Our Work in Neurotechnology: You Asked; We Answered

The Battelle NeuroLife® Neural Bypass Technology skips damaged areas of the nervous system to allow the brain to communicate directly with muscles. The technology and the strides we’ve made in recent years often bring up questions – not just from those in the neuroscience, but from regular folks around the world. 

We recently hosted an I Am A … Ask Me Anything session on Reddit and received more than 250 questions. 

The Participants

Here’s who helped answer the questions:

Ian Burkhart, In 2010, Ian broke his neck while diving into the ocean. Four years later, at The Ohio State University, he underwent an operation to implant a chip in his brain that detects part of the neural activity in his motor cortex. That activity is translated by algorithms in a computer system invented at Battelle. The “thoughts” are transferred from a computer to a special sleeve on Ian's arm that stimulates his muscles to respond to his brain's commands, moving his hand in a purposeful way.

Dr. Marcie Bockbrader, Assistant Profession & Residency Research Director, Residency Research Course Director, The Ohio State University Wexner Medical Center.

David Friedenberg, Battelle Statistician and Algorithm and Data Analysis lead for NeuroLife.

The Questions

Check out some of the highlights from the Ask Me Anything session. View the full thread here. 

So I'm guessing this device bypasses the spinal injury to directly send signals to his arm? If so what kind of materials did you need to use/ invent yourself to help prevent the body from rejecting the brain implant. And if you have to constantly take immunosuppressant drugs to reduce the chance of a rejection.

MAB: The brain-computer interface uses a brain “chip” including the Blackrock Microsystems Utah microelectrode array. For the implantable hardware, see: https://blackrockmicro.com/neuroscience-research-projects/human-research-systems/.

There is always neuroinflammation after chip implantation, but eventually a scar forms around the electrodes. The scar can be a problem for recording the brain, but we’ve been able to work around it and get good recordings for about five years. No immunosuppressant drugs are needed to prevent rejection. The most important ways to minimize inflammation are to: (1) prevent/treat infections immediately, whether they are from the urinary tract, skin or other body areas; (2) take probiotics (several lines of evidence suggest that gut health affects the central nervous system; (3) manage stress (again, there’s a link between stress and whole body inflammation; (4) optimize nutrition (there is evidence that low sugar, high omega 3 diets are anti-inflammatory); and maybe even (5) take melatonin every night at bedtime.

As the physician-researcher on the team (MAB), one of my jobs is to keep Ian healthy and catch infections early.

The bypass all happens outside Ian’s body, going through a computer workstation, decoder software, and a wearable muscle stimulator that is calibrated to activate forearm muscles that generate the movement that Ian thinks about doing. Because none of these parts are implanted, we can reanimate Ian’s paralyzed arm without worrying about rejection of implants in his arm or hand.

In what form does the technology receive brain signals? For example, is it like a direct current coming from different directions, at different intensities? Or is it a continuous/layered stream in some form?

MAB: The technology acquires brain signals through an implanted, 96-channel electrode array. Each channel detects voltage that is continuously generated from active neurons in its neighborhood. The sampling rate is 30kHz; this means that the number of voltage measurements that we collect is 30,000 per second and this streams continuously in real-time while Ian uses the system. We have to use signal processing methods to mathematically translate the “raw” voltage data into normalized response intensity that we can use as neural features for our machine learning decoder algorithms. For our initial proof-of-concept report, with signal processing details see: https://www.nature.com/articles/nature17435  

Does the chip become "part" of the brain? For instance, does the chip carry on signals from one neuron to the other that its in between, if at all? Or is it always an end/start point and never an intermediary?

MAB: The chip becomes scarred into the brain, but not part of the brain. And the eventual intent is to remove it when it is no longer functional (or something better comes along). It acts as a microphone to listen to local neural activity. It doesn’t carry signals from one neuron to another, although others are working on new technology to do that (particularly for memory enhancement).

What measures are taken to assure this technology cannot be accessed/hacked locally or over WiFi/Bluetooth or whatever it is you use to communicate back to your computer?

MAB: Right now, the tech can’t be hacked because it is all wired: the brain chip is connected by cable to a computer, the computer is offline (not on an Ethernet or WiFi network), and the decoded instructions for movement are transferred by USB to a muscle stimulator that is in direct contact with Ian’s arm. We are working towards remote control of devices (car) for mobility, and that will require a secure, non-hackable network. If brain activity were hacked, what the hacker would see is a large stream of numbers, ranging from -1000 to 1000 microvolts (though numbers can be larger with environmental noise). The real concern would be if the connection to Ian’s arm were networked, and thus, hackable – we don’t want a hacker taking control of Ian’s hands! This is why we are currently using a wired system.

What is the delay between his brain and the movement of his arm? Is it significantly greater than without the chip?

DF: In the original system, the lag was usually around one second. Recently we've been exploring using deep learning algorithms and have found that they can speed up the response time by about 200 milliseconds. The response time is a really important metric that we are constantly trying to improve. You can check out some of our recent work on the topic here and here

In the future, will the visual stimulus that helps guide Ian to perform a particular task or movement be less necessary?

DF: Yes! We typically use visual cues when we are calibrating our algorithms. For instance, we'll ask Ian to think about moving his hand in the same way a virtual hand on a computer monitor. This gives us a dataset where we know exactly when he is thinking of different hand movements which is precisely what we need to train our machine learning algorithms. Once the algorithm is trained, Ian can use the system without the visual cues. For instance, he'll play Battleship and move when it's his turn.
We're thinking about how we can eliminate the algorithm calibration step (and the associated visual cues) all together. Typically, we calibrate the algorithms every session, but for this technology to be practical for all-day, everyday use we need to minimize or ideally eliminate the need to calibrate the algorithms every day.

What's been the most exciting and/or surprising moment for you all over the last few years? Have there been any outcomes that you never really expected?

DF: Certainly, the first time Ian was able to move using the system was a big thrill for all of us. We had been preparing for that moment for a long time and weren't sure it would work until it actually did.

MAB: For me, the most exciting and surprising moment occurred when Ian began to be able to use the system while doing other things – this means that he now treats the neural interface as “part of him”. For example: We’ve done cognitive testing (using Digit Span) and found that Ian can do challenging cognitive tasks while multitasking to use the BCI system – just like a healthy person can scratch his nose or drink from a cup while answering a question. And Ian can use the system for fun activities too. He played Battleship and beat our grad student while using the BCI to move the game pieces.

IB: One of the most surprising things is how much we have been able to accomplish with so little. Personally, I’m excited that the system works and the hope it provides for the future to myself and others with similar issues.

However, stay tuned for more exciting findings that are currently in press detailing some positive unintended consequences of the 4+ years of using this bypass system.

How long did it take to be able to train your mind to use the interface, or was it instantaneous?

IB: It took me about 4 months (3 days a week for 3.5hrs each) to become comfortable with the response of the system. Before my spinal cord injury I never thought about “how” I was moving my arm so I needed to learn exactly what I needed to think about. This improved once I was given feedback from my hand moving with the help of muscle stimulation.

Why did you actually need so much training - are the electrodes not quite in the right place in the brain, not sensitive enough, too sparse?

IB: I was using the system right away but in order to get a comfortable response that I could rely on it took sometime. The electrodes are placed in an optimal position in the motor cortex however our method of analyzing neural data involves looking at groups of neurons instead of individual spikes.

Why start with the arm? The hand is one of the most complex muscle groups. Why not a leg? Do you see the need to have multiple chips/surgeries for different body parts?

MAB: We started with the arm because returning hand function to someone who has lost it is an important way to restore independence. Several studies among those with 4-limb paralysis (also called tetraparesis) report that return of hand function is a very high priority of people with cervical spinal cord injury. There are reasonable alternatives (e.g., wheelchair) to restore mobility for those who want to get around their house or in the community, even though walking will always be preferred to the wheelchair. There are a few assistive devices (e.g., U-cuff, text-to-speech interface, mouth stick) that can help a person with paralysis interact with their environment, but there is no substitute for manual dexterity when it comes to making a sandwich or pouring milk for lunch. We wanted to give Ian and others back the freedom that comes with being able to use his hands again.

Having said that, other groups are working on leg control and ambulation. For some people, an epidural spinal cord stimulator (see Harkema's work) may help restore lower limb function. So there may be other, better, solutions for lower limb paralysis. Time will tell.

You could certainly do multiple surgeries and implant multiple chips for different body parts, with a reasonable expectation that they would work well independently or as an assembly. However, from my perspective as a physician and a researcher, I always have to ask whether the risk is worth the benefit. Any implant can get infected, any surgery can have complications. For these reasons, I would definitely not recommend this type of neurosurgery and implant to an otherwise able-bodied, healthy person.

Has the system has been proven with other similar patients? Could you extend this to other parts of the body? 

MAB: The neuroimplant from Blackrock Micro has been used in BCI systems for a few dozen humans. Yes, you can extend the general concept to other parts of the body and other types of prosthetics. Some of the systems were designed to control communication devices for people with conditions like ALS, others have been used to control robotic limbs (see Jen Collinger and Robert Gaunt's work from the University of Pittsburgh) for people with stroke, spinal cord injury, or motor neuron disease. There are also other peripheral nerve devices under development that help restore sensation from and control to artificial limbs for amputees. Do a google search for Greg Clark from the University of Utah and Kevin Walgamott (the prosthetic user) for more info - they are testing the "Luke" arm, a neuroprosthetic to replace Kevin's dominant, left arm, which he lost in an accident. The technology is amazing!

How much money and time have you spent in the development of this system?

MAB: It's taken at least $15 million USD to get to this point, although, to be truthful, we've lost count. All of this money has come through donations to The Ohio State University and from Battelle's internal research and development budget. We're looking for additional funding options to carry the project forward into the future. As for time, it's important to note that many people working on neuroimplant technology have come before us. Ian was just the first human to reanimate his own paralyzed limb using this version of the BCI technology. Not counting the years of other people's efforts that our work depends on, it took Battelle and Ohio State about 6 years to get where we are ... and we're still working to improve the whole BCI system.