Introduction

The spinal column is an integral part of the human body allowing trunk flexibility. The spine is subject to normal age related process but also to certain disorders such as degenerative disc disease (DDD, accelerated aging) and disc prolapse [1]. The prevalence of a disc prolapse increases with age, occurring in 1–3 % of the population, and is highest in the age group 30–60 years. Incidence of prolapse is highest in the lower lumbar or lumbosacral spine (>90 %) [24]. This region is subjected to high biomechanical loads, which presumably account for the development of disc prolapse [1, 5]. It is still debated whether DDD leads to or protects against disc herniation [1]. There is also debate as to whether they are even interconnected [6].

Since the 1950s several studies have investigated the failure mechanism of the intervertebral disc covering the whole range from quasi-static pure compression to more complex cyclic loading situations: The first pioneers were Virgin [7], Hirsch [8], Brown et al. [9], Hardy et al. [10], and Roaf [11], followed by Farfan et al. [12] and Lin et al. [13] in the 1970s using uniaxial testing devices. In the 1980s, Wilder et al. [14, 15], Liu et al. [16, 17], as well as Adams and Hutton [1822], further improved the testing techniques using special fixture designs that allowed combinations of compression with torsion or flexion with lateral bending. Later tests by Callaghan and McGill [23] included a second dynamic axis, e.g., used for flexion/extension. Regarding all aforementioned tests, dynamic loads were applied to a maximum of two degrees of freedom (DOF). In a study by Drake et al. [24] in addition to the dynamic loads of two DOF a static load is applied to a tertiary DOF.

It was shown that loading rate and complex loading condition greatly influence the mode of failure of the intervertebral disc [2527]. However, existing test apparatuses are either designed for a maximum of two independently controlled DOF with high loading rates [23] or six DOF with low loading rates which allow quasi-static flexibility measurements of spinal motion segments [2838]. Only two dynamic systems with six DOF are known to the authors: a robotic system by Fujie et al. [39], later used for spinal segments by Hurschler et al. [40] and a hexapod system by Ding et al. [41], designed to be suitable for all biomechanical joints and tissues, including spine [42]. The robotic system (KR 16, Kuka Robotik GmbH, Gersthofen, Germany) used by [40] is able to achieve very rapid motion ranging from 156 to 614°/s depending on the axis, while loads are limited to about 160 N. The custom designed hexapod system by Ding et al. [41] is capable of very high loads (up to 21 kN and 2.5 kNm). The only limitation is the movement speed, which is between 60°/s (lateral bending) and 135°/s (axial rotation).

The cause of disc herniation is still not well understood. It is assumed that heavy lifting and extreme postures can cause small injuries starting in the inner anulus [22]. Such injuries are accumulated over years until its structure is weakened and finally a single loading event leads to a sudden failure of the last few intact lamellae. However, finite element studies suggest that complex loading also leads to high shear strain [43]. Therefore, failure may also start from the outside close to the endplate working its way into the disc until nucleus material can be extruded through this channel.

A dynamic disc-loading simulator was developed and validated to allow complex loading in order to provoke these kinds of failures. The requirements for this apparatus were to allow:

  • Pure loads in all six degrees of freedom [flexion/extension (FE), lateral bending (LB), axial rotation (AR), laterolateral shear (SLL), anteroposterior shear (SAP), axial compression/decompression(AC)],

  • Any load combination,

  • High dynamic movements [±22.5° at 5 Hz (equal to 707°/s) for rotation and 100 mm/s translation],

  • Angle/position control, and

  • Loads up to 2 kN and 40 Nm.

After calibration and validation of this machine, a first test series was performed to investigate whether mechanical failure or even disc herniation can be provoked with healthy discs.

Methods

Disc-loading simulator

The general set-up of the dynamic disc-loading simulator is similar to the conventional material testing devices: a crosshead with two clamping blocks is guided by one column each (Fig. 1). The shafts are rigidly fixed to a baseplate that is mounted on a table. Overall dimensions of the machine are 1135 × 735 × 1655 mm (width × length × height) and weighs approximately 200 kg.

Fig. 1
figure 1

CAD drawing (left) and photograph (right) of dynamic disc-loading simulator with position of the coordinate system. 1 Crank handle, 2 swing, 3 counterweight, 4 angle transducer, 5 crosshead with clamping blocks, 6 specimen, 7 XYZ-table, 8 electric drive, 9 column, 10 six-component load cell, 11 path transducer (laser)

The crosshead is cardanic, allowing rotation about two independent axes. An XYZ table can produce independent translation along each of the three axes and provides the final rotational DOF. The construct was built primarily from aluminum to reduce weight and thus the inertial forces and moments. Non-moving parts of the frame (shafts, clamping blocks), that are crucial for stability, as well as some minor parts, were made from stainless steel. The specimen is aligned to the centre of the machine’s coordinate system defined according to ISO 2631 and given by the position of the crosshead [44, 45] (Fig. 2).

Fig. 2
figure 2

Detailed information about the movements applicable to the specimen and the machine’s coordinate system

Movement of the crosshead around the y axis produces flexion/extension and is driven directly. The swing (LB) is moved by a lever arm connected to the drive shaft of the motor by a cardan joint. AR is driven directly. Translation of the XYZ table in the x- (SAP) and y-directions (SLL) is coupled to the corresponding drive shaft by a lever arm. Translation in the z direction (AC) is transferred by a spindle drive (EPB 2505, THK, Ratingen, Germany) with a thread pitch of 5 mm. All translational movements are guided by linear bearings. The axes of all DOF are connected to the corresponding drive shaft by backlash-free couplings (JAW, Orbit, Wolfenbüttel, Germany). A counterweight was added to the crosshead to balance the FE axis.

All translational and rotational movements are driven by electric motors (Baumüller, Nürnberg, Germany) with a torque of 6–29 N m. To increase the maximum torque (29–224 Nm), gears (Neugart, Kippenheim, Germany) were added to all axes except the translation in z direction. A “b maXX 5000”-system power supply and driver (Baumüller, Nürnberg, Germany) is used.

The rotational stroke is ±22.5° for all axes and measured by angle transducers (Kübler, Villingen-Schwenningen, Germany) with a range of 45° and a resolution of 12 bit (0.01°). For safety reasons, the software limits were set to ±20.5°. The translational stroke is ±7.5 mm for both the laterolateral and anteroposterior directions and ±19 mm for axial decompression/compression. Software limits were set to ±6.5 and ±17.5 mm, respectively. Three Laser sensors (ODSL 9, Leuze, Owen, Germany) with a resolution of 0.01 mm are used for measuring the translations.

Reaction forces up to ±10 kN (z axis) and ±2 kN (x-/y-axis) as well as moments up to 100 Nm (all axes) are measured by a six-component load cell (GTM, Bickenbach, Germany) placed directly beneath the specimen. Load and position are controlled by a Flextest-40-controller (MTS, Eden Prairie, MN, USA) that generates a ±10 V signal as input for the cascaded actuator controller (Fig. 3). Detailed information about all components used for the apparatus is given in Table 1.

Fig. 3
figure 3

Cascaded control loop configuration of the apparatus

Table 1 Detailed information about the main components used for the dynamic disc-loading simulator

Mounting of spinal motion segments with a height of up to 120 mm (including embedding) is possible. To align the centre of the intervertebral disc of the functional spinal unit (FSU) with the global coordinate system of the machine, the crosshead can be raised or lowered 100 mm using a crank handle. In addition, the swing height can be adjusted by 50 mm independent of the crosshead.

Operation of the machine and programming of arbitrarily complex loading protocols are done using commercial software (Basic Testware/Multipurpose Elite, MTS, Eden Prairie, MN, USA).

Position control analysis

A spring of defined mechanical properties (86/1/2, Knoerzer, Pfullingen, Germany) was used as surrogate of a real motion segment to validate position control. The spring properties were chosen so that static forces and moments do not exceed 600 N and 50 N m, respectively. This ensured that typical in vitro loads are accessible. The ends of the spring (30 mm each) were embedded in polymethylmethacrylate (PMMA, Technovit 3040, Heraeus Kulzer, Wehrheim, Germany). Flanges were screwed to the PMMA blocks so that the spring could be mounted in the dynamic disc-loading simulator. The middle of the spring was aligned with the origin of the machine’s coordinate system in an unloaded condition (Fig. 4).

Fig. 4
figure 4

Close-up photograph of the spring used as test specimen for position control analysis mounted in the dynamic disc-loading simulator

A sinusoidal waveform with an amplitude of ±17° (FE, LB) and ±8° (AR), respectively, was chosen for the target position. Translational amplitudes were ±2 mm (SAP, SLL) and ±3 mm (AC), respectively. All values were selected to stay well within the safe limits of the machine but still exceed typical physiological limits [44, 4649].

Each of the six DOF was tested individually while the remaining axes were held stationary. Displacement was applied as a sine waveform at frequencies of 0.5, 1.75, and 3.0 Hz, simulating low, medium, and fast dynamic testing speeds typically used for in vitro tests. This resulted in a maximum angular speed of 320°/s and a maximum translational speed of 36 mm/s. A total of 20 cycles were recorded of which the middle six median cycles were analysed for position error. Data were acquired at 100 Hz and analyzed using MATLAB (MATLAB 2013b, The MathWorks, Natick, MA, USA).

Ovine specimen study

Eight ovine lumbar motion segments were gathered from five healthy sheep aged between 3 and 5 years. The motion segments were immediately stored at −20 °C. Prior to testing, the specimens were thawed at 6 °C for 12 h. After dissection the vertebral bodies were cut in the middle of the transverse plane. To provide a clear view on the posterior anulus to observe a possible herniation during the tests, the posterior elements of the motion segments were removed. Care was taken to preserve anterior and posterior longitudinal ligaments. Screws were placed in the vertebral body of the motion segment to improve embedding in the PMMA (Technovit 3040, Heraeus Kulzer, Wehrheim, Germany). Flanges were screwed onto the PMMA to mount the specimens in the dynamic disc-loading simulator in an unloaded condition.

For preconditioning of the specimen, a compressive load of 130 N was applied for 15 min, simulating a typical spinal load of a standing sheep [50]. Testing was performed using the following protocol: 0°–12° flexion, 0°–9° right lateral bending, and 0°–4° right axial rotation combined with constant axial compression of 800 N typically occurring in vivo during dynamic activities [50]. The motions were applied cyclically with a frequency of 0.5 Hz. During the test, specimens were filmed from the posterior direction. Testing was stopped after 1200 cycles, which was deemed sufficient for the initiation and possible extrusion of nucleus material as found in a preliminary study with similar ovine specimens. Upon test completion, the discs were again frozen at −20 °C for a maximum of 4 weeks and thawed at 6 °C over night prior to the day of scanning. Imaging of the potential defects was performed in a dedicated 11.7 T small animal ultra-high field MRI system (UHF-MRI, BioSpec 117/16, Bruker Biospin, Ettlingen, Germany) using an experimental protocol (MR-method: FLASH, contrast: T1, Resolution: 75 µm isotropic, acquisition time: 9 h 25 min) [51]. All data were received with a 40 mm quadrature transmit/receive coil. Discs were scanned with no external load applied. Analysis of video documentation and scan images was done by visual inspection. Additionally, the scan images were used by specialized software (Avizo standard v5.6, FEI Visualization Sciences Group, Burlington, MA, USA) for segmentation and 3D-reconstruction of the disc and lesion volume.

Results

The new dynamic disc-loading simulator enables an arbitrary combination of all six DOF. Loads of up to ±10 kN in axial compression, ±2 kN in both shear directions and moment up to 100 Nm around all axes can be applied (currently limited by the load cell). It allows high speeds of up to 707°/s and 100 mm/s.

Position control analysis

The actual position during motion typically displayed a consistent sinusoidal waveform at all frequencies, except for higher frequencies of both shear directions (Fig. 5).

Fig. 5
figure 5

Target and actual sine curves of typical (axial rotation, 1.75 Hz, left) and of highest error (laterolateral shear, 3.0 Hz, right). Encircled areas mark the effect of backlash

Generally, the normalized error between target and actual position increased with rising testing frequency (Fig. 6). The highest median errors of 2.4 % were yielded by both shear directions (AP and LL) at 3.0 Hz (0.048 mm at 2 mm amplitude) where they also yielded the highest interquartile range (IQR) (SAP 4.4 %, SLL 6.1 %). The remaining four axes were almost similar with a median error of ≤2 % and an IQR of ≤2.4 % at all frequencies. The lowest error (median and IQR) was found for the FE-axis at 0.5 Hz (median 0.43 %/0.072°, IQR 0.40 %/0.068°).

Fig. 6
figure 6

Error between target and actual position with respect to the particular amplitude of the six axes (LB lateral bending, FE flexion/extension, AR axial rotation, SAP anteroposterior shear, SLL laterolateral shear, AC axial compression). The mean value is seen as a “+” sign

During testing, the FE, LB, and AR moments typically reached ±50, ±50, and ±30 Nm, respectively. Forces in SAP, as well as SLL direction were ±250 N, while AC was ±500 N.

Ovine specimen study

The first in vitro study in this dynamic testing machine showed that with the chosen complex motions failure of ovine discs can be produced in each specimen (Table 2). Three different kinds of lesions could be identified: four herniations, two protrusions, and two delaminations. Two modes of failure were observed when herniation or protrusion occurred: either anular failure (two cases) or endplate junction failure (four cases) (Table 2). Video documentation showed in three cases of herniation that the lesion developed within the first few cycles. One herniation was not visible because the sequestrum was covered by the posterior longitudinal ligament. Endplate lesions were generally noticed by a crackling noise while an anular lesion was barely audible. The sequestra were gradually pressed out and reached their maximum volume outside the anulus after a few cycles (one case) and after 500 cycles (two cases) (Fig. 7). In case of protrusion, a bulge on the left posterolateral side was noticed during testing. UHF-MRI scans revealed structural details of both failure patterns (Fig. 8). Failure was seen in the mid to left posterior part of the discs. UHF-MRI scans showed delaminations as areas of no signal inside the anulus.

Table 2 Overview of the ovine specimen study’s results
Fig. 7
figure 7

Gradual development of a herniation. After 500 cycles the herniation does not increase further

Fig. 8
figure 8

Reconstructed volume of a disc (red) and lesions (green) in craniocaudal view (left). Sagittal 11.7 T MRI section in the area of the lesion (right). L1/2 anular lesion (top) and L4/5 endplate lesion (bottom). For better visibility nucleus and anulus are highlighted in color

Discussion

This paper introduces a new custom-designed dynamic disc-loading simulator with six DOF. It is suitable for most kinds of biomechanical testing of FSU and allows high speeds, which previously could only be attained by commercial robots, e.g. Kuka KR 16 (Kuka Robotik GmbH, Gersthofen, Germany). For in vitro tests with angular speeds <53°/s and displacement speeds <9.4 mm/s, the positioning accuracy can be specified as ≤0.08° and ≤0.026 mm, respectively. It does not reach the described high accuracy of the hexapod system by [41] but matches or exceeds the aforementioned commercial robotic system.

Position control analysis

The position control analysis demonstrated a general increase of error between target and actual position with increasing cycle frequency. This was to be expected, as in studies by Ding et al. [41] and Holsgrove et al. [32], since the required torque of the electric motors increases to the square with test frequency due to inertia of the moving parts, thus disturbing the control system.

For all test frequencies, higher errors were seen in each of the shear directions than in the remaining directions. When compared to the third translational DOF, the gears have a higher backlash (0.20° vs. 0.08°), there is additional mechanical laxity from the joint heads of the lever arms, and the 2 mm amplitude is equal to a rotational angle of the drive shaft of only about 20° (vs. 140°).

Although AR has the least moment of inertia of all rotational axes and a gear with the lowest backlash (<0.08°), it did not yield the highest accuracy. This can be explained by the smaller amplitude of ±8° used for the axis, since errors were normalised to amplitude.

Ovine specimen study

Before using human specimens a first study was performed with ovine specimen applying a dynamic, complex loading protocol with four DOF. Four herniations and two protrusions, with either endplate junction failure or anular failure, were produced. The goal in this study was to show that it is possible to provoke disc failure under extreme conditions. Therefore, bending angles were chosen to be four times the values obtained by Wilke et al. [47] for the range of motion (ROM) of ovine lumbar FSUs at 7.5 Nm. Such failure could also be produced in vitro with an increased pressure by injecting gel into the nucleus [27]. Interestingly, the same ratio was found recently in a clinical study by Rajasekaran et al. [52] which showed that 117 of 181 patients (65 %) with herniations had an endplate junction failure.

The defects could nicely be visualized by the UHF-MRI. Unfortunately, it is not practical to scan all specimens also before testing because one scan with this high resolution requires about 10 h of scanning time and warms up the specimens which may have an effect on the failure mechanism. Therefore, only one of the specimens was scanned exemplarily prior to testing, which allowed comparison of the UHF-MRI images after failure with the status before. This specimen did not show any indications of defects. Previous histologic studies of sheep from the same source aged between 3 and 5 years also showed no marked pre-existing defects [53].

Even though only eight specimens were used, it seems that there is a tendency toward endplate junction failure (4 times) over anular failure (2 times). This implicates the bone, rather than the disc, as the structure primarily susceptible to damage.

Usually, bone mineral density (BMD) of ovine vertebral bodies is 6.5 times higher than human specimens [54, 55]. However, it is not known how the BMD affects the mode of failure. Since a similar trend was present in the current study, it may be concluded that ovine specimens, even having a higher BMD, is a useful failure model of the human disc.

A recent ovine in vitro study by Wade et al. [26], using an apparatus previously introduced by Adams et al. [56], is more comparable to the test conducted in the present study since flexion (10°) was achieved by compression (approximately 10 kN). Wade et al. [26] only found anular failures and endplate fractures, but no endplate junction failures. It may therefore be concluded that the biomechanical approach of load application has an important influence on the mode of failure. Therefore, further investigation in this area is imperative.

While mechanical fixtures integrated into uni- or biaxial testing machines are used for a specific load scenario and thereby limited in their design, the dynamic disc-loading simulator presented allows any type of load combination, thus rendering it highly versatile.

The apparatus is prepared for several enhancements. A water bath can easily be added when testing for several hours. The usage of an external motion capture system as was previously done by [57] is possible since the view of the specimen is unobstructed. Intradiscal pressure measurements can also be conducted and are automatically synchronized with load and displacement measurements by connecting the pressure sensors to the analog input of the Flextest-40-controller. Furthermore, a second load cell can be installed on the cranial side of the specimen if desired.

A limitation of the apparatus is the maximum angular speed about the FE axis; this motion direction does not reach the desired value of 707°/s but rather only 400°/s, owing to the fact that a balancing weight was added. Currently, load control is available for axial compression only. Implementation of further load controlled DOF is under investigation. Another limitation is that any displacement results in reaction forces, dependent upon the specimen’s stiffness typically seen in dynamic in vitro tests. A possible solution would be load- or hybrid control, however, load control for dynamic testing is technically challenging.

Conclusions

A new, custom-designed, electromechanical, 6-DOF dynamic disc-loading simulator was developed and validated. It is highly versatile and expandable if desired. This device provides new possibilities for fundamental spine biomechanics research. An initial study on ovine specimens showed that herniation can be provoked with complex loading. Experiments in this new dynamic disc-loading simulator will improve understanding of the initiation of intervertebral disc lesions, especially in human disc specimens. This device is now ready to perform many studies on discs with different degrees of degeneration from different regions of the spine to differentiate between different load scenarios or other parameters such as number of load cycles or velocity. The findings from this first experiment form the foundation for future experiments to find the critical load cases.

These findings may have a great impact on the credibility and confidence of medical court expert testimony. Another notable application includes implant failure testing (subsidence of cages, dislocation of nucleus implants or disc prosthesis, failure of anulus sealing methods, pedicle screw loosening, etc.) under complex physiological conditions or following ASTM or ISO standards.