Future research The material models used to define the properties of each element within the model will be sequentially replaced by ones specific to each individual tissue within an anatomical structure. However, the cumulative effect of so many additional variables will necessitate experimental validation against both animal models and post-mortem human subjects to improve the credibility of any predictions made by the model. We believe this approach will in the future have the potential to enable objective comparisons between the mitigative effects of different body armour systems to be made with resultant time and financial savings.


Statistics from

Key messages

  • Numerical injury models have the potential to complement physical and animal models in the future.

  • An anatomically accurate 3D finite element model of cervical neurovascular structures is currently being developed.

  • Currently, the model simulates the penetration of fragment simulating projectiles into all tissue types using algorithms based on 20% gelatin.

  • Experimental comparisons with both animal models and post-mortem human subjects are planned to validate the material properties chosen.

  • This model has the potential in the future to enable objective comparisons between the injury mitigative effects of different body armour systems to be made.


Neck injury due to explosively propelled fragments experienced by UK service personnel deployed on current operations is responsible for significant mortality and long-term morbidity.1 ,2 The mortality of a penetrating neck wound from a penetrating explosive fragment is 42%; in addition, 16% of survivors have significant long-term complications such as hemiplegia secondary to stroke. Between January 2006 and December 2010, 94% of deaths from penetrating explosive fragments to the neck were from neurovascular damage (spinal cord, carotid artery, internal jugular vein or vertebral artery).1 Post-mortem analysis suggested that 16 deaths could potentially have been prevented had the soldiers been wearing their issued neck collars at the time of injury.2 However, no soldier during this 5-year period was wearing these neck collars at the time of injury, reflecting a general dislike of their design.

There is a desire to refine current methods of ballistic cervical protection and potentially develop new ones in order to improve user acceptability; however, this necessitates the manufacture of multiple designs of prototypes2 each of which require human factors assessments to determine their acceptability for performing representative military tasks. Numerical simulation of injury to the neck has the potential to complement other methods of body armour assessment, namely, ergonomics trials and ballistic protective materials testing. Such trials are costly both financially and in terms of time. The ability to rule out a particular design of personal protective equipment on medical grounds prior to ergonomics assessment would enable other designs to be tested instead with resultant time and financial savings. Little information has been published in the open literature on existing numerical injury models used by the UK Ministry of Defence or its allies to date,3 ,4 but a brief summary on what is available is included in the supplementary online material. Existing numerical simulations of penetrating injury that include the neck region are not suitable for this analysis as they do not model the individual anatomical structures of the neck, with most even combining the neck and head as a single anatomical unit despite their fundamental anatomical differences.3 ,5–7 The aim of this research is to develop an accurate method of predicting the damage to individual cervical neurovascular structures from an explosively propelled fragment. This will potentially enable objective comparisons to be made between the mitigative effects of different body armour systems once they have been incorporated into the model.

Concept of the model

The neck model has been constructed using a finite element approach in which the cervical anatomical structures are represented as parallepoid blocks termed ‘elements’. Each element is assigned a ‘material model’, which can be thought of as a set of algorithms that represent the specific biomechanical responses of that individual tissue or material under ballistic impact. This finite element model will therefore be based upon the complex interaction of the finite elements representing the cervical anatomical structures with the elements representing the projectile as it passes through them. The geometric representation of the anatomical components within the neck was originally derived from a commercially procured 3D mesh (V.5.0 Male Human Anatomy, Zygote media group, American Fork, Utah, USA). This triangulated mesh had been generated from coordinates derived by CT and MR scans and accurately described the surfaces of anatomical structures down to a resolution of 0.25 mm. The mesh was subsequently converted at Dstl Porton Down into discrete finite elements (Figure 1), and each element ascribed a material model. An expanded explanation of the concept behind this finite element approach towards modelling the penetration of a high velocity projectile through tissue can be found in supplementary online material.

Figure 1

Triangulated surfaces of high fidelity anatomical mesh prior to filling procedure (A) and after utilisation of eulerian mesh (B).

Geometric representation and scaling of cervical anatomical structures

The triangulated mesh of cervical anatomical structures was based upon coordinates derived from CT and MR scans, but it was not possible to ascertain from the commercial developer on who the original scans were undertaken raising concerns that the size of the neck and the structures within it were not representative of the military population whose injuries this model would predict. Publications such as the UK Defence Standard (DefStan)8 and NATO STANdardising AGreements (STANAG)9 have traditionally provided standardised anthropomorphic measurements for military-specific populations, but they provide no greater detail than a range of neck circumferences. In addition, the original Zygote mesh only provided coordinates for the surfaces of anatomical structures and did not describe their inner structures. For example, it did not show the thickness of an arterial wall or delineate a cervical vertebra into inner cancellous and outer cortical bone, each of which has different biomechanical properties.

To provide these data, a review of CT angiograms performed on UK service personnel was undertaken,10 which was able to provide military-specific cervical anthropometric data such as vessel wall thickness and distances from vessels to the skin. The cervical anatomical structures within the mesh were subsequently scaled to that representative of a 50th percentile male UK service person, meaning no further adjustments had to be made. An additional unpublished review of these same CT scans demonstrated that for the purposes of our model, the cortical bone of cervical vertebrae should be meshed with a thickness of 0.50 mm with the remainder represented as cancellous bone. These findings were in keeping with cadaveric studies of cervical vertebrae which suggested that cortical shell thickness varies between 0.44 and 0.89 mm.11 ,12

Currently, the model only incorporates those cervical structures that have been shown to be directly associated with predicting mortality from explosive fragmentation, namely, skin, bone, arteries, veins and nerve components (Table 1). The high fidelity of the originally procured mesh however will potentially enable incorporation of further structures in the future, such as those associated with significant long-term morbidity rather than mortality such as the brachial plexus and larynx (Table 1).

Table 1

Essential and desired cervical anatomical structures requiring modelling as determined by clinical morbidity and post-mortem mortality analysis2

Material model representation

Each element within the model must be assigned a ‘material model’, which is a set of algorithms that represent the specific biomechanical responses of that individual tissue under ballistic impact. For example, a blood vessel could, at its simplest, be considered as a cylindrical tube of a single tissue type (requiring a material model to represent it) surrounding a single type of fluid representing blood requiring a second material model. The fidelity of the model can be increased by representing the blood vessel wall in its true three individual layers instead of a single homogenous layer; however, each layer in turn will require its own material model to represent its individual biomechanical properties, greatly increasing the complexity of the model. A basic material model for each tissue type requires a value for the density of the material as well as two additional types of algorithms. The first describes the ‘strength’ of the material and represents strain versus stress in different directions; the second is the ‘equation of state’ (EOS) of the material and represents how pressure develops under a given level of hydrostatic compression as well as any accompanying change in internal energy due to such deformation.

Ascertaining the values experimentally required to populate these material models is still highly challenging13 as the high compressive strain rates (100–2500 s−1) and large deformations characteristic of typical impact scenarios require a fresh sample of each tissue type using techniques that have only been developed relatively recently.14 ,15 A review of the open literature prior to the commencement of this project demonstrated very limited original experimental data14 ,16 ,18–30 from which to derive these material models (Table 2). Further research is currently directed at deriving experimental values for these EOS, which can replace the existing estimations, thereby refining the fidelity of the model.

Table 2

Material models used to represent anatomical structures within the latest iteration of the finite element neck model.

Convergence study

In an analysis of this type, it is important to use elements of a suitable diameter to maximise the accuracy of its predictions. As the size of each element decreases (thereby increasing its fidelity) the accuracy of the model potentially increases. However, the representation of such complex geometries at high fidelity requires significant demands on computing power, necessitating practical constraints to be accepted so that it remains useable. A convergence study (Figure 2) aimed to ascertain the most appropriate element size when modelled in two dimensions, that is, the point at which a change in resolution does not affect the predictive output of the model. Convergence for the velocity of the projectile occurred with an element size of 0.5 mm, as demonstrated by overlapping lines in the figure. However, the neck must be modelled in three dimensions, and this additional dimension hugely increases the total number of elements required to perform the analysis. Currently, the model uses 1 mm wide elements resulting in a total of approximately eight million elements. Although the computing power required to perform the large number of calculations required for a 3D model with element sizes of 0.5 mm does exist in the UK, it is currently limited to a small number of institutions. Access to one such source has recently been agreed with the eventual aim to be able to undertake such calculations at Dstl Porton Down following a capability review in the near future.

Figure 2

2D mesh sensitivity convergence study using a 6 mm diameter steel sphere with an impact velocity of 300 m/s. The material model derived for 20% gelatin has its maximum correlation to that found experimentally when the size of each individual element is 0.5 mm or less.

Incorporation of projectiles and body armour components

Fragment dimensions will be based on internationally standardised fragment simulating projectiles (FSPs),9 enabling comparisons with existing animal and ballistic simulant-derived experimental data. The primary FSP will be the chisel-nosed cylindrical 1.10 g FSP, which remains the internationally recognised benchmark for experimental comparisons of body armour protective materials.31 ,32 This particular FSP however was originally chosen as it was the most common fragment produced by one particular World War I ordnance casing and therefore may not be representative of the fragments produced by contemporary explosive devices.32 Recent analysis of CT scans performed on UK soldiers with penetrating explosive fragments retained in their necks has suggested that the smaller 0.49 g cylindrical FSP may instead be a more representative testing projectile for ballistic protective materials to protect the neck.31 In addition, a spherical FSP will be incorporated, which is potentially more representative of pre-formed munitions, such as from mines or rocket propelled grenades.31 The dimensions of these projectiles are accurately described in the FSP standardising agreement,9 enabling meshed constructs to be easily imported into the model (Figure 3) and subsequently converted into finite elements. By using a 3D scanner, physical objects, such as the ballistic neck collar component of body armour (Figure 3), can also be imported and allow the model to be used for design assessments.

Figure 3

Meshed images of a chisel-nosed cylindrical fragment simulating projectile and current short OSPREY neck collar prior to importing into the model.

Quantification of the wounding potential of projectiles within the model

Previous numerical simulations of injury have generally used an infinitely thin shot-line to determine the path of wounding (Figure 4A) with any anatomical structure along this line assumed to be damaged. Another suggested method has been to use a cylindrical tract of destruction with a same width as that of the projectile (Figure 4B). However, neither of these two methods reflects the complexity of the permanent wound tract (PWT) found in tissue which is actually required in the model (Figure 4C); this is determined by both projectile (eg, size, shape, effect of tumbling) and tissue factors (eg, variations in density between anatomical structures, tissue planes). A systematic review of the wound ballistics literature between 1902 and 2012 clarified that the PWT comprises a central permanent wound cavity surrounded by an area of irreversibly damaged tissue, each of which is formed by different mechanisms.33

Figure 4

Graphical comparison of different potential mechanisms of quantifying projectile wounding effects: (A) Infinitely thin shot-line, (B) Cylinder of variable length but using width of projectile and (C) Permanent wound tract.

Accurately determining the dimensions of the PWT in tissue for a variety of projectile shapes and impact velocities is challenging. The aforementioned ‘biological variation’ inherent to such testing means that large numbers of animal experiments must be undertaken just to provide a small amount of statistically valid information on just a single projectile. The most promising potential approach identified in the systematic review was based on research undertaken in the 1970s and used the mass of tissue that required debridement by a surgeon following wounding.33 Such a method would inherently account for both projectile factors and tissue factors; however, the experimental results produced algorithms describing the line of best fit with such poor correlation that this approach cannot be used with the existing limited data set alone, which would necessitate further testing.

The current iteration of our model uses PWT dimensions based on the permanent cavity produced in 20% gelatin (Figure 5). This tissue simulant enables the incorporation of projectile factors and produces depths of penetration using both 0.49 and 1.10 g cylindrical FSPs comparable with that found in goat34 and pig muscle.35 Measurements made using high speed photography of these FSPs penetrating gelatin will also enable the dimensions of the temporary cavity to be incorporated into the model. However, evidence would suggest that although the temporary cavity size is comparable in magnitude with that produced in muscle, the permanent cavity dimensions are likely to be less than that in live tissue, as gelatin has a greater tendency towards collapse. Such an approach will therefore invariably underestimate tissue damage and we would encourage further research to establish methods to quantify the dimensions of the PWT based on a more biofidelic model.

Figure 5

Measuring wounding parameters produced by a projectile experimentally penetrating 20% gelatin (A) compared with that predicted by the numerical model (B); Permanent cavity width (1), Depth of penetration (2) and Projectile width (3).

Future research and potential methods of model validation

A working anatomically accurate 3D finite element model of an FSP traversing the neurovascular structures of the neck has been developed (Figure 6). As the material model based on 20% gelatin currently used by each element in the model is replaced by one specific to each individual tissue type, the model as a whole will clearly require validation. The use of freshly killed animal surrogates in conjunction with pre-ballistic and post-ballistic firing CT scans will enable a numerical model of this surrogate to be built. Testing is most likely to be undertaken with goats or pigs as these animals have the greatest amount of evidence for ballistic testing to date; the skin of goats in particular is believed to be closest to human in terms of projectile retardation.34 Animals would be slaughtered humanely using a Schedule 1 method as approved by the Home Office with testing starting as soon after death as possible to ensure that the material properties of the tissues through which the projectile passes are as close to that of a live subject as possible. There is no evidence describing the effect on projectile penetration as tissues age after death; however, it is likely that the process of rigor mortis, which increases the rigidity of muscle tissues in humans 3–5 h post-death, would in some manner affect its material properties. The time to initiation and effect of rigor mortis on muscle is also believed to vary by breed and anatomical location as well as the presence of stress pre-mortem and refrigeration post-mortem;36 it will therefore be of great importance to clearly document variables such as breed, storage conditions and times between death and firing commencing. The elements within the model will be populated with the material models that are identified to be specific for each tissue type, and the actual results can be compared with that predicted by the model.

Figure 6

A screenshot of the latest iteration of the model demonstrating the passage of a cylindrical fragment simulating projectile through the neck: spinal cord (A), muscle (B), vertebra (C), internal jugular vein (D) and common carotid artery (E).

No animal model has the 3D cervical anatomy representative of humans. The use of fresh frozen and refrigerated post-mortem human subjects (PMHS) for ballistic research37 has presented a new opportunity to potentially validate the neck model as their anatomy will be broadly representative. However, little objective evidence exists as to the effect of decomposition, refrigeration and freezing on tissue material properties,38 especially to ballistic impacts, such that this method cannot be used alone. An experimental trial of FSP penetration into the necks of PMHS is planned and will hopefully provide further validation of the finite element model.


A working, anatomically accurate 3D finite element model of an FSP traversing the neurovascular structures of the neck has been developed. Currently, the huge computing power required to perform the large number of calculations required for a 3D model with optimal element sizes limits its application; however, we are confident that within a few years computer processing and memory capabilities will enable simulations to be run in multiple establishments held by the Ministry of Defence.

Currently, the model simulates the penetration of an FSP into all tissue types using elements coded with a material model based on 20% ballistic gelatin on the simplification that all tissue types act like homogenous muscles. As the material models are replaced by ones specific to each individual tissue within each anatomical structure, the model as a whole will require validation. This will necessitate experimental comparisons with both animal models and PMHS to validate the material properties chosen. Although the model is currently at a stage where it is too early for any conclusions to be made, we believe its eventual iteration will have the potential to enable objective comparisons between the mitigative effects of different body armour systems to be made. Such a comprehensive approach towards developing an accurate injury model of the neck has never previously been attempted and we believe the considerable effort required in the construction of the model will be offset by the resultant time and financial savings in body armour selection.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors Guarantor of this work: JB. Conceived of and designed the numerical model: JB, DP and TN. Searched the literature: JB, DP and TN. Preparing and editing the manuscript: JB, DP, TN and MM.

  • Funding None.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.