A research team, led by Pompeu Fabra University (UPF) with the collaboration of Hospital del Mar, has created a unique database in the world containing 16,807 different three-dimensional virtual models of the spine, specifically of the thoracic and lumbar areas, using computational systems. It is an open science project and the database has been shared with the entire scientific community through the European research repository Zenodo. Thanks to the European project Disc4All, led by UPF, these models can also be viewed and downloaded via the Spineview application, co-developed by UPF and the company InSilicoTrials.

This database has been created as part of research aimed at improving the diagnosis and personalized treatment of spinal deformities, in line with the growing trend towards personalized medicine. This deformity, which is highly prevalent among the adult population, causes curvature of the spine in relation to the pelvis, giving rise to so-called sagittal imbalance. If the patient undergoes an operation, the risk of post-surgical complication of mechanical origin is by no means minor, and there is a great need to improve the prognosis and personalized planning.

The results of the research have been published recently in the journal Scientific Data (Nature). In addition to the BCN MedTech research unit at the Department of Engineering at UPF and Hospital del Mar, the study has involved researchers from two Italian institutions: the technology company InSilicoTrials in Trieste, and the IRCCS Istituto Ortopedico Galeazzi in Milan.

Towards more precise diagnoses and treatments for spinal deformity

This study comes in response to the need to create effective and scientifically validated models to analyse deformities of the thoracolumbar part of the spine, which were not available until now. “Using the database we have created, it will be possible to generate personalized computational models of the thoracolumbar part of a specific patient and obtain more accurate diagnoses and treatment plans, which leads to personalized therapies” – explains Jérôme Noailly, head of the Biomechanics and Mechanobiology (BMMB) research area of the UPF BCN MedTech Unit.

As a starting point, the research has taken as a reference a mechanical model of the standard structure of the spine, in accordance with previous studies, with the vertebrae, ligaments and articular contacts with the spine, as well as the intervertebral discs, but without considering individual morphological variations. Thanks to the numerical model of finite elements -a method widely used in different fields of engineering that discretizes the basic components of a structure into basic elements-, it is possible to calculate the impact that a movement or mechanical load (simulated virtually) might have on each of the tissues present in the model of the column and mapped along the discrete mesh. Using a simile, a similar method could be used for calculating the strength of the different elements of a bridge.

Based on this mechanical structure of the spine, “in this research, we have generated 16,807 personalized 3D models of the thoracolumbar part”– explains Morteza Rasouligandomani, a researcher at the UPF BCN MedTech Unit and first author of the article. First, two-dimensional images of the spine of 42 real patients were obtained using radiology techniques (EOS). Secondly, these 2D images enabled reproducing the three-dimensional structure of the thoracolumbar bone structure of the spine of these patients.

Next, the standard mesh of finite elements with the other tissues, used as a starting point, was adapted to the external 3D structure of the spine of each patient through numerical shape-transformation processes. This allowed obtaining complete 3D models, both of the inside and of the outside the thoracolumbar part of the spine of these 42 patients. Finally, using a statistical modelling system, variations of these prototypes were generated to create a virtual population of as many as 16,807 models of the thoracolumbar part of the spine based on real interindividual morphological differences. The biomechanical response of these models was also subsequently validated with experiments based on information and data from donor patients.

A second investigation by UPF creates 169 models of intervertebral discs to progress with the personalized diagnosis of chronic lower back pain

In the same European repository (Zenodo) where these 3D models of the spine can be found and through the same access and visualization platform (Spineview), the UPF BCN MedTech Research Unit has recently shared 169 three-dimensional models of the intervertebral discs of the thoracolumbar part of the spine. These discs are the largest avascular structure -where blood does not flow- of the human body, so they act like a sponge to absorb water and mechanical loads, and to transport the nutrients that come from the two vertebrae above and below to the resident cells.

This database is part of a second investigation by UPF, recently presented in an article in the journal Frontiers in Bioengineering and Biotechnology, which is also based on the methodology of finite elements. “In this research, we analyse how a certain physiological mechanical load affects the intervertebral discs according to the specific local morphologies of the different elements that form them and may vary among patients” – explains Estefano Muñoz-Moya, a researcher at the UPF BCN MedTech Unit and first author of the article. The results have great potential to contribute to personalized diagnoses and therapies for each patient and prevent the degeneration of the intervertebral discs and harmful effects on health, such as chronic lower back pain or herniated discs. These discs can degenerate slowly or rapidly and to different degrees for different reasons, for example due to habits such as a sedentary lifestyle or smoking, or due to genetic factors, but always in interaction with the movements or mechanical loads that a person frequently develops.

To generate the different models of intervertebral discs, imaging of the lumbar part of the spine of 36 patients has been performed, yielding 3D models of the external and internal shapes of 169 intervertebral discs (between 4 and 5 per person). Subsequently, a procedure similar to that of the previous research was followed: with numerical shape transformation processes, the standard finite element mesh (defined in previous research) of a mechanical model of the intervertebral disc -that represents all the tissues of the organ- was adapted to the morphology of the discs of each patient. Then, the impact that a physiological mechanical load (simulated virtually) would have on the mesh of the intervertebral discs of different models of real patients with their respective morphologies was calculated, with the aid of machine learning techniques.

This study has revealed that the impact of mechanical loads on the intervertebral discs does not depend only on the height of these structures, the most widespread view in medicine to date, but on their global or holistic composition. Thus, we find that the mechanical loads that are redistributed in the different tissues of the intervertebral disc result from the interactions between a multitude of morphological variables. Therefore, some discs could naturally be more susceptible to degeneration than others, due to the particularity of the mechanical environments that are generated in them caused by the weight of the body. Knowledge of these relationships between morphology and mechanical loads could improve the exploitation of medical imaging techniques such as magnetic resonance imaging to define personalized risk factors.

Reference articles:

Rasouligandomani, M., del Arco, A., Chemorion, F.K. et al. Dataset of Finite Element Models of Normal and Deformed Thoracolumbar Spine. Sci Data 11, 549 (2024). https://doi.org/10.1038/s41597-024-03351-8

Muñoz-Moya E, Rasouligandomani M, Ruiz Wills C, Chemorion FK, Piella G, Noailly J. Unveiling interactions between intervertebral disc morphologies and mechanical behavior through personalized finite element modeling. Front Bioeng Biotechnol. 2024 Jun 10;12:1384599. doi: 10.3389/fbioe.2024.1384599. PMID: 38915337; PMCID: PMC11194671.

Web with multimedia and interactive images of the models generated by the two previous investigations (thoracolumbar part of the spine and intervertebral discs):

https://spineview.upf.edu/