Five questions … on the ‘bionic spine’ for Professor David Grayden
BY GARRY BARKER
Scientists at the University of Melbourne, Royal Melbourne Hospital and Florey Institute for Neuroscience and Mental Health have found the holy grail of bionic research.
They have developed technology to closely monitor, record and interpret human thoughts and use them to control a bionic limb, or even a vehicle, without opening the skull or implanting sensors directly on the brain. It has been dubbed “the bionic spine”.
It is a world first in biomedical engineering that uses electrodes on a tiny stent, called a stentrode, inserted through the jugular vein to lie beside the brain.
The electrodes pick up the tiny microvolt electrical signals generated in brain neurons. Trials so far have been on sheep, whose thoughts are unknown, but when clinical trials begin with humans next year, their thoughts will be known, detected, recorded, interpreted and sent to a computer via a radio transmitter implanted under the skin of the chest.
Algorithms running in the computer will interpret those signals and send instructions to a bionic limb or vehicle to help disabled people with spinal cord damage, even motor neurone disease.
But the Melbourne researchers see great potential for other fields, including amelioration of Parkinson’s disease, prediction and control of epileptic seizures and treatment of severe depression.
Professor David Grayden (BE(ElecEng) (Hons) 1990, BSc 1991, PhD 1999) is a team leader in the group of 40 medical and bioengineering specialists working on the project.
Professor Grayden and his team are charged with identifying, recording and interpreting brain signals collected by the stentrode and with building the algorithms to instruct an external device.
He is Professor and Deputy Head (Academic) of the Department of Electrical and Electronic Engineering.
1. Collection and interpretation of brain signals has been possible for years. What makes this development so important?
Until now, detecting and recording detailed signals from neurons required electrodes inserted directly into the brain. But most degraded and failed after six months – partly due to loss of insulation but more because the brain built up scar tissue or glia (non-neural tissue occurring naturally in the brain) around the electrodes, blocking contact with neurons.
The stentrode has no contact with the brain. Electrodes are firmly placed in the blood vessel, the wall of which grows over the stentrode, securing it in place, so it can continue in use in everyday life.
2. Trials so far have been on sheep, but, as you say, it is difficult to get useful information. You can’t tell a sheep what to think. But clinical trials with humans, disabled and not, are to begin in 2017. What are the goals then?
Before we can begin clinical trials with humans we must prove the stentrode is safe and convince surgeons that our implants will not cause harm. Nitinol, the nickel-titanium alloy used for the stent, has been proven biocompatible. The stent is a standard item, first developed to remove blood clots in the brain that have caused strokes. In our case, electrodes are fixed to the stent by Dr Nick Opie (PhD 2012) and his team at Royal Melbourne Hospital.
The vein we use is about three millimetres in diameter, into which we insert a tube. The stent and its connecting wires are then pushed through the tube, which has an internal diameter of less than two millimetres. The tube is then withdrawn, leaving the stentrode and its wires in place. In our trials with sheep the wires connect to a recording unit on the sheep.
For very long-term use in humans we plan to have the wires emerge from the blood vessel and connect to a device that has a battery, a processor and a radio transmitter. When the first humans receive this device next year, we will launch straight into decoding the signals produced by their thoughts.
3. Recording signals from the brain is now fairly common in modern medicine. What makes your research different?
It is a by-product of the way neurons behave in activating and communicating with each other that they create electrical fields. It was discovered many years ago that you could record those fields and make sense of them. That’s how EEG (electroencephalography) developed. But we want to record those minute electrical fields over very small areas of the brain and infer from them what the neurons are doing.
So, if you are sending an instruction to “lift your leg” there will be a sequence of activations of those neurons that send commands down to the spinal cord and then to the legs saying “activate that muscle or this muscle”.
4. So it’s finding out what the voltage changes mean and then creating appropriate instructions to a bionic device?
Yes. We can use different techniques such as EEG, MEG (magnetoencephalogram), which measures magnetic fields in the brain, and MRI (magnetic resonance imaging) to look at brain activation as people imagine actions.
With the stentrode we can look at a signal picked up by a single electrode, see the characteristics of it and, if certain frequencies become activated, then that means something is happening and we can decode that.
Another way is to look at what the relative activity is between multiple electrodes. There may be a sequence of instructions or we can target where the signal is coming from, just as multiple microphones on a submarine can detect the position of a sound source.
We would say to someone, “imagine raising your arm”, they imagine it, we record the signal and then we say “lower your arm” and we get hopefully a different signal. We can analyse that directly or, a common approach now, feed it into a machine-learning algorithm in a computer to decipher it. From that we can set up templates of what different actions look like.
Machine learning is very helpful but it needs to adapt over time because people adapt the way they do their thinking. There is an interaction – you have machine learning to decode the patient’s activity but the patient is also learning how best to work with the system.
Such learning happening from both angles is more likely to make the system work.
5. And the goals of your team? And what about other applications?
Success depends largely on our ability to communicate with the machine, the computer that converts the signals from the brain neurons into instructions that replicate and transmit the intent in the thoughts of the patient. You can put electrodes into the brain, but if you can’t make sense of the signals being recorded, then there’s not much point.
The idea of using a stent came from Dr Tom Oxley (a neurologist from the University of Melbourne interested in vascular systems and electrophysiology). Tom was looking for support for his idea.
Professor Terry O’Brien (MB BS 1988, MD 2000), head of medicine at the Royal Melbourne Hospital, suggested he speak to Professor Tony Burkitt and me. We were at the Centre for Neural Engineering at the time. We thought the idea was fantastic and agreed to support it.
We immediately saw what the challenges were and were familiar enough with the bionic eye and ear to know the sorts of challenges that would be met, such as finding biocompatible materials and working in very small spaces.
Tom had been speaking to DARPA (the US Defense Advanced Research Projects Agency) and we joined in with a formal application for funding, which was successful.
It might be possible to record from other parts of the brain, perhaps use it for epilepsy monitoring. I am also interested in the prospect of stimulating the brain through the stentrode’s electrodes. Could we deliver electric current to the brain to stop a seizure, treat severe intractable depression or stimulate lower brain centres for control of Parkinson’s disease?
An implanted device to treat Parkinson’s disease is already available but its use involves invasive surgery with all the inherent risks of entering the brain.
Electrical stimulation delivered there can alleviate some of the symptoms of Parkinson’s. Severe shaking and “freezing”, when the patient can’t move, are treatable.
Read more on Pursuit.