The design intent was for the Loughborough University Surrogate Neck (LUSN) to be as representative of the 50th percentile male in as many aspects as possible. It is noted that the ‘average’ male that is amalgamated from various sources may not indeed represent any single individual. However, in line with previous surrogate developments, the intent here was to design a cost-effective solution that matches key average dimensions, measured from a large target demographic; whilst accepting that the result may be an ATD that exactly matches no living human. The assumption is that the ATD will respond in a human like manner and produce ‘average’ outcomes, as though the target demographic had been tested and the results averaged. This ‘representative’ ATD is then assumed to have relevance for the entire demographic and a cost, which is proportional to this greater utility.

A neutral posture of the 50th percentile skeletal male (stature: 1758 mm; age: 32 years) in a standing position was defined and constructed in Siemens NX CAD software using three-dimensional cervical vertebrae models adapted from Magee (2009), as the basis. The vertebral body geometry (heights and depths) and intervertebral disc spacings were matched to mean values of the 50th percentile male (Magee, 2009). The location and orientation of the second and seventh cervical vertebrae (C2 and C7) were used as key references for defining the cervical lordosis. Harrison’s Posterior Tangent model (Harrison et al., 1996; Harrison et al., 2004), considered to be the most appropriate method of indicating the level of cervical lordosis (Magee, 2009), was used to fit a circular arc to the lordotic profile of the cervical spine. Between C2 and C7, the arc length ( L ) represents the length of the cervical spine and the chord of this arc represents the height ( H ). The height is measured as the distance between the postero-inferior body of C7 vertebra to the postero-superior body of C2 vertebra and the arc length is calculated along the posterior longitudinal ligament. The geometrical spinal model is represented in Figure 1 and is defined by the following equations Eqs 1, 2: H L = Height Length = Chord Arc   length = C S = 2 R ⁡ sin θ R 2 θ = sin θ θ   , (1) 8 x + 7 2 5 x = 2 θ , (2) Where, H represents the height (chord) of the cervical spine, L represents the length (arc length) of the cervical spine and x represents the height of a vertebral body in degrees of the total lordotic arc. The total arc, known as the ARA, is defined by 2 θ and dictates the sum of the relative rotation angles (RRA) between adjacent vertebrae. The lordotic profile of the cervical spine has a summated height of the vertebral bodies and discs of 136.5 mm. The height to length ratio of 0.97 (Harrison et al., 2004) resulted in a circle radius of 159.5 mm, upon which the posterior borders of the vertebral bodies were aligned. C2-C7 sagittal vertical analysis (SVA) was used to define the global orientation of the cervical spine such that the C2 inferior posterior corner had a relative anterior translation of 3.9 mm to the C7 superior posterior corner (Magee, 2009).

The articulation of the neutral posture of the cervical spine required the application of the intervertebral ROM at the appropriate instantaneous centres of rotation (ICR) locations. The maximum reported physiological limits of intervertebral motion for a 50th percentile male were defined by the data presented in the review of Waiter (2006) and are summarised in Table 1.

ICRs are an important determinant in predicting the motion path of the cervical spine and provide a quantitative location to apply an intervertebral ROM to a superior vertebra with respect to an adjacent inferior vertebra. Motion in the sagittal plane has been the major focus of previous research and as such most studies on ICRs and ROM are in this plane. To the best of the author’s knowledge no such quantitative data exists for the lateral flexion response of the cervical spine. It was therefore assumed that the ICR locations of the vertebrae in the sagittal plane could be reasonably applied for the case of lateral flexion. Normalised values of ICR locations, accounting for differences in the sizes of individual vertebral bodies, have been reported in the literature (Amevo et al., 1991; Charlton, 2015) and are presented in Table 2. The two studies report similar normalised values of ICRs for the C2/C3 to C7/T1 segmental levels. The ICR locations of the upper cervical spine have rarely been quantified. The most complete dataset was provided by Chancey et al. (2007) who reported C0-C1 and C1-C2 ICRs with respect to anatomical landmarks of the skull. Their findings agreed with Van Mameran et al. (1992). Chancey et al. (2007) also reported bony landmark positions with respect to the external auditory meatus (EAM) location. To account for likely differences in the geometry between individual skull models, the absolute ICR measurements were normalised by calculating geometrical ratios of the skull, for example the EAM to Nasion versus the overall head length ratio was compared. The ICR locations for the upper cervical spine are also shown in Table 2.

These intervertebral ROM and ICR location values were applied to the neutral cervical spine model in a first order mode. A proportion of the total intervertebral ROM was applied such that all vertebrae and the skull, above and inclusive of the C7 vertebra were rotated about the ICR of C7-T1 by one-fifth of its total ROM. The process was repeated for all vertebrae above and inclusive of C6 and these were rotated about the C6-C7 ICR and so forth, until the final iteration involved rotating the skull about the C0-C1 ICR. The above process was then repeated for each fraction of the total ROM until all vertebrae were positioned at their end ROM. A visual representation of the orientation and location of the cervical spine and skull in its flexion mode can be seen in Figure 2.

A geometrical approach was taken to inform the number and location of joints in the LUSN’s design. The point to point distance between the C7-T1 ICR and the C0-C1 ICR was considered along the mid-sagittal plane for the cervical spine in the neutral, full flexion and full extension locations. The distance between C7-T1 to C0-C1 ICRs was found to increase and decrease for flexion and extension, respectively, such that a single segment–two joint design that spanned the two ICR locations could not satisfy the loci of motion from extension to flexion. A third joint was therefore added to the model and its location was defined by the point of intersection of two circles with radii equal to half the distance between the C7-T1 ICR and the C0-C1 ICR in full flexion (Figure 3). This location is represented by a ‘red cross’ in Figure 3. The length of segment one (S1) and segment two (S2) are therefore equal. The resulting orientation of the head (Frankfort horizontal plane) is represented by the ‘blue x-z axes’, where positive x and z are towards the anterior and inferior skull, respectively.

The LUSN constitutes three encapsulated ball-type joints with superior and inferior bodies, acting to restrict ROM of each joint through contoured mating surfaces, in combination with a plain bearing axial joint to provide a high degree of axial rotation ROM. The upper assembly consists of a truncated hemispherical ball joint and plain bearing to represent the human occipito-atlanto-axial joint (C0-C2 vertebrae). The human C0-C1 joint primarily permits flexion/extension motion whilst the C1-C2 human joint accounts for approximately 50% of the total axial rotation ROM of the cervical spine. The majority of this rotation is known to occur in the neutral zone, i.e. with low resistance to motion. The C0-C1 joint of the LUSN was therefore designed to permit the calculated ROM in flexion/extension and lateral flexion and the C1-C2 joint was limited to pure axial rotation and a ROM of 80° (40° to the left and right). The LUSN middle joint is similar to the C0-C1 upper joint mechanism but the high ROM was achieved using two mating truncated hemispheres (i.e., a ball with threaded joint), each accounting for half of the total ROM. The truncated hemispheres and bolt inserts are assembled in to the upper and lower segment bearing surfaces using threaded holes. The contours of the two bearing surfaces defined the ROM limit at the joint, whilst additional clearance was given to the truncation of the hemispheres to avoid interference of the bolts with the inner surfaces during maximal ROM. The LUSN lower joint design was an adaptation of the upper joint given the similarities in DOF and ROM. The slope of the first thoracic (T1) human vertebra was maintained on the T1 mechanical component and the flat surface acted as a ROM limit for the mating, contoured surface of the C7 component. The truncated hemispherical extrusion was constructed with its centre at the C7-T1 ICR location and the truncation allowed for maximal ROM and clearance to avoid interference of the bolt with the inner surface. Across each joint, a mixture of stainless steel and phosphor bronze were chosen to ensure low resistance at the joint. Figure 4 shows the human cervical spine CAD model and LUSN three-joint equivalent structure in combination with the biofidelic surrogate head of Miyazaki et al. (2017).

The resistive elements were inspired by the cervical spine’s longitudinal ligaments, attaching anteriorly, posteriorly and laterally across each of the three LUSN joints. An analytical solution of the force required to displace the neck a known distance was used to inform the nominal material properties required for the passive response of the neck (i.e., cross-sectional area, length and Young’s modulus). Passive bending moment responses of human volunteers (20 male volunteers, stature: 1775mm; age: 19.5 ± 1.4 years) in flexion/extension and lateral flexion have been reported by McGill et al. (1994) and these values compare well to the cadaveric cervical spine response values of McElhaney et al. (1988). The bending stiffness of the human neck is known to be higher in flexion versus extension, though the difference in maximum bending stiffness was found to be relatively small (0.128 Nm/deg versus 0.106 Nm/deg) and similar to the lateral flexion stiffness of 0.13 Nm/deg. The values reported are a valid guide but must be interpreted with caution given the disparity in response reported both inter and intra respectively for human samples.

The geometry of the neutral posture was used to define the coordinates of the elastic element attachment locations, in order to determine the original length and perpendicular distance of the line of action of the resistive element with respect to the joint centre location. The total angular displacement of the head relative to the torso, as reported in McGill et al. (1994), was proportionally split across the three joints relative to their total possible ROM. The three joints were then analysed in isolation such that the respective angular displacement value was applied to the joint. The joint was considered to be in equlibrium at this moment in time as though the total moment, M , as reported in McGill et al. (1994) was applied to each joint. The 3D CAD (Siemens NX 8.5) model was then used to determine coordinate data for the new locations of the element attachments and Pythagoras’s theorem was used to calculate the elongation of the posterior and/or anterior elastic element. The force exerted by the anterior and posterior elements across each joint was then calculated for the respective bending mode.

The material stiffness characteristics of woven Polyester elastic (Stretchline Limited, width: 25mm, thickness: 3 mm) samples were quantified through tensile testing on an Instron Dynatup 9250HV machine. A nominal length between the loops was defined as 30 mm (approximately the length of the anterior middle joint elastic element). The loops were stitched to ensure that a standard sized (8 mm diameter) Instron testing pin could be used to secure the sample. A tensile loading profile was created such that an absolute ramp rate of 500 mm/min (maximum capacity of the machine) was applied until a load of 200 N was reached. The sample was then unloaded to its original length and the cycle was repeated 100 times. The sample was visually inspected after the loading protocol and the stitching was assessed for signs of damage. From the theoretical calculations, the maximum elongation of the elastic elements was predicted to be 43%, 38% and 22% at the top, middle and bottom joints, respectively. The stiffness modulus of the elastic sample was therefore predicted in the range of 0.2–0.6 mm/mm strain corresponding to a maximum of 18 mm elongation. An assessment of the stress versus strain gradient within this region resulted in a predicted modulus of 6.86 MPa, which is in agreement with the required modulus from the theoretical prediction.

The Polyester elastic was cut into lengths defined by the CAD model and closed loops were formed at both ends with a bartack zigzag stitch. The elastic was either directly attached to a LUSN via a slotted plate and then stitched or attached retrospectively via a pin through the loop of the elastic. The configuration allows for lengths to be changed and therefore pretension of the elastic elements to be easily modified, which in turn can be used to define a non-neutral initial posture. The axial rotation plain bearing assembly is independent of the upper resistive elements and therefore is still able to rotate with a negligible resistance to motion, whilst the torsional stiffness of the lower neck elastics resist further axial rotation. Figure 5 shows the assembled neck in the neutral posture with elastic elements attached, along with annotations for clarity.

The degree of biofidelity is defined as the ability of the LUSN to appropriately represent available data on the 50th percentile male (e.g., mass, range of motion and bending moment response). Specifically, the surrogate neck, when constrained at the lowest most component (representative of the human T1 vertebra) and subjected to a pure bending moment or axial torque at the upper most component (representative of the base of the skull, referred to as C0), should rotate nominally to an angle that is in agreement with that reported for the human in an equivalent load case. To quantify the biofidelity with respect to the definition, two separate experimental investigations were conducted, like those described in the literature (Luck et al., 2014).

The total mass of the surrogate neck considered all components except the T1 components (T1 body, T1 pins and T1 bolt insert). In-situ these components are rigidly mounted to the thorax and contribute 0.44 Kg of mass. Kang et al. (2016) reported mass and inertia of the upper (C0-C4) and lower (C4-C7) cervical spine. The equivalent upper and lower split on the LUSN is at the centre of the mid ball joint. A comparison of the total and upper and lower distributed mass is presented in the results. Estimates of the human cervical spine’s principal moments of inertia have rarely been reported, and those that have provide a wide range of values, due to differences in measurement methodologies and rigid body assumptions. For comparison, the theoretical inertial properties of the LUSN were calculated in the CAD software. The I_xx, I_yy and I_zz, account for moments of inertia about the sagittal, coronal and transverse planes, respectively.

Based on the methodology adopted by Luck et al. (2014) and originally presented by Nightingale et al. (2002), the quasi static pure bending response of the LUSN was quantified. A loading frame was additively manufactured from ABS and rigidly mounted to the top of the C0 skull attachment plate of the LUSN. Two lengths of lightweight nylon cable were attached to the loading frame and masses were added at a constant arc guide of diameter 0.28 m. The cable attachments were located equidistance apart to ensure that equal and opposite forces were applied to the LUSN. The pulleys were lightweight with low inertial resistance and were positioned such that the vertical distance between the upper and lower pulleys were maximised to ensure that the force application resulted in rotation of the LUSN and minimised translation throughout the entire ROM. The bending rig mounting plate was designed to allow for the LUSN to be rotated 360° and rigidly mounted in four positions (90° increments), such that the application of the bending moment resulted in flexion and extension (about y-axis) and left and right lateral flexion (about x-axis). An annotated schematic of the setup is shown in Figure 6A.

In each of the bending orientations, the neutral posture of the surrogate neck was checked prior to load application and a calibrated digital protractor was used to ensure a consistent repositioning of the surrogate neck between trials. In the neutral posture, a camera (resolution 1024 x 640 pixels) was used to capture a calibration image of the LUSN after each repositioning, which would serve as reference images, from which the angular displacement of the neck would be calculated for each load case. The calibration images were captured with a 30 cm steel rule in view and this allowed for a pixel-to-mm calibration factor to be calculated. Additionally, retro-reflective circular markers were located on rigid arms of the lightweight loading frame to be tracked during post-processing.

Prior to testing, the LUSN was preconditioned with 10 cycles of 2.75 Nm applied bending moment. The masses (0.25 kg to 2.75 kg) were then added to each cable incrementally using a load, hold and unload approach, resulting in an applied moment of 0.69 Nm—7.55 Nm. After being secured to the cables, the masses were lowered slowly until the cable became taut and was then released from the hands of the tester. An image was captured 30 s after the application of the load to account for creep within the structure. The load was then removed, and the neck was placed back in the neutral posture before a calibration frame was captured and a subsequent load was applied. The full loading process was repeated five times in each of the principal loading orientations (about the x and y–axes) of the LUSN. Exemplar images of the sagittal and lateral loading cases are shown in Figures 6B,C, respectively.

Manual image processing was conducted using Image Pro Analyser software. The intra-tester repeatability of the testing was investigated first by quantifying the accuracy of the repositioning of the neutral posture of the LUSN and secondly by quantifying the accuracy of the digitisation of landmarks in the image analysis software. To investigate the repositioning of the LUSN, the landmarks of the lightweight fixture and features on the lower neck segments were digitised across each of the calibration images. A three-point circle feature was used to create a circle for each of the landmarks to quantify the coordinates of the circle’s centre location and the diameter of the circle. The accuracy of the landmark digitisation was investigated by selecting one calibration image from each bending mode and repeating the digitisation of four landmarks from each image, five times. The landmark data for both investigations were compared, and the mean and standard deviation values were calculated.

A bespoke axial torque test rig (Figure 7) was developed to investigate the angular displacement response for axial torques applied about the z-axis. A simplified and manual bespoke test rig was constructed to quantify the axial torque response of the LUSN about the z-axis, based on the test setup used by Myers et al. (1989). A lightweight hex bar was attached to a rectangular top plate and bolted to the top of the LUSN, guided through a closed linear bearing. The sub-assembly was free to axially rotate and translate with negligible resistance to motion. The ability to axially translate was important due to the change in length of the neck during axial rotation. The constraint against this natural motion artificially increases the axial stiffness of the cervical spine (Myers et al., 1989). The torque was applied manually by the tester using a torque wrench (Kennedy Tools; model TW4) with a range of 0.3 Nm—4 Nm, a resolution of 0.1 Nm and accuracy of ±3% via the lightweight hex bar.

A low-speed loading rate of between 25 ( ± 5) degrees per second, was utilised for the battery of tests on the LUSN. The loading rate was equivalent to the loading procedure used by Luck et al. (2014) to quantify the characteristics of the Thor-M and Myers et al. (1989) to quantify the responses of PMHS. Furthermore, the H3SN data presented by Myers et al. (1989) was acquired with a loading rate of 30° per second. Therefore, the effect of loading rate on the response of the LUSN was minimised and ensured that a comparable dataset was collected. Prior to data collection, the LUSN was preconditioned with 10 cycles of axial rotation from 0 to 50°.

A high-speed video camera (Photron Fastcam SA1.1) recorded the application of the torque and was focused on the dial gauge of the torque wrench and also the lightweight, rigid markers on the upper neck attachment plate. The frame rate of the camera was 240 FPS to allow for analysis of the applied torque and angular displacement response. In addition, the applied loading rate was quantified by investigating the change in angular displacement of the torque wrench across frames of the high-speed video, ensuring that the torque was applied consistently. A total of ten trials were collected and analysed for the axial torque response.

The resonant frequencies of the human skull have previously been investigated using computational and experimental techniques to further the understanding of dynamic loading effects on the risk of head and brain injury. Ruan and Prasad (1996) used a finite element head model to investigate the effect of boundary conditions and skull thickness on the resonant frequency response of the head. Their results implied that by applying the boundary conditions at the lower neck instead of at the atlanto-occipital joint (upper most neck joint), the frequency content at each mode was reduced, with a significant effect at the lower modes (one to three). It is assumed that this would be due to an increase in the effective mass of the system. Khalil and Viano (1979) investigated the modal response of a single dry 50th percentile male cadaver skull, using a force hammer to excite the skull in several locations. 11 resonant frequencies were identified and the lowest was reported as 1385 Hz. Hakansson et al. (1994) conducted in vivo testing using human volunteers and found between 14 and 19 resonant frequencies for each of the six volunteers (2 males and 4 females). The lowest mean resonant frequency was 972 Hz and the standard deviation was ±119 Hz.

The dynamic loading of high-speed, short-duration impacts (e.g., baseball pitches) are likely to excite a wide range of frequencies. A biofidelic surrogate neck should not inhibit these responses before their influence on the resulting head response is understood. Therefore, the effect of surrogate neck constraint on the modal response of a biofidelic surrogate head (Miyazaki et al., 2017) was investigated. The modal response of the surrogate head was investigated under three constraints:

1) Freely suspended via lightweight bungee cords (typical constraint for laboratory high-speed projectile impact testing),

The surrogate head was fully assembled to contain all surrogate components with exception of the skin, which is known to dampen the frequency response of the skull and increase the difficulty of mounting an accelerometer. The head was constrained by one of the three stated conditions and a force hammer (Bruel and Kjaer, Type 8206-001) was used to apply an excitation load at locations around the surrogate head. The response was measured using a triaxial accelerometer (DJB Instruments, Type AT/10-6), mounted to the contralateral side of the surrogate head with synthetic wax and recorded using the Siemens LMS Test Lab 15.0. The excitation and measurement locations are summarised and visualised in Table 3.

Table 4 presents the comparative percentage distribution of the total ROM across the joints of the LUSN and the equivalent summation across the human intervertebral joints (Waiter, 2006) in flexion/extension and axial rotation/lateral flexion, respectively. The LUSN and the cervical spine model were articulated through to their end ROM in flexion, extension and lateral flexion and the coordinates of the C0-C1 ICR were compared. The LUSN, using only three joints with distributed ROM, precisely matches the location at the end ROM and can closely approximate the predicted human loci of motion when considering the articulated human vertebral model. The maximum horizontal and vertical offset differences of the C0-C1 ICR were found to be 0.61 mm, 1.43 mm, 1.14 mm, and −0.31 mm, −0.20 mm, 0.05 mm for 50% ROM in flexion, extension and lateral flexion, respectively.

Figure 8 presents rendered overlay images of the LUSN in full extent of ROM with respect to the neutral posture, in each of the principle planes.

The total mass of the LUSN is 1.62 Kg and has agreement with the total mass of the 50th percentile human cervical spine (1.6 Kg) (Kang et al., 2016). The total and distributed mass and moments of inertia, when considering an equivalent upper and lower cervical split, are presented in Table 5.

Figures 9A–C presents the mean angular displacement and 95% confidence intervals (CI) for each bending load case calculated for the LUSN. The results are presented as best-fit polynomial curves with an intercept of (0,0) and therefore represent an interpolation of the 11 measured angular displacements. For clarity, the 11 applied moments and corresponding angular displacements are presented as markers on the curves. The applied axial torque with resulting mean axial rotation and 95% CIs are presented in Figure 9D. Additional datasets (e.g., cadaver) have been interpreted as a series of x, y coordinates from those graphically presented by the respective authors and then interpolated using polynomial best-fit curves. Where possible, these results have been presented up to the maximum tested on the LUSN (7.5 Nm for bending modes and 3 Nm for axial rotation). The Thor-M surrogate neck and Duke Adult Head and Neck Model (DAHNM) response curves (bending and axial rotation) were interpreted from the quasi-static load cases with the necks in their no muscle cables and relaxed muscle states, respectively (Luck et al., 2014). The flexion, extension and lateral flexion H3SN and cadaveric data sets were interpreted from those presented by McElhaney et al. (1988) under tension-bending load cases. The axial rotation H3SN and cadaveric response curves were interpreted from the constant velocity tests of Myers et al. (1989). Additional response curves of human volunteers in the passive neck state are presented for the bending modes (McGill et al., 1994).

The bending moment response of the LUSN was found to be repeatable with a maximum deviation in angular displacement of ±0.77° across the applied loads. The higher loading cases generally resulted in a slightly higher measured deviation in angular displacement between the five repeats. The axial rotation response of the LUSN was also found to be highly repeatable across the range of applied axial torques. The maximum standard deviation (±1.95°) occurred at an applied torque of 2 Nm, where the range of axial rotation was 56–61°. Whilst the LUSN exhibits a highly repeatable response throughout the tested range, it is not possible to quantify this repeatability with respect to other surrogate necks, as the data has not been presented in the literature.