Based at Imperial's White City Campus, we are a research group with a focus on Spine Biomechanics. We use a range of tools to better understanding in the areas of spinal injury, spinal deformity and spinal surgery.
Our lab has state-of-the-art ex vivo testing capabilities, including bespoke testing rigs, a 6 DOF robot arm, a C-arm, pressure needles, water baths, and high-speed X-ray. We also have access to advanced imaging technologies, including micro-CT, 9.4T MRI, and microscopy.
We use novel computational approaches (finite element modelling, msk modelling, digital volume correlation (DVC), machine learning) to develop workflows to provide clinicians with information to inform patient treatment strategies, to better predict risk of injury, and to assess scoliosis brace designs.
We collaborate globally, with ongoing projects with colleagues in New Zealand, USA, Portugal, South Africa, Germany, Australia, Sri Lanka and India.
You can explore our recent publications below.
Results
- Showing results for:
- Reset all filters
Search results
-
Journal articleRaftery KA, Levy H, Adamson L, et al., 2026,
Three-dimensional analysis of interbody cage-apophyseal ring contact to predict endplate subsidence following transforaminal interbody fusion
, Clinical Biomechanics, Vol: 134, ISSN: 0268-0033BackgroundThere is a higher risk of subsidence following transforaminal lumbar interbody fusion (TLIF) relative to other approaches. Decreased subsidence risk is associated with anterior cage placement, speculated to be because of increased apophyseal ring contact. However, this hypothesis is largely based on in vitro evidence and, to date, has not been investigated in a clinical cohort.MethodsPre-operative and post-operative computed tomography (CT) images from 42 TLIF patients were used to segment the endplate and cages. The apophyseal ring boundary was manually landmarked for each endplate. The pre-operative endplates were rigidly registered with the post-operative cage position, and an iterative closest point approach was used to calculate the contact area between the cage, apophyseal ring, and endplate. Subsidence was categorised based on severity (No Subsidence: <2 mm; Moderate Subsidence: 2-4 mm; Severe Subsidence: ≥4 mm) from post-operative CT.FindingsApophyseal ring contact was significantly lower in both Moderate and Severe Subsidence, relative to No Subsidence (Moderate: −19.6 ± 7.0%; Severe: −21.5 ± 6.5%; P < 0.05), and negatively correlated with subsidence depth (P < 0.05). Injury risk analysis demonstrated that a 50% subsidence risk was associated with 45.7% (38.4–53.6%) apophyseal ring contact. Suprajacent endplate apophyseal ring contact, but not subjacent, was significantly predictive of subsidence at the respective endplate (P < 0.05).InterpretationThe risk of subsidence in TLIF patients can be mitigated by ensuring that at least half of the interbody cage surface area is in contact with the peripheral endplate rim, particularly at the endplate superior to the cage.
-
Journal articleRaftery KA, Tavana S, Davis B, et al., 2026,
In vivo prediction of intervertebral disc strains and segmental kinematics from clinical MRI during lumbar extension
, Frontiers in Bioengineering and Biotechnology, ISSN: 2296-4185Introduction: Excessive intervertebral disc (IVD) strains and vertebral body motions are associated with lower back pain (LBP). Quantifying these strains and motions may aid in predicting the success of candidate LBP treatments and enable better prediction of pre-operative instability and post-operative implant failure, but cannot currently be obtained in routine clinical assessment. Thus, the aim of this study was to evaluate the feasibility of utilising clinical measures of spinal alignment, IVD geometry, and disc degeneration to predict in vivo IVD strains and vertebral translations. Methods: Fifteen participants presenting no LBP were subjected to one unloaded and one supine extension-loaded MRI scan. MRI-based digital volume correlation (DVC) was used to quantify the principal and shear strains of lumbar IVDs and anterior-posterior, cranial-caudal, and total translation of the vertebral bodies (L1-S1). IVD height, anterior-posterior IVD height ratio, segmental lordosis, lumbar lordosis, lumbar height, sacral angle, and Pfirrmann grade were evaluated using the reference MR images. Multivariate linear regression was used to predict level-wise strains and translations.Results: IVD strains and vertebral translations were successfully predicted from clinical measures of spinal alignment and disc degeneration, but only at the L4-L5 and L5-S1 levels. Specifically, greater minimum principal IVD strains and vertebral anterolisthesis were associated with a reduced anterior-posterior IVD height ratio at L4-L5 (p < 0.01). Greater peak minimum principal strains and anterolisthesis were associated with taller IVDs in the L5-S1 segment (p < 0.05). In the same segment, increased sacral angle was associated with greater peak minimum principal strains (p < 0.05) but lower anterolisthesis (p < 0.01). Discussion: This study demonstrates the potential of utilising radiographic variables to predict the biomechanical behaviour at the segmental level, giving rise to future
-
Journal articleRaftery KA, Levy H, Singh R, et al., 2026,
Intervertebral disc distraction stiffness predicts endplate subsidence following transforaminal interbody cage expansion: an ex vivo study
, European Spine Journal, ISSN: 0940-6719Purpose: Expandable cages have the potential to mitigate the currently high subsidence rates following transforaminal lumbar interbody fusion (TLIF), but are liable to over-distraction in situ. This may be due to the undefined patient-specific expansion threshold of the intervertebral disc (IVD) space. This study aimed to elucidate whether IVD properties affect the torque required to expand the cage within the IVD space, and determine the association between achieved torque, distraction stiffness, and subsidence severity.Methods: Fifteen cadaveric L3-L4 and L4-L5 samples were prepared with the TLIF approach. Under 100N compression, the torque required to expand the cage per half-turn, alongside the changes to IVD and cage height, were recorded until maximum cage expansion. Subsidence depth was measured after subsequent cyclic loading, and the surface area of removed IVD tissue was quantified post-test.Results: Peak torque was inversely associated with preloaded IVD height (B: -0.34, p < 0.001) and the percentage of IVD removed (B: -0.04, p < 0.01). IVD distraction stiffness was associated with preloaded IVD height only (B: -0.19, p < 0.001). There was no association with IVD or facet degeneration. When subsidence depth was normalised to bone mineral density, a positive correlation was observed with peak torque and cage expansion stiffness (both p < 0.05).Conclusion: The torque required to expand interbody cages in situ is relevant to subsidence risk, and depends on IVD geometry and the amount of residual tissue. Thus, short IVDs should be thoroughly prepared to alleviate excessive stiffness during cage expansion.
-
Journal articleBrans VA, Constantinou AP, Kibble MJ, et al., 2025,
Ultrasound-triggered gelation for restoring biomechanical properties of degenerated functional spinal units
, Advanced healthcare materials, ISSN: 2192-2640Lower back pain is closely associated with intervertebral disc (IVD) degeneration and is a leading cause of global disability. Existing treatment options are unable to provide suitable long-term outcomes, and emerging strategies employing injectable biomaterials are hindered by factors including limited native tissue integration and depth- or time-constrained gelation mechanisms. To overcome these issues, the present research evaluates a new concept employing ultrasound to remotely trigger in situ implant formation. The concept centers around an implant precursor biomaterial consisting of an anionic polysaccharide solution containing thermally sensitive liposomes loaded with ionic crosslinkers. Ultrasound-mediated heating to 4–5 °C above normal body temperature triggers liposomal release of the crosslinking species, thereby initiating hydrogel formation. Optimization studies define the implant precursor material (1.5% wt/v sodium alginate seeded with calcium-loaded liposomes (10–15 mm calcium chloride) and 6% wt/v glass microspheres) and the ultrasound parameters (0.95 MHz, 1.6 MPa amplitude, 87% duty cycle). Proof-of-concept experiments in degenerated ex vivo bovine IVDs indicate partial restoration of biomechanical function, with the implanted biomaterial well-integrated into the disc tissue and without material herniation. These results offer promise for treating intervertebral disc degeneration, with continued refinement of biomaterials and protocols being essential for achieving robust in-disc efficacy.
-
Journal articleSlater TD, Raftery KA, van Heeswijk VM, et al., 2025,
A multiscale ex vivo method to investigate intervertebral disc strain and fiber recruitment in anterolateral bending using 9.4T MRI-DVC and DIC microscopy
, JOR Spine, Vol: 8, ISSN: 2572-1143Study DesignEx vivo, multiscale analysis of disc strain using ultrahigh-field MRI-based digital volume correlation (MRI-DVC) and differential interference contrast (DIC) microscopy.ObjectiveTo evaluate the relationship between three-dimensional strain distributions and collagen fiber recruitment in porcine cervical intervertebral discs under flexion and lateral bending.Summary of Background DataFlexion combined with lateral bending is often linked to disc herniation, yet the strain patterns and fiber-level changes in the annulus fibrosus are not well understood. Multiscale characterization is essential to uncovering failure mechanisms.MethodsFour porcine cervical motion segments were scanned in neutral and anterolaterally (AL)-bent postures using 9.4T MRI, with 3D strains calculated via DVC. Samples were sectioned and imaged with DIC microscopy to quantify collagen fiber recruitment based on fiber crimp patterns, using a crimp grading scale (0 = fully straight, 1 = semi-crimped, 2 = uncrimped).ResultsMRI-DVC revealed an inhomogeneous strain distribution in AL-bent discs, with higher magnitudes compared to the neutral discs. Fiber uncrimping was greater in the AL-bent discs (mean crimp grade: 0.44, mostly straight) compared with the neutral discs (1.56, predominantly crimped). Across the bending axis, the anterior-right region exhibited higher strains than the posterior-left (minimum principal strain ~25% greater), which correlated with the presence of sequential lamellae having straight and fully-crimped fibers. A greater amount of fiber uncrimping was observed in the posterior-left than anterior-right disc regions.ConclusionThis study confirms the suitability of MRI-DVC combined with DIC microscopy for relating macroscopic strains to microscopic fiber crimp, and for identifying regions of high strain across multiple length scales. Under AL-bending, this methodology revealed that the disc's posterior region exhibited taut fi
-
Journal articleSlater T, Wilke H-J, Gurusamy G, et al., 2025,
Lumbar disc herniation modelling: a review of ex-vivo mechanical models and a comparison with clinical data
, European Spine Journal, Vol: 34, Pages: 4353-4368, ISSN: 0940-6719PurposeEx-vivo herniation models are essential for studying lumbar disc herniation mechanisms, but their morphological accuracy remains unclear due to limited validation against patient-derived clinical data. This review collates clinical lumbar disc herniation characteristics and evaluates whether existing models replicate real-world pathology. By identifying the most morphologically relevant models, this study provides a stronger foundation for improving mechanistic herniation models.MethodsA systematic review following PRISMA guidelines identified clinical studies detailing herniation characteristics and experimental models of ex-vivo lumbar disc failure. Models were categorised by loading conditions (complex ultimate compression; cyclic; and intradiscal pressurisation), then compared to clinical data to assess their validity.ResultsIn patients, extrusions (50%) and protrusions (34%) are the most common lumbar disc herniation types, with paracentral herniations (61%) predominantly occurring at L4-L5 (49%) and L5-S1 (42%). Structural failure patterns varied, with annulus fibrosus failure reported in 35–81% of cases and endplate junction failure in 19–68%.Among 25 analysed models, all loading types induced herniations, but often with different damage patterns. Complex ultimate compression caused abrupt failures and fractures, while cyclic led to progressive annular damage. Intradiscal pressurisation highlighted nucleus pulposus migration pathways. Within a single herniation model, the damage mechanisms seen were similar between discs.ConclusionsClinical herniation patterns show significant variability, while ex-vivo models yield more repeatable outcomes. Cyclic, complex ultimate compression, and intradiscal pressurisation models provide valuable mechanistic insights but differ in physiological relevance. Researchers must consider the physiological relevance of the applied load and the differences between animal and human discs when selecting a model. Fu
-
Journal articleSlater TD, Gagliostri B, Kibble MJ, et al., 2025,
A comparison of five animal models for acute intervertebral disc herniation research
, JOR Spine, Vol: 8, ISSN: 2572-1143Study DesignMicrostructural investigation of mechanical load induced acute disc herniation on five animal models.ObjectiveTo compare how spinal discs in different animal models herniate under a standardized complex compressive load.Summary of Background DataAnimal models in disc herniation research offer reduced degeneration-associated variability, lower cost, and greater availability compared to human specimens. However, there is limited consensus regarding which species is best suited for modeling human herniation, making a comprehensive comparison of species-specific herniation mechanisms necessary.Materials and MethodsA standardized shear and compressive load, designed to herniate intervertebral discs, was applied to isolated discs of five cadaveric animal models (n = 30, 6 specimens per group): bovine tail, bovine lumbar, ovine lumbar, porcine lumbar, and porcine cervical. The segments were flexed (7°), and a shear-compressive load was applied at a crosshead displacement rate of 40 mm min−1, until a force drop, or a displacement limit was reached (~80% of disc height). Microstructural analysis was undertaken to identify failure modes.ResultsClinically relevant herniation features were observed in all models—including endplate and annulus fibrosus (AF) tearing, AF delamination, vertebral body (VB) fracture, nucleus pulposus (NP) extrusion into VB, and radial NP movement. Bovine lumbar, porcine cervical, and porcine lumbar segments exhibited high rates of radial NP movement (84%, 100%, and 67%, respectively), with ovine lumbar discs displaying VB fracture (84%) and NP extrusions into the VB (67%). Bovine tail discs showed minimal damage but were characterized by sequential lamellar AF tears (67%).ConclusionsPorcine cervical, bovine lumbar, and porcine lumbar discs are suitable for annulus-failure herniation research, although porcine cervical discs may be the most appropriate due to exhibiting the highest rate of releva
-
Journal articleWang AP, Slater T, Raftery KA, et al., 2025,
Endplate preparation for anterior cervical discectomy and fusion: does the amount of endplate removed affect cage subsidence risk?
, European Spine Journal, ISSN: 0940-6719Purpose: Subsidence after anterior cervical discectomy and fusion (ACDF) is a common complication that may be influenced by the degree of endplate removal prior to cage insertion. The optimal degree of endplate removal remains unclear; therefore, we performed a series of ex vivo experiments to elucidate the relationship between the aggressiveness of endplate preparation and subsidence risk.Methods: Human cadaveric subaxial cervical endplates were partially decorticated either conservatively (n = 10) or aggressively (n = 9). The degree of endplate removal was quantified using microCT. Subsidence was modelled by measuring the strength and stiffness of each specimen when an interbody cage was axially compressed into the endplate.Results: Conservative endplate preparation resulted in less endplate removal than aggressive endplate preparation (mass: 150 vs 301mg, p < 0.001; volume: 47 vs 88mm3, p = 0.01; thickness: 0.02 vs 0.16mm, p = 0.004). There was no significant difference between the two groups with respect to endplate strength (2.04 vs 2.04kN, p = 0.99) or stiffness (2.38 vs 2.41kN/mm, p = 0.89). Bone mineral density (BMD) was similar between the two groups (271.6 vs 271.9mg/cm3, p = 0.98) but positively correlated with endplate strength (effect size 0.68, p = 0.001).Conclusions: When performing partial cervical endplate decortication, the degree of bony endplate removal did not significantly predict endplate integrity during ex vivo compression testing, but greater BMD was associated with increased strength. The degree of endplate removal should be based on individual patient factors and intraoperative findings to achieve the ideal cage-endplate interface.
-
Journal articleRaftery K, Kargarzadeh A, Tavana S, et al., 2024,
Disc degeneration influences the strain magnitude and stress distribution within the adjacent trabecular bone
, Frontiers in Bioengineering and Biotechnology, Vol: 12, ISSN: 2296-4185Introduction: Up to one in five will suffer from osteoporotic vertebral fracture within their lifetime. Accurate fracture prediction poses challenges using bone mineral density (BMD) measures. Trabecular bone strains may be influenced by the underlying intervertebral disc (IVD). Understanding how disc degeneration alters load distribution to the vertebra may demonstrate that supplementing fracture risk tools with IVD metrics could improve predictions. The aim of this study was to assess the influence of IVD degeneration on the stress and strain magnitude and distribution in the trabecular bone of adjacent vertebrae.Methods: Ten human cadaveric lumbar bi-segment specimens (20 IVDs, 9 degenerated, 11 nondegenerated) were µCT-imaged under 1000N. Digital volume correlation was used to quantify axial, principal, maximum shear, and von Mises strain in the superior and inferior regions of the vertebra. Volumetric BMD from quantitative-CT was used to calculate Young's modulus, which was then registered with the von Mises strain field to calculate internal von Mises stress.Results: Two bi-segments fractured during mechanical testing, resulting in N = 8 endplate regions per group. Trabecular bone adjacent to degenerated IVDs presented higher maximum principal and shear strains in the anterior region, relative to non-degenerated (peak ε1: 6020 ± 1633 µε versus 3737 ± 1548 µε, p < 0.01; peak γmax: 6202 ± 1948 µε versus 3938 ± 2086 µε, p < 0.01). Von Mises stress distribution was significantly skewed towards the anterior region in the degenerated group only (28.3 ± 10.4 %, p < 0.05). Reduced disc height correlated with increased central-region axial compressive strain, decreased central-region BMD, and increased anterior region von Mises stress (all p < 0.05).Discussion: Disc degeneration may encourage high strains to be experienced within the anterior regio
-
Journal articleTamanna R, Matthew K, Luca H, et al., 2024,
Comparison of four in vitro test methods to assess nucleus pulposus replacement device expulsion risk
, JOR Spine, Vol: 7, ISSN: 2572-1143BackgroundNucleus replacement devices (NRDs) are not routinely used in clinic, predominantly due to the risk of device expulsion. Rigorous in vitro testing may enable failure mechanisms to be identified prior to clinical trials; however, current testing standards do not specify a particular expulsion test. Multiple methods have therefore been developed, complicating comparisons between NRD designs. Thus, this study assessed the effectiveness of four previously reported expulsion testing protocols; hula-hoop (Protocol 1), adapted hula-hoop (Protocol 2), eccentric cycling (Protocol 3), and ramp to failure (Protocol 4), applied to two NRDs, one preformed and one in situ curing.MethodsNucleus material was removed from 40 bovine tail intervertebral disks. A NRD was inserted posteriorly into each cavity and the disks were subjected to one of four expulsion protocols.ResultsNRD response was dependent on both the NRD design and the loading protocol. Protocol 1 resulted in higher migration and earlier failure rates compared to Protocol 2 in both NRDs. The preformed NRD was more likely to migrate when protocols incorporated rotation. The NRDs had equal migration (60%) and expulsion (60%) rates when using unilateral bending and ramp testing. Combining the results of multiple tests revealed complimentary information regarding the NRD response.ConclusionsAdapted hula-hoop (Protocol 2) and ramp to failure (Protocol 4), combined with fluoroscopic analysis, revealed complimentary insights regarding migration and failure risk. Therefore, when adopting the surgical approach and animal model used in this study, it is recommended that NRD performance be assessed using both a cyclic and ramp loading protocol.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.