Combining biomechanical finite element analysis and multi-parametric MRI to assess mechanical and structural damage in cervical spondylotic myelopathy
Manuel Taso1,2,3,4, Pierre-Jean Arnoux2,4, Léo Fradet4,5, Arnaud Le Troter1,3, Jean-Philippe Ranjeva1,3,4, Kathia Chaumoître4,6, Pierre-Hugues Roche4,7, and Virginie Callot1,3,4

1CRMBM UMR 7339, Aix-Marseille Université, CNRS, Marseille, France, 2LBA UMR T 24, Aix-Marseille Université, IFSTTAR, Marseille, France, 3CEMEREM, AP-HM, Pôle d'imagerie médicale, Marseille, France, 4iLab-Spine international associate laboratory, Marseille/Montréal, France, 5Mechanical Engineering, Ecole Polytechnique de Montréal, Montréal, QC, Canada, 6Service de Radiologie, Hôpital Nord, AP-HM, Pôle d'imagerie médicale, Marseille, France, 7Service de Neurochirurgie, Hôpital Nord, AP-HM, Trauma Center, Marseille, France

Synopsis

While diagnosis of cervical spondylotic myelopathy is easily done with MRI, patient outcome is still difficult to predict. It is nonetheless associated to a strong mechanical cause as spinal cord’s (SC) compression is the first event leading to tissue alterations and neurological deficits. This work proposes an original approach using biomechanical numerical simulation, to apprehend the mechanisms of SC compression by the disk, and multi-parametric MRI, to probe the consequent microstructural alterations (axonal loss, demyelination …). Thanks to spatial normalization, first results on 3 patients are presented, allowing co-localization of personalized simulation of mechanical stress and structural MR alterations.

Target audience

Scientists and clinicians involved in spinal cord (SC) MRI and compressive SC diseases.

Introduction

Cervical spondylotic myelopathy (CSM) is a common SC pathology, whose diagnosis relies on MRI. It may lead to severe neurological impairment1. However, accurate quantitative prognosis biomarkers are still lacking. The relationship between the mechanical cause of the disease (SC compression) and the tissue alterations leading to clinical deficits are also unknown. In this context, biomechanical finite element analysis (FEA), through numerical simulation, is a promising tool to give some insights into the primary cause of SC tissue damage. In this work, a proof-of-concept of a unique framework, combining biomechanical FEA and multi-parametric MRI (mp-MRI), as well as automated MR post-processing, is proposed to provide new insights into the mechanical origin of the structural damage encountered in CSM.

Material and methods

Three patients (1M/2F, age 56±20yo) with cervical myelopathy were included. MRI was performed at 3T using standard coils, with approval of the Local Ethics Committee.

mp-MRI consisted in multi-echo GRE T2*-w sequence (0.5x0.5x5mm3, 5 echoes, 8 slices), monopolar single-shot SE-EPI DTI (30 directions, b=0 and 800s/mm2, 0.9x0.9x10mm3), and HASTE-ihMT sequence2,3 (500 saturation pulses, 500μs each every 1ms, Δf=±7kHz, alternation between single and dual frequency saturation, 0.9x0.9x10mm3). All images were acquired with ECG-synchronization, axially, with one slice covering the site of maximal compression (2 cases at C5/C6, 1 at C4/C5). Mp-MRI metrics were estimated offline (ihMTR, FA, ADC, $$$\lambda_{//}$$$ and $$$\lambda_{\bot}$$$).

FEA: a finite element model of the cervical spine (SM2S – Spine Model for Safety and Surgery)4,5 (fig.1) with biofidel geometry drawn from in vivo CT for bones and MRI data (AMU atlases6) for the SC, and including mechanical properties (viscoelastic for soft tissues, elastoplastic for bone) derived from in vitro data4,5,7 was used. For each patient, an estimated compression ratio at the most affected level was calculated based on the T2*-w MRI and using an invariant database8. To replicate the SC compression observed with MRI (30, 31 and 36% of the antero-posterior cord diameter, for patient #1, 2 and 3, resp.) a quasi-static imposed displacement was applied to the intervertebral disk (IVD) of the concerned functional unit (example for C5/C6 on fig.1). The computed stresses (Von Mises stress, scalar derived from the stress tensor) were used to investigate potential damage occurrence.

MRI/FEA combination: for each case, the outputs of the FEA were projected from a 3D-mesh into a 2D-plane (to match the MR image acquired at maximal compression site), then scaled and resampled to match the MR resolution. After mp-MRI normalization9, the deformed geometry from the FEA was realigned to the subject’s T2*-w slice using affine registration (FLIRT,FSL) and applied to the VM-stress map. Co-localized VM stress and mp-MRI metrics were then quantified in different WM/GM regions of interests9,10 (fig.2b-c).

Results and discussion

Mp-MRI and mechanical stresses for the three cases are presented on fig.3. Although quantitative comparison has not been performed yet, the simulation led to morphological deformations that were qualitatively similar to those seen on MR images (e.g. when comparing GM morphology from the MRI (fig.3b) and stress maps (fig.3c)). Then, for all patients, it can be seen that higher stresses (red to light-blue pixels, fig. 3c) were encountered in anterior GM and WM regions (motoneurons and motor pathways) compared to the posterior regions (p<0.0001 HSD Tukey-Kramer) and maximal (in average 676±43kPa) in the GM anterior horns. On the MR side, widespread alterations can be seen (10 to 15% in average as compared to reference control values11), especially on the FA and ihMTR metrics (fig. 3d,e, vs. fig. f,g), with similarities between site of maximal MR alterations and maximal VM stresses (yellow arrows).

While biomechanical analysis provides some information that could be linked to the primary lesion occurring, mp-MRI provides sensible markers of tissue alterations related to primary and secondary lesions. Moreover, the numerical simulation only gives a static image of the tissular stress encountered, not taking into account yet the temporal degenerative aspect of the pathology.

Therefore, the mechanical and structural information appear complementary for a better understanding of the pathophysiology of CSM. Analysis of other mechanical parameters (directional stress, pressure) could also provide pertinent information about specific tissue alterations. Finally, mechanical and MR analysis distal to the compression site could also help in understanding and assessing the extent of secondary lesions (e.g Wallerian degeneration12).

Conclusion

A proof-of-concept of a unique framework combining biomechanical FEA and mp-MRI is proposed. This methodology, applied on a wider cohort, and associated with longitudinal MR follow-up and multivariate analyses, could potentially help determining prognostic markers of surgical outcome, reference treatment for CSM.

Acknowledgements

A*MIDEX ANR-11-IDEX-0001-02

References

1 Northover et al, Skeletal Radiol, 2012. 41(12): p. 1543-6; 2 Varma et al, MRM, 2015. 73(2): p. 614-22; 3 Girard et al, MRM, 2015. 73(6): p. 2111-21; 4 El-Rich et al, J Biomech, 2009. 42(9): p. 1252-62; 5 Wagnac et al, J Biomech Eng, 2011. 133(10): p. 101007; 6 Taso et al, MAGMA, 2014. 27(3): p. 257-67; 7 Ichihara et al, J Neurotroma, 2001. 18(3): p. 361-7; 8 Fradet et al, Spine, 2014. 39(4): p. E262-9; 9 Taso et al, Neuroimage, 2015. 117: p. 20-28; 10 Lévy et al, Neuroimage, 2015. 119(0): p. 262-271; 11 Taso et al, ISMRM 2015, p.681; 12 Cohen-Adad et al, Neuroimage, 2011. 55(3): p. 1024-33.

Figures

Fig. 1 SM2S finite element model of the cervical spine with detailed meshing of the C5/C6 functional unit and illustration of simulation result (geometry and Von Mises stress)

Fig. 2 Example of T2*-w MRI (a), automated GM/WM segmentation (b) and WM tracts delineation (c) of a patient with a C5/C6 compression

Fig. 3 mp-MRI for the 3 patients and corresponding simulated stress maps, along with normal ihMTR and FA maps11. The red arrows highlight the SC compression on the T2 and T2*-w scan. The yellow arrows highlight the area of maximal Von Mises stress and tissue alterations observed on mp-MRI metrics



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0847