It is no exaggeration to say we currently stand on the cusp of a new wave of innovations that will totally transform how humans live. At present, we are very much finding our feet, and imagining the potential applications for technologies such as artificial intelligence (AI), and machine learning. What will these technologies ultimately lead to? No one knows precisely, but we are starting to see some of the early signs of what it might mean in the field of medicine. One such avenue of investigation is the idea of using AI to help people with traumatic brain injury (TBI).
TBI - A Challenge to Diagnose and Treat
TBIs can be notoriously difficult to diagnose, especially in their mildest form, because the brain damage may not be immediately obvious. According to Marianne Sturr, director of the spinal cord unit and the brain injury unit at the Bacharach Institute for Rehabilitation in New Jersey, USA, "the brain is the most difficult to identify a problem, because if you don't do a test for that specific area, you'll miss it,". And treatment is highly dependent on a range of factors outside of the doctor’s control, including the brain’s own ability to heal and rewire itself, a process known as neural plasticity. It is therefore clear that better tools to detect injury and assist the brain to recover are much needed.
Drawing the Connection Between AI and Brain Injury
At first, it may not be obvious how AI could play a factor in the recovery of a TBI, but by pulling together several separate mechanisms and uses, it is possible to envision a future whereby this new technology plays a vital role in patient recovery from the point of initial diagnosis, through the immediate medical intervention, to rehabilitation and ultimate recovery. We will look at two potential applications in more depth; 1) using AI to assist in the diagnosis of concussion, and 2) how AI can read the brain, and how this could be used to enable patients with severe loss of function to communicate.
Looking for signs of concussion, even years after the event
In a fascinating and revolutionary study published in the European Journal of Neuroscience, scientists aimed to prove that concussions can, in later life, cause “persistent anomalies of the brain structure and function that interact with the effects of normal ageing”. To do this, they used machine learning to analyse a series of detailed images and other ‘markers’ which together created a ‘multi-dimensional’ signature of a brain affected by concussion in the earlier life of an individual. This created a series of ‘automated classifiers’ of the condition, which when applied to brain images, showed a 90% accuracy of detecting concussion-related damage. This may, in the future, be used to create a tool to assist medical staff to quickly detect the existence of past concussion in patients. By enabling clinicians to reliably spot signs of previous concussion, patients will be more assured of receiving the care they need, even if it is years after the event that caused the initial damage.
In a similar line of research, Brightlamp, a start-up company based in the USA, is using AI and machine learning to detect more immediate signs of concussion. The technology simply requires a mobile phone to flash a light into the eye of a patient – which could be used by paramedics or for sports people who have suffered a serious blow to the head. It will measure the pupil dilation of the individual and send the information to a cloud-based neural network, which will then look for the characteristic markers of a person with concussion. The technology is said to have a 98% accuracy rate.
Mind-reading – science fiction comes closer to reality
An article in the Cerebral Cortex medical journal published in October 2017 described a study in which neuro-scientists used a neural network to work out how brains encode visual images. Researchers then used an algorithm based on this to decode the brain. The results were remarkably good given the infancy of this technology. In half of all attempts, the software was able to correctly pick which video the participant was watching (from a choice of 15).
And in very recent work, scientists in Japan were able to get closer to the holy grail of reading images in the brains of individuals. By using a Deep Neural Network (DNN), they were able to improve on past attempts to achieve this goal by using a hierarchal approach to processing which is similar to that used by the brain. Information gleaned from different parts of the brain was then filtered through the DNN, to make sense of the noise. This was then passed through another layer which attempted to further refine the picture by comparing it to a series of natural images. A neutral observer was then able to select which of two photos the person had been looking at in 99% of cases. The software was not effective at depicting imagined photos, but when used for simple shapes the recreated image was recognisable in 83% of cases.
This technology still has years of development ahead, but it seems almost inevitable that one day, we will be able to see into the mind of those who have suffered a serious TBI and are unable to communicate. This will open up an extraordinary set of opportunities to give these people a voice so they can reconnect with their loved ones and the world at large.
Support for brain injury victims
Even when compared to ten or even five years ago, AI, deep learning, and machine learning has moved so quickly that there seems little doubt that we will one day see technology that will transform the diagnosis, treatment, and communication of those with mild to severe TBI’s. Rather like the beckoning age of self-driving cars, the future is almost tangible. Until then, we are reliant on the specialist skills and care of clinicians and rehabilitation therapists to do their very best to put back what has been lost following a catastrophic brain injury.
At IBB, our personal injury team, led by brain injury legal expert, Malclom Underhill, has the expertise and knowledge to advise and represent you if you wish to seek compensation for a traumatic brain injury. To talk about how we might be able to help, please phone us on 0333 123 9099, email us at firstname.lastname@example.org or fill in our contact form. Any discussions you have with us will be in the strictest of confidence.