Silicon replacement parts for the brain


Proposed hippocampal prosthesis. The biomimetic model is a microchip that transforms hippocampal input into output signals that mimic what the damaged hippocampus would have produced.

Brain chips. Long a staple of science fiction, these wafers of silicon circuitry that interface with the nervous system are quickly becoming a reality as a result of pioneering new work in neuroengineering

You’re familiar with prosthetic hands, legs, etc. used to replace missing originals. Recent research brings us closer to creating “neural prostheses” that can replace or bypass damaged or malfunctioning brain tissue with microchips designed to replicate the lost function.

Neural prostheses can be categorized as sensory, motor or cognitive. Sensory neural prostheses feed information from the outside world directly into the nervous system. Cochlear implants (sometimes called “bionic ears”) are the first, best known and most widely used sensory prosthesis.

Normally, specialized neurons in the inner ear called “hair cells” convert sound into electrical/chemical signals that are then sent to the brain by the auditory nerve. An infection can destroy the hair cells but leave the nerve largely intact. In these cases a cochlear implant can convert sound into a pattern of electrical currents used to directly stimulate the nerve. More than 200,000 cochlear implants are already in use. An analogous retinal implant used to restore minimal vision in those with certain forms of blindness was recently approved by the FDA.

Motor neural prostheses record and decode the action potentials of neurons that control movement. The recorded signals can then be used by paralyzed individuals to control computers or other devices.

Restoring Cognition

The most ambitious neural prostheses, though, are of the cognitive variety. These seek to replace the function of damaged brain tissue and restore normal mental function. Those with certain memory impairments may be among the first to benefit from such a device.

A structure deep in the brain called the hippocampus is crucial for forming certain types of memory. Perhaps the most famous neurological patient in history, patient HM, underwent a surgical procedure in 1953 that largely destroyed his hippocampus in an attempt to alleviate epileptic seizures. The seizures were markedly reduced, but HM was left unable to create conscious long-term memories of anything he experienced after the procedure.

Now Theodore Berger and colleagues at University of Southern California and Wake Forest University have demonstrated a functioning neural prosthesis that can replicate the function of parts of the hippocampus and restore memory in rats.

Experiences and thoughts, including memories, are represented in the brain as patterns of all-or-nothing voltage changes across billions of neurons in the brain. Berger and his team have spent years recording the patterns of electrical activity produced by neurons within the hippocampus as well as by neurons that send signals into and out of the hippocampus. From this they’ve developed computational models that can accurately predict the hippocampus’s output from a given pattern of input, and are using these models to develop a “memory prosthesis” that can circumvent a damaged hippocampus.

Hippocampal Prosthesis 2

Figure showing (a) a cross section of the hippocampus and its normal circuitry, (b) conceptual representation of replacing CA3 with a microchip (c) hippocampal slice in which the CA3 field has been removed and dentate neural activity is fed into a microchip. The chip performs the same transformations as biological CA3 neurons and transmits the output through an electrode array to CA1 neurons, completing the circuit.

In an earlier test of these models (illustrated in the figure on the right), the team extracted thin slices of rat hippocampus and kept the slices alive temporarily in a dish. Tiny electrodes placed on the surface of the slice recorded the activity of neurons that feed into a subregion of hippocampus called CA3. Other electrodes recorded the activity of neurons (in a subregion called CA1) that carry the output from CA3. A microchip that implemented a computational model of CA3 function was fed the recorded input and then electrically stimulated the CA1 neurons based on the received pattern of input. The team cut the neuronal connections feeding into CA3, effectively removing it from the circuit. Nonetheless, the microchip was able to replicate the function of CA3 and generate the correct output.

More recently, and more dramatically, Berger and his colleagues used a similar technique to restore short term memory in intact rats.

The rats were trained to perform a “delayed nonmatch-to-sample” task known to rely on the hippocampus. They were placed in a box with two extendable levers. On each trial, one lever was extended into the box and the rat pressed it once. Then, after a delay of between one and thirty seconds both levers were extended and the rat had to press the lever it had not pressed earlier in order to receive a reward. The task required the rat to remember for the duration of the delay which lever it first pressed.

Once the rats achieved proficiency in this task they were implanted with 32 tiny wire electrodes for recording and stimulation, eight each in CA3 and CA1 on each side of the brain. The researchers found that their model could predict the output of the hippocampus (CA1 activity) based on the recorded pattern of activity in CA3. Furthermore, the pattern of CA1 activity predicted the animals’ memory performance on a given trial, with stronger activity for trials in which the animals remembered correctly.

The crucial test came when CA1 function was blocked by a drug injection. The drug caused the rats’ performance to drop to 50% correct (no better than random guessing), but when a microchip that implemented the team’s model of CA1 function was used to stimulate CA1 with a pattern of activity similar to that recorded on correct trials, the rats’ performance bounced back to almost the same level as without the drug.

While this isn’t quite as dramatic as removing or destroying the entire hippocampus and then reinstating its function with a brain chip, it’s a clear step in that direction.

The researchers achieved similar results in monkeys, this time recording patterns of activity from different layers of the prefrontal cortex (the front-most part of the wrinkly outer portion of the brain). They first showed that they could use a computational model to predict the neural activity in deeper layers from that recorded in shallower layers. They again showed that the pattern of recorded activity predicted whether the animal made an accurate response on each trial. Then they injected a drug known to disrupt activity in prefrontal cortex and impair short term memory performance: cocaine. Finally, as in the rats, they stimulated the deep layers at the appropriate times with the same pattern that was recorded on successful trials. Not only did the drugged monkeys’ performance improve with stimulation, it actually surpassed their performance without the drug!

Enhancing Cognition

As that last result suggests, these prostheses also offer the possibility of improving on healthy brain function. In fact, the neural prostheses in the studies just described enhanced performance not just in animals with compromised brains, but also in those that had not received any injection.

Perhaps this shouldn’t be so surprising. After all, the researchers are essentially reverse engineering a small chunk of the brain. Once they’ve decoded the patterns associated with successful performance on a given task, then it becomes possible to consistently impose those patterns on the brain with artificial stimulation.

Even more promising was the observation that when stimulation was applied during 20-40% of trials over the course of several sessions, the animals’ performance kept improving, and not just on trials in which they received stimulation. Even when the animals were not being stimulated by the microchip, both memory performance and the recorded pattern of neural activity in the hippocampus were enhanced - as if the artificial stimulation was helping to train the brain networks responsible for key aspects of task performance.

Stimulant drugs such as methyphenidate (Ritalin) and amphetamine (Adderall) have been shown to improve attention in healthy individuals as well those with ADHD. Drugs are being developed to sharpen memory in those with Alzheimer’s disease and other forms of dementia, but these efforts are likely to lead to drugs that boost memory in healthy individuals as well. Transcranial direct current stimulation and other forms of neurostimulation have been shown to boost an impressive array of mental functions. Soon we will need to add neural prostheses to the growing list of ways that healthy brain function can be enhanced.

The possibility of widespread use of cognitive enhancers by healthy people raises a host of ethical concerns. We as a society will soon have to decide whether their use is appropriate and under what circumstances.

Reconsidering Meat Chauvinism

The development of cognitive prostheses also raises profound questions about the nature of the self. If my hippocampus was replaced with an artificial one, would I still be me? When someone loses a leg or an arm, we tend not to think this changes who they are. But the brain is different. It is the seat of the mind – the machinery behind all we think and do. Wouldn’t changing it necessarily change us? What if the replacement part functioned exactly as the original, such that there was no noticeable difference in thought or behavior?

To push the point further, if parts of the hippocampus or layers of the prefrontal cortex can be replaced by a microchip, what’s to keep us from ultimately replacing the whole brain with computer hardware? If we can replicate the function of some brain circuits in silicon, then with sufficient understanding of the brain and the processing power to model it, couldn’t we replicate them all? Would the resulting electronic brain be conscious in the same way we are? Would it have to be afforded all the same rights as biological humans?

It’s easy to dismiss the idea out of hand. Our experiences with machines make it difficult to imagine them possessing the spark of human consciousness. But this kind of “meat chauvinism”, as it’s sometimes called, overlooks recent advances in neuroengineering like the ones I’ve just described and undermines the central pillar of the brain sciences – what Francis Crick, co-discoverer of the molecular structure of DNA, called “the astonishing hypothesis” – that the mind and consciousness arise entirely from the physical mechanisms of the brain.

We can accurately simulate all manner of physical systems with computers. We create computational models for everything from bridge designs to global weather patterns. There’s little to prevent us from successfully modeling the physical systems that produce the mind as well.

What do you think? Post your comments below.


Berger TW, Hampson RE, Song D, Goonawardena A, Marmarelis VZ, & Deadwyler SA (2011). A cortical neural prosthesis for restoring and enhancing memory. Journal of neural engineering, 8 (4) PMID: 21677369

Hampson RE, Song D, Chan RH, Sweatt AJ, Riley MR, Gerhardt GA, Shin DC, Marmarelis VZ, Berger TW, & Deadwyler SA (2012). A nonlinear model for hippocampal cognitive prosthesis: memory facilitation by hippocampal ensemble stimulation. IEEE transactions on neural systems and rehabilitation engineering : a publication of the IEEE Engineering in Medicine and Biology Society, 20 (2), 184-97 PMID: 22438334


One thought on “Silicon replacement parts for the brain

  1. My son was, for reasons unknown, born missing the Corpus Callosum and Septum Pellucidum and diagnosed with Perisylvian Syndrome due to the malformation of both sides of the outer portion of his brain. For obvious reasons, this article interests and amazes me. He is three years old now and I have always wondered if there might ever be some medical advancement or technology that could help him. This gives me hope.

Share Your Thoughts

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>