By: Sara Hoover
Mathematics professor Dr. Robert Kozma may not be bending spoons and levitating glasses,
but he is working on technology that allows people to control machines with their
In his Computational Neurodynamics Lab at the U of M’s FedEx Institute of Technology,
Kozma’s team monitors the cognitive behavior of animals and humans using a nonintrusive
technique, a scalp EEG – an array of electrodes put on the head like a hat. This will
allow for direct brain-computer interface by using brain waves to interact with the
computer. Think of a keyboard-less computer that eliminates the need for hand or voice
“What’s becoming a reality is that you can talk to the computer and it talks back
to you,” said Kozma, who recently moved from the computer science department to mathematics
in the College of Arts & Sciences. “There’s a next step using interaction even without
voice. This type of interface can be used for disabled people and elderly. If they
cannot move or are limited in their movement, they turn on certain devices like the
Kozma’s “brainy” research has
far reaching possibilities.
(Lindsey Lissau photo)
The research has three prongs of use: video/computer gaming; to support people with
disabilities or physical constraints, such as the elderly; and to improve control
of complex machinery, such as an aircraft and other military and civilian uses.
The direct brain-computer interface would give those with physical constraints or
those operating complex machinery “extra arms.”
“The possibilities are limitless because if the principle works, you measure certain
patterns in your brain and that’s what that computer will execute,” said Kozma of
his research that was begun 14 years ago. Later this month, he will become the director
of the U of M’s newly created Center for Large-Scale Integrated Optimization and Networks
(CLION), which covers this brain wave research.
The main collaborator is Dr. Leonid Perlovsky, principal research physicist at the
Air Force Research Laboratory & visiting scholar at Harvard University. The research
is patented through the U of M’s FedEx Institute of Technology and the Air Force Research
The National Science Foundation originally supported the fundamental theoretical research
and mathematical aspects of brain waves. The project has ongoing support from the
Air Force for the modeling aspect and private foundation support.
The other use, not related to movement or controlling a machine, is the bio-medical
application for brainrelated diseases. The lab already collects data from patients
with epilepsy at Le Bonheur Children’s Medical Center and other sources. Additional
behaviors applicable are Parkinson’s, schizophrenia and sleep apnea.
Collaborators working on the bio-medical aspect are: Dr. Walter J. Freeman, professor
emeritus of neurobiology at University of California Berkeley; Don Tucker, CEO of
Electrical Geodesics Inc., a company that produces brain-monitoring devices; and Dr.
James Wheless, chair of the division of pediatric neurology at Le Bonheur and director
of its comprehensive epilepsy program and Neuroscience Institute.
Freeman, who works on discovering brain-wave properties and translating them into
applicable forms for clinical forums, states the goal is “to understand the ways in
which normal neural and mental function can break down” to therefore help those “who
carry the responsibility for treating patients with neuro-psychiatric disorders.”
To that end, graduate students have been very involved and work with Le Bonheur to
do the signal processing, since Kozma’s lab does not yet have equipment in-house.
Computer science doctoral student Mark Myers (MS ’05) is the lead researcher of the
brain modeling and data processing of the Le Bonheur patients. He works with a brain
model called the K4, a biologically inspired neural network that enables modeling
of the limbic system in the brain. It allows Myers to simulate the brain in its healthy
electrical operating state and then move the brain into an off-balance state. This
off-balance state contributes to neural pathologies, such as seizures, Parkinson’s
and schizophrenia. The K4 has been successful in allowing researchers to capture the
brain’s normal state, move into an abnormal state that puts it into seizure behavior
or other states and then adjust the brain back into a normal state to overcome that
By analyzing data from seizure patients age 3 to 12, Myers hopes to improve the quality
of life for patients with a brain implant called the vagus nerve stimulator. Although
considered one of the first brain implants to prove its efficacy in reducing overall
seizure symptoms, it’s on all the time, pulsing at regular intervals much like a pacemaker.
“Our goal is to find the prediction algorithm that once fired up or detected will
enable the pulse generator to only turn on as the seizure is approaching, continue
pulsing during the seizure activity and shut off,” said Myers. “The end goal is for
the individual to not even know they had a seizure.”
They would essentially create an on/off switch for the brain implant. They have discovered
indicators to differentiate pre-seizure states from actual seizure states. With the
on/off ability, the implant would become an individually tailored system for each
seizure patient’s needs.
“There’s a flare-up, much like a sun flare before the actual eruption occurs,” said
Myers. “That’s something we’ve been able to detect through signal processing methodology.
We found several minutes before the seizure event, a prediction marker that looks
just like the seizure event.”
Current users are pediatric seizure patients, but the concept is applicable to adults
and other disorders similar in character, such as migraines or bipolar illness.
“Diseases that are similar in character to epilepsy,” said Wheless, referring to its
other uses. “Where you’re fine, something happens and you’re not fine and then you
recover. Intermittent problems like horrible migraine headaches, we can tweak that
on/off switch to shorten a day or two disability to none or minutes. It works for
neurologic or psychiatric problems that are intermittent but severe when they hit.”
Once the team proves their algorithm enables robust seizure prediction, they plan
to make recommendations to medical communities.
“This study has enabled me to move my career into an altruistic-type setting,” said
Myers. “I can take what I’ve learned and move into an area of research that’s not
only extremely fascinating, but that is good.”
Although the project secured funding for theoretical and movement aspects, it has
not captured support for experiments, which is the next step toward commercialization.
“We have the laboratory experiments as far as the neuroscience is concerned with collaborators
at Berkeley and Le Bonheur, but don’t have support for that,” said Kozma.
Using their patent as a platform, the team has applied for funding through the FedEx
Institute and the National Institutes of Health.
Kozma’s group is also in the early stages of finding brain waves that predict movements,
in the hope of helping individuals with prosthetics.
“There’s brain activity that precedes movement we don’t necessarily think about,”
said Wheless. “If you open the door, you don’t think, ‘I’m going to turn the door
handle, look down at it, make sure my hand is on the door.’ There’s brain activity
that actually precedes those motor movements.”
Similar to how they found seizure prediction markers, the plan is to use the data
to analyze pre-motor movements, the changes in the brain that occur before there’s
actually movement, and apply that to someone who has a prosthetic device to allow
them to better manipulate it.
Since the brain is usually multitasking, Kozma’s team will have to pick up the signal
for the desired task from all the other things going on in the brain. In the next
few years, they will work on reliably finding pre-motor movements in the brain.
When some people hear “brain wave monitoring,” they think mind reading, communicating
with the dead and lottery numbers.
“People always get concerned when you’re looking at brain activity,” said Wheless.
“We’re not looking at what you think of your mother. You can’t interpret those kind
of things. It’s looking at brain patterns that predict things like seizures and movements.”
Myers is eager to continue his research under the tutelage of Kozma and be part of
the changing landscape of human-computer interaction.
“Currently, brain-computer interfacing is in its extreme infancy stage,” he said.
“I love this stuff. I’m absolutely in love with it. Short of my wife and kids, this
is up there.”
To see a video of Kozma's research, click here.