AI can predict bone fractures in cancer patients
Image credit: Dreamstime
A team of US researchers has used artificial intelligence (AI) to create a digital twin of a patient’s vertebra that will be able to predict the risk of bone fracture in people who have suffered from cancer.
As medicine continues to embrace machine learning, a new study suggests that scientists may be able to use AI tools to predict how cancer affects the probability of spinal fractures.
Every year, over 1.6 million cancer cases are diagnosed in the US, 10 per cent of which experience spinal metastasis, which occurs when the disease spreads from other places in the body to the spine. One of the biggest clinical concerns patients face is the risk of spinal fractures due to these tumours, which can lead to severe pain and spinal instability.
“Spinal fracture increases the risk of patient death by about 15 per cent,” said Soheil Soghrati, associate professor at Ohio State University. “By predicting the outcome of these fractures, our research offers medical experts the opportunity to design better treatment strategies, and help patients make better-informed decisions.”
While many of the changes the body undergoes when exposed to cancerous lesions are still a mystery, computational modelling can be the key to a better understanding of its effects on the spine, said Soghrati.
His team was able to train an AI-assisted framework called ReconGAN to create a virtual reconstruction of a patient’s vertebra, that can be used to help doctors analyse the effects of tumours on patients' bones. The findings of the study were published in the International Journal for Numerical Methods in Biomedical Engineering.
Unlike 3D-printing, where a virtual model is turned into a physical object, the concept of a digital twin involves building a computer simulation of its real-life counterpart without creating it physically. Such a simulation can be used to predict a system’s future performance or, in this case, how much stress the vertebra can take before cracking under pressure.
By training ReconGAN on MRI and micro-CT images obtained by taking slice-by-slice pictures of vertebrae acquired from a cadaver, researchers were able to generate realistic microstructural models of the spine. Using their simulation, Soghrati’s team was also able to virtually enlarge the model, a vital capability for the understanding of changes in a vertebra’s geometric shape.
“What really makes the work in a distinct way is how detailed we were able to model the geometry of the vertebra,” said Soghrati. “We can virtually evolve the same bone from one stage to another.”
The researchers used CT/MRI scans from a 51-year-old female patient whose lung cancer had metastasised to simulate what might happen if cancer weakened some of the vertebrae and how much stress the bones could take before fracturing.
The model predicted how much strength parts of the vertebra would lose as a result of the tumours, as well as other changes that could be expected. Some of their predictions were confirmed by later clinical observations.
For a field like orthopaedics, using a non-invasive tool like a digital twin can help surgeons understand new therapies, simulate different surgical scenarios and envision how each patient's bone will change over time.
“The ultimate goal is to develop a digital twin of everything a surgeon may operate on,” he said. “Right now, they’re only used for very, very challenging surgeries, but we want to help run those simulations and tune those parameters even more.”
Much more work is needed to make this vision a reality, Soghrati said. ReconGAN was trained on data from only one cadaveric sample, and more data is needed for the AI to be perfected.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.