A team of University of Arkansas for Medical Sciences (UAMS) research scientists recently began using a powerful computer system that they hope will aid them in better understanding the mechanisms of mental illness.
In June, the Brain Imaging Research Center (BIRC), a division of UAMS’ Psychiatric Research Institute (PRI), installed a Dell PowerEdge system, equipped with two 20-core central processing units, three NVIDIA Tesla P100 graphics processing units, and nearly one terabyte of memory, as well as 500 terabytes of digital storage. A terabyte is 1 trillion bytes.
The purchase, funded in part by significant gifts from PRI donors, will enable researchers to conduct computationally intense data analysis tasks like deep learning--a powerful new form of artificial intelligence--on the brain imaging data acquired from up to 17,000 research subjects, more than one hundred times larger than the current capacity of the research center’s computers. This population-level scale of analysis will allow the researchers at UAMS to identify the complex, hidden patterns of brain organization driving human behavior and cognition, which are the roots of mental health and illness.
“Understanding the complexity of organization of the human brain has advanced greatly in recent years through the use of magnetic resonance imaging scanners like the one we have at the Brain Imaging Research Center,” said the center’s director, Clint Kilts, Ph.D. “This new computer uses next-generation technology to analyze complex spatial and temporal information within brain imaging data and will support data sharing with research imaging groups worldwide. It will power the Brain Imaging Research Center’s mission to understand the brain mechanisms of mental health and illness, create state-of-the-art learning opportunities for our trainees and enable Arkansas to contribute at this scientific forefront.”
A single magnetic resonance imaging (MRI) scan has 140,000 voxels, according to Keith Bush, PhD, a computer scientist and assistant professor of psychiatry in the research center. A voxel (volumetric pixel) is the three-dimensional counterpart of a pixel. Most research studies require as many as 80 scans, said Bush, requiring most facilities like the BIRC to compress the images down to 200 to 300 data elements. With the new computer, research center staff will be able to analyze brain images faster, and with less data reduction, he said.
"We’ll be able to complete analyses in a week that would have taken us months in the past, and work with larger imaging data sets on the order of 17,000 individuals, roughly the size of cities like Bryant or Maumelle,” said Bush. “We’re looking forward to working with large data repositories generated by the BIRC or contributed from national and international imaging research groups, something we couldn’t have done before.”
Ming-Hua Chung, PhD, a mathematician and post-doctoral fellow whose graduate thesis research explored the mathematical means of analyzing large volumes of data, plans to use the computer system to analyze data collected in BIRC studies over the last seven years.
“We’ll be looking at factors such as child abuse that causes users to abuse drugs, to study the impact trauma has on the brain,” said Chung. “We’ll look at age and sex differences and their impact on drug use differences. There is so much potential here to analyze raw data sets that were too big in the past.”