Not a member yet? Register for full benefits!

Fundamental Obstacles to Developing Reliable Neuroprosthetics

Much of the basics of neuroprosthetics - devices that are implanted semi-permanently or permanently into the body and interface directly with the electrical or chemical neural signals within the brain or central nervous system - are here, existing in one form or another. Practical brain implants have been being tested for nearly a decade now, and scarcely a month does not go by without one significant stride forwards in neuroprosthetics making the press.

Why then, are they not mainstream yet?

Well, not to put too fine a point on it, neuroprosthetics work, yes. But, and this is a big but; they do not yet work reliably. There are several fundamental obstacles which must be overcome in order for neuroprosthetics, and as direct extension, brain repair, BMI and augmented brain function, to be a reality.

  • Stable interfaces between electromechanics and neural tissue

    One of the major problems with taking any external electrode array, and embedding it deep within the brain, is you have to cut your way down and embed the array in an open wound. There is currently no way around that.

    So, whilst initially you gain a strong signal, and many if not most of the electrodes are hooked into the electrical output of nerves, over time micro scar tissue forms, and cells begin to die back, away from the wound site. The electrical signals become fainter, as they have to fight through more and more dead matter, until finally, they are undetectable. When that occurs, the neuroprosthetic must be resited in the brain, a complex and dangerous operation.

    Alternatively, several inroads have been made on electromechanical devices that extend themselves, pushing individual electrodes through the scar tissue, into fresh brain matter. Of course, to do this, they traumatise the brain a little more, and fresh scar tissue soon forms around the new location, necessitating an even deeper push. The distances involved are tiny, but over a matter of years, such an approach would be like a tumor, slowly digging through the brain. Hardly acceptable.

    MEMS elecromechanical, neuron seeking digging electrode
    Credit: Caltech

    Other approaches on the outside of the brain, or outside of the skull, have better luck maintaining a stable interface, reading brainwave patterns rather than individual neurons themselves. Of course, what they gain in stability, they sacrifice in speed and accuracy, unable to determine specific signals, just general concepts.

    Still other attempts are being made using nanotube 'artificial neurons' instead of electrodes, and some using light to pierce the brain and read deep inside. It is too early to tell how much success either approach will have.

    What is clear is until we have a stable method for reading or writing to the brain's neural circuitry long-term, without damaging it, mainstream neuroprosthetics will not occur.

  • Decoding of nerve signals/ nerve signal networks in order to understand how to stimulate for intended effect

  • This issue and the one below, do kind of go together. Both rely on the deep seated necessity of decoding and reverse engineering the brain, understanding the codes that drive it in the same way that coded data drives a more conventional computer system.

    There is a great deal of work undergoing in this area. Whilst fMRI does not directly examine the neurons individually, it examines oxygen flow in the blood in the brain, identifying areas of general activity. There are at least a dozen such studies run every month, and the number is increasing. Large, well financed efforts such as the blue brain project are attempting to recreate the brain in silico. Smaller projects are growing neurons in petri dishes, and wiring them up to robotic systems or simulators, to study how they function.

    In 2008, MIT successfully replicated half a mouse brain, for the same reason. Fruit fly networks, because they are so close to human brains, are continually poked and prodded. A great deal of time and money is being sunk into research to understand how to decode the brain, follow what it does, how it does it, and what every code signal precisely means.

    Due to the sheer scale of the problem, it will be many years before we understand it all, let alone have the computing power to realistically analyse it. Yet, unless there is a civilisation destroying event, it is a foregone conclusion that we will eventually fully understand the brain.

    Long before that has been achieved, we will be regularly using the information we do have, to write signals back into the brain, informing it of specific tasks. That is happening even now. Haptic prosthetic arms, which feed back touch sensation information back into the peripheral nervous system, do so thanks to years of research on decoding the electrical signals passed up and down the nerves of the arm, when a finger or thumb encounters resistance. By duplicating those signals precisely, and feeding them into the stumpy remnant of arm nerves from a severed arm, the feeling of touch has been replicated precisely.

    Grasping and feeling the lettuce with sensors in a haptic, prosthetic arm

    As computing power advances, and we decode more of the signals, it becomes possible to interface back with ever greater fidelity, increasing the amount of data prosthetics can send. Deep brain stimulators, at the moment just blanket the brain with current, halting an epileptic attack in its tracks for example. However, when they do that, they also blanket and cancel out other signals, randomly. As we gain in understanding about which signals where, do what, we will eb able to precisely pattern and target such discharges, affecting only the brain systems, that have gone wrong.

  • Decoding of nerve signals/ nerve signal networks in order to understand precisely what is being read

    This issue and the one above, do kind of go together. Both rely on the deep seated necessity of decoding and reverse engineering the brain, understanding the codes that drive it in the same way that coded data drives a more conventional computer system. Much of that which applies above, also applies to reading from the brain. However, there is one big difference.

  • Whilst writing back can frequently be accomplished from outside the brain, tapping into the peripheral nervous system directly, to send sensory data back to the brain, reading nerve firing must be accomplished from within the brain itself. Thus, we are left with the quandary of understanding and decoding the brain, to follow what signal goes where to do what, then having to reach inside, potentially disrupting those signals, to be able to read signals even deeper inside the brain, in real-time.

    How this might be achieved, without causing disruption on all the signals a reader would have to pass through, is still unclear.

    Neurons used to control a Robot
    With thanks to New Scientist for filming

References & Further Reading

Using Nanotech as Neuroprosthetics

Fruit Flies: Cells with double vision

BMI: Basic Introductions

Brain Reading: Diffusion Spectrum Imaging

Brain Reading: fMRI

Brain Scans Read Intentions

BrainGate: Video Introductions

CBS 60 Minutes: Harnessing the Power of the Brain

Neurons as Artificial Control Circuits

Neuroprosthetics, Brain Emulation and Mind Uploading: The ultimate VR concepts

Podcast: EPOC Brain Machine Controller

Brain Machine Interface Enabled Wheelchair

Brain blankets for BMI

Interrupting the Brainstem

VR Interfaces: Mind Flex

Blue Brain Project - creating a simulated brain

Woman with bionic arm regains sense of touch

Staff Comments


Untitled Document .