Cutting-edge techniques to address spinal degeneration
While clinical observations provide insights into people’s health, converting such evidence into well-informed disease prevention and treatment remains challenging. This is the case in spine degeneration and low back pain, where tissue and organ degradation that lead to clinical symptoms are often highly multifactorial. “The complex interplay of factors is beyond our natural capacity of analysis,” explains Disc4All(opens in new window) project coordinator Jérôme Noailly from Pompeu Fabra University(opens in new window) in Barcelona, Spain. “However, computer models and simulations can help us to retain only the most important cause to consequence relationships, providing us with sufficient understanding for well-informed predictions and actions.”
Understanding processes triggering spine degeneration
Noailly’s aim in this project was to provide a predictive mathematical tool to help spine care professionals make more informed decisions. However, he was finding it hard to make his models that depict what is going wrong mechanically in the spine correlate with what spine care professionals are trying to address and patients are actually feeling, which is pain. The Disc4All project, supported by the Marie Skłodowska-Curie Actions(opens in new window) programme, enabled Noailly to bring mathematical models closer to desired clinical support through the combination of artificial intelligence (AI) and machine learning. A core consortium of 12 beneficiaries brought together expertise in computer and data science, experimental and computational biology, bioinformatics, biomechanics and medicine. This was achieved through integrating evidence-based data with biological and mechanical computational models. The tools developed during Disc4All were applied to a range of data sources, which include both human cohorts and laboratory experiments. The aim was to provide a clearer picture of what is actually causing a patient intervertebral disc degeneration in the lower back, a major cause of lower back pain. “We started looking more into the biological processes that actually trigger the degeneration of the spine,” he adds. “By gathering data on this, I thought we might be able to identify personalised and enhanced descriptors of the causes of pain.”
Grouping molecules according to function
AI and machine learning enabled Noailly and his colleagues to process this data rapidly, identifying objective risk factors of spine degradation along with demographic and psychological data. “Medical image biomarkers were converted into personalised parameters in mathematical biomechanical models,” Noailly explains. “This enabled us to compute and explore different biophysical mechanisms and the relation of each of these mechanisms with the personalised biomarkers.” Model simulations led to the identification of potential biomarkers, i.e. certain molecules that might be directly related to pain but not screened clinically. Noailly was able to see patterns in biological processes that are not easily captured in traditional ways. “It’s a little like how astronomers describe faraway phenomena without actually seeing them,” he remarks. “They use complex models, and this is what we are doing here.”
Meaning from biological and patient data
By applying machine learning to extract meaning from large amounts of biological and patient data, the project hopes to eventually help surgeons achieve a clearer picture of the causes of lower back pain in individual patients. “AI-enhanced computer modelling is an efficient way of building up a personalised spinal model for precision medicine,” says Noailly. “We are able to interpret data quickly and provide predictions as to whether personalised features represent a biological risk for intervertebral disc degeneration. This can then be leveraged to prevent or handle low back pain.” The aim is to eventually bring this pioneering computational modelling concept into clinical settings. For healthcare providers, this could mean less reliance on expensive and time-consuming technologies such as CT and MRI scans.