With the help of TACC
supercomputers, researchers create OpenfMRI project to enable data intensive
analyses of the mind. Researchers of neuroscience at The University of Texas at
Austin, bridge psychology, neuroscience and computer science to understand how
the brain creates cognitive functions. fMRI machines map neuronal activity
based on the blood oxygen levels in the brain. When a neuron is active, the
brain sends extra oxygenated blood, which has a distinct magnetic signature. By
recording these signatures at different locations in the brain, neuroscientists
can connect and pinpoint various functions―and potentially dysfunctions―with
amazing specificity. The age of telepathy is nearly upon us. Brain researchers
can now identify what you are looking at using only fMRI scans of your brain.
However, actions, motivations and feelings are still hard to identify. They
created OpenfMRI, a web-based, supercomputer-powered workflow that makes it
easier for researchers to process, share, compare and rapidly analyze brain
scans from many different studies.
Currently, the project has 18
datasets, consisting of data from almost 350 human subjects. The data comes
primarily from four main partners―Stanford, Harvard, the University of Colorado
and Washington University. The pipeline that researchers developed allows to
automatically process, visualize and analyze raw fMRI data, using the powerful
Lonestar supercomputer at the Texas Advanced Computing Center (TACC). When fMRI
scans are taken, they contain a lot of noisy information that must be cleaned
up. In the automated workflow, the supercomputer first determines what parts of
the fMRI images represent brain tissue. Next, it computationally reconstructs
the 3D surface of the brain based on structural images and projects the data
from the fMRI scans onto that surface. Finally, it takes each subject's brain
and warps it to correspond to the average brain so a researcher or doctor can
ask, across a group of individuals, which areas are turning on during a
specific activity. Each of these steps requires large-scale computational
power, but they can be done quickly, using Lonestar.
More information: