Sometimes, a new problem calls
for an old solution. For Mukesh Dhamala, assistant professor of physics, that
meant going back to an early-19th-century theory to refine his
analysis of information flow inside the brain.
Dhamala’s latest research
appears in a paper he co-authored that will be published in the January 11 edition
of Physical Review Letters. Basically, the paper
addresses the problem of mapping information flow patterns in dynamic systems -
for example, the patterns of neural interactions in the brain that underlie thought
and behavior.
Dhamala and his co-authors argue
that the best way to approach the issue is through two relatively old systems
of theoretical physics. Their work uses Fourier transforms, equations first
laid out in the early 19th century, and continuous wavelet
transforms, introduced in the seventies. These methods are widely used in
science and engineering, but have not yet been brought to bear on these particular biological problems. The old equations, Dhamala says, offer a better way to approach the
question of cause and effect inside the brain – that is, to figure out which
way impulses flow through chains of neurons.
Physics equations aren’t an obvious
choice for understanding biological problems. But new advances in medical
imaging have produced new kinds of data for scientists to work with. Scientists
are still looking for the most accurate – and the most revealing – ways to
analyze that data.
Up to now, researchers have been
using what are known as parametric modeling techniques. In these methods,
researchers first build a mathematical model for the data at hand and then use
the model to analyze the data.
There are two problems with the
parametric approach, Dhamala said. First, the results are unavoidably shaped by
the model itself – build a different model, and a different analysis will come
out.
Second, the models often fail to
fully account for the complexities of the data, Dhamala said. Measurements are
now so fine that scientists can see the spikes of electrical activity as an
individual ion passes from neuron to neuron. But parametric models are not suitable
for analyzing these “spike trains.”
The nonparametric approach that
Dhamala and his colleagues have proposed can handle that level of detail,
however. Their methods allow researchers to calculate their results directly
from the data, without building a mathematical model first.
“It’s just a more principled way
to see the data,” Dhamala said.
Dhamala, who is also associated
with the Brains and Behavior program and the Center for Behavioral Neuroscience
at Georgia State, has a Ph.D. in theoretical
physics. His research, however, straddles the line between physics and biology,
using the theories of the former to dig through the data of the latter. It’s an
unusual niche, but an important one.
“The brain is so complex,” Dhamala
said. “However, modern biomedical imaging/recording technology offers us a
unique opportunity to increase our understandings of the brain, both in health
and disease. But, the analysis requires a multidisciplinary approach.”
