Neuroscientist Ben Seymour’s approach to understanding how the brain processes pain has its roots in engineering. He aims to develop neuroengineering technologies to treat chronic pain.
By using pain as a ‘teacher’, I aim to decipher the ‘software code’ for pain in the brain...I build a model of pain as a control system.Dr Ben Seymour
I aim to design intelligent targeted technologies to re-wire the brain to reduce pain. I’ve been using engineering theories such as control theory and signal processing to explore the concept of pain as a ‘teaching signal’ that exists to protect us from harm. My goal is to understand our sense of pain: what is pain and why do we experience it?
Pain offers a unique window into core processes of perception and action in the brain. For example, if you accidentally touch a hot stove, pain drives our reflex to quickly withdraw our hand to limit the chance of sustaining a burn, but it is, in fact, learning to avoid the hot stove that saves us from multiple future burns.
My toolkit for studying the brain includes data science, sensor networks and robotics. The use of data science (machine learning, AI, deep learning, network theory), although already widespread in research, looks at patterns of information related to pain in different parts of the brain.
By using pain as a ‘teacher’, I aim to decipher the ‘software code’ for pain in the brain. Using a mix of brain-imaging experiments and machine learning, artificial intelligence (AI) and simulations, I build a model of pain as a control system. I then use these models to try and discover what goes wrong with the pain system in patients experiencing chronic pain – a leading cause of disability across developed and less developed countries – and ultimately design target treatments to reduce chronic pain in clinical situations.
I undertook my clinical training in hospitals in Manchester, London and Cambridge. I completed my specialist training as a consultant neurologist in 2012 and was awarded a Wellcome Trust Intermediate Clinical Fellowship based at the Department’s Computational and Biological Learning Laboratory.
My Fellowship has involved periods of research in Japan. I’ve looked at, for example, the potential for using AI techniques to treat phantom limb pain in amputees and to unconsciously remove a fear memory from the brain using a combination of AI and brain scanning technology. I’m also the Principal Investigator at the Center for Information and Neural Networks (CiNet) in Osaka and an honorary Consultant Neurologist at Addenbrookes Hospital in Cambridge.
Engineering provides a formal, mathematical way of looking at the brain. It is by viewing the brain through an engineering lens that we can start to make sense of the variability in the way we feel pain in different situations. My research uses a domain of control theory called ‘reinforcement learning’ to help provide a framework for understanding pain behaviour and its underlying neural activity. Namely, how the brain controls our actions in order to minimise future harm. You can read more about this theory in a future paper of mine due to be published in Neuron
I use robotics to provide simulation frameworks to test theories and to model disease. The idea here is that there is a 'reality gap' in our current models of information processing in the brain between the theory and practical reality. Since these models are getting quite complicated, if we really want to test their validity and make predictions, we need advanced simulation platforms and even hardware 'models'. In the case of pain, therefore, we need to show how and why having a pain system helps to protect us against harm, and how different control architectures for doing this lead to slightly different behaviours. A paper I co-authored provides an overview. It is titled Decision-making in brains and robots – the case for an interdisciplinary approach and published in Current Opinion in Behavioural Sciences.
New technologies to target pain are within reach. This includes clinical applications in diagnostics and therapeutics. I led an international collaboration that introduced a back pain biomarker based on machine learning and deep learning of brain imaging data. This has allowed us to characterise differences in brain activity patterns in patients with chronic back pain versus healthy people. In turn, this has now led to collaborations with industry, to try and find new ways to translate animal models of pain to humans – something that is a notoriously difficult problem. I’m also working on a new AI-neurofeedback method that we’re developing between the clinical school and engineering as a new treatment for back pain, thanks to a recent University-administered Wellcome Developing Concept Award.