Published in: Berkeley Neuroscience News | July 2, 2015

by Sarah Hillenbrand

BMI_loopFor most of us, it is easy to recall a time we felt frustrated because a machine failed to do what we asked. Whether Siri misunderstood you, you wrote code that crashed, or you dropped your phone in the toilet, you are probably aware that humans and machines don’t always play nicely together.
Although at times it may not seem like it, our dependence is illusory. For patients who have suffered a loss of function in their limbs, however, the stakes are drastically higher. Such a loss can be devastating. In this case, technology can dramatically improve patients’ lives. Machines may ease, rather than cause, frustration. Yet even modern prosthetics restore only a fraction of the functionality of biological limbs.

Brain machine interfaces (BMIs) offer a potential solution to this problem. BMI research focuses on creating neuroprosthetic devices. They turn brain activity into movement and stimulation into sensation. The hope is that someday BMIs can offer mobility and a sense of touch to patients with movement disorders. The creation of BMIs poses unique challenges that test the limits of our knowledge of brains, machines, and everything in between.

 

Brains

 

Professor Jose Carmena works with BMIs as tools to better understand how the brain learns. In turn, he applies that understanding to creating better neuroprostheses. Carmena is a Professor of Neuroscience, Electrical Engineering and Computer Science and Co-Director of the Center for Neural Engineering and Prostheses (CNEP) at UC Berkeley and UCSF.

Recently, Aaron Koralek, a graduate student in the Carmena lab, demonstrated that rodents could achieve control over highly abstract tasks using BMIs. They employed connections between evolutionarily older “gas and brake” type motor pathways in the brain and the newer, “smarter”, more “human” areas responsible for things like logic and rule-following.

BMI_moveBy identifying the brain circuitry behind learning, researchers can develop neuroprosthetics that give patients both precise control and a natural feel. “The brain has to own the prosthetic,” says Carmena. “You want naturalistic control without it demanding all of your concentration.”

To achieve this, Carmena seeks to divide the labor of learning between brain and machine. The brain’s ability to learn is known as neuroplasticity, while the process of developing smarter algorithms to decode the brain’s activity is known as machine learning. By striking the right balance, BMIs can exploit the best features of biological and synthetic systems.

Maintaining the stability of the circuit over days to preserve motor memories is also essential. Ideally, an artificial limb should flexibly recreate a patient’s repertoire of motor behaviors without having to recalibrate in between sessions. It should also be robust to interference from simultaneous tasks. In other words, a BMI patient should be able to walk and chew gum at the same time.

BMIs that meet these criteria for reliability are on the bleeding edge of neurotechnological capability. The implants must be reliable, mobile, and long-lasting. Fortunately, Berkeley neuroscientists have access to insights from some of the brightest minds in circuit design, often conveniently located right next door.

 

Machines

 

thinkerFrom an engineering standpoint, BMIs need to be able to sense the outside world, store and communicate data about that environment, and possibly even stimulate the brain. All these activities require energy. If the implant will be with a patient for the rest of her natural life, powering it becomes a very real obstacle.

Currently, brain implants are connected to a power source by passing wires through the skull. These can heat up and cause infection. Furthermore, carrying a battery around isn’t high on any patient’s wish list.

Elad Alon, Professor of Electrical Engineering and Computer Science, heads a project at the Center for Energy Efficient Electronics Science. The Center’s primary goal is to develop electronics that consume dramatically lower amounts of energy – good for our changing climate, good for BMIs. Alon makes energy-efficient integrated systems, which are circuits that electromagnetically harvest energy from their environment. Instead of the patient removing and recharging a battery, it recharges itself. So much for wires and infection.

Jan Rabaey, Professor of Electrical Engineering and Computer Science, finds that BMIs are a good testing ground for his goals in circuit design. Rabaey’s circuits push the limits of miniaturization, allowing researchers to listen in on even single cells in the brain. “BMI, with its need to directly read and write into neurons, as well as its compelling applications, is a perfect match to this goal,” says Rabaey. That is why, in 1999, he founded the Berkeley Wireless Research Center. The Center has served as an accelerator, translating research into clinical applications more quickly and efficiently.

 

…And everything in between

 

Berkeley’s BMIs combine state-of-the-art gadgetry with the latest in neuroscientific advances. Michel Maharbiz, a Professor of Electrical Engineering and Computer Science, finds himself happily situated at this intersection. Maharbiz aims to hybridize the best of biology and the best of silicon chips. “I like building these weird little gadgets that sit in the middle between the biological and non-biological and try to do stuff,” says Maharbiz.

Neural Dust: An Ultrasonic, Low Power Solution for Chronic Brain-Machine Interfaces. Dongjin Seo, Jose M. Carmena, Jan M. Rabaey, Elad Alon, Michel M. Maharbiz. arXiv:1307.2196.
Neural Dust: An Ultrasonic, Low Power Solution for Chronic Brain-Machine Interfaces. Dongjin Seo, Jose M. Carmena, Jan M. Rabaey, Elad Alon, Michel M. Maharbiz. arXiv:1307.2196.

Recently, graduate student Dongjin Seo published a theoretical paper with Carmena, Rabaey, Alon, and Maharbiz, to present an idea that had been percolating amongst the five of them. They proposed using ultra-miniaturized recording devices called “neural dust”. The miniscule sensors could potentially be permanently embedded in brain tissue to continuously record neural activity. The implementation still faces obstacles, but their proposal could overcome one major hurdle. Combined sensing and data transmitting nodes could use ultrasonic frequencies to carry information as well as harvest electromagnetic energy.

Maharbiz’s group also works on a variety of gadgets that are ready for commercialization now. Maharbiz and Rabaey, along with three UC Berkeley graduate students, co-founded Cortera Neurotechnologies. This startup makes recording devices used for micro-electrocorticography (micro-ECoG). In current procedures for ECoG implantation, a neurosurgeon lays a sheet of sensors directly on the surface of the patient’s brain. Micro-ECoG shrinks these sensors to allow for finer-scale recordings.

Until now, BMIs had relied on more invasive electrodes that penetrate the brain’s surface, while ECoG had been used for recording from the brain, but not for controlling a BMI. With micro-ECoG recordings, BMIs may be controlled from the brain’s surface, providing better signal than ECoG and a less invasive, less risky implantation process than penetrating electrodes.

Maharbiz’s group and others are part of an emerging subfield in neuroscience. Their goal is to figure out exactly what kind of information micro-ECoGs pick up, what to do with it, and whether it can be used to control a BMI.

 

The future: SUBNETS

 

The Defense Advanced Research Projects Agency (DARPA) recently awarded Berkeley BMI researchers a grant for a project called Systems-Based Neurotechnology for Emerging Therapies (SUBNETS). The new project’s goal is to design implants that can detect signs of anxiety and depression. These same implants may then stimulate the malfunctioning circuits to correct or reroute activity in them.

A BMI that can target specific areas in the brain can provide personalized, targeted treatment for individuals’ symptoms. In contrast, taking a pill is untargeted, as the medicine circulates through the bloodstream and may cause side effects. Stimulating the dysfunctional circuits more precisely could alleviate symptoms, and could be used to treat problems as severe as post-traumatic stress disorder (PTSD) or traumatic brain injury (TBI), both common among returning war veterans.

Edward Chang, neurosurgeon and neuroscientist at UCSF, oversees the SUBNETS project as Principal Investigator. Chang coordinates the efforts of researchers in multiple fields, including Carmena, Alon, and Rabaey. Joining them on the SUBNETS team are Berkeley professors Jon Wallis and Robert Knight in Psychology and Neuroscience and Fritz Sommer of the Redwood Institute for Theoretical Neuroscience, and Michael Merzenich of Posit Science, a company that makes mobile brain training apps. Together, the team is now embarking on a quest to bring the latest in BMI training to patients not only in their hospital rooms, but on their mobile phones as well.

The SUBNETS project promises to greatly expand the reach of emerging BMI technologies to include neuropsychiatric disorders. “It’s not that we’ve solved sensorimotor problems – that’s still a huge challenge,” says Jen Sloan, the Managing Director of CNEP. “But we are starting to apply what we know towards creating new treatments.” If the next ten years of BMI research are as fruitful as the last, then we can expect vast improvements in treatments for motor and cognitive disorders alike.

 

Labs and projects:

Center for Neural Engineering and Prostheses

Center for Energy Efficient Electronics Science

Berkeley Wireless Research Center

Systems-Based Neurotechnology for Emerging Therapies