Head Detectives

UNM engineers investigate the mysteries of the mind

With its high ceilings, displays of modern art, and expansive views, this gleaming building certainly doesn't look anything like your typical detective office. Then again, your typical gumshoes don't work here. These special investigators aren't searching for clues to a crime; they're trying to crack a much more difficult case: how the mind works.

This modern HQ for these high-tech detectives is The MIND Institute, a non-profit neuroscience research center just north of UNM's main campus. Multidisciplinary teams at the institute collaborate with researchers around the country and use the latest imaging technology to sleuth for new ways of understanding human behavior, especially mental illness.

Collaborating to Find Clues

One of the head detectives is Vince Calhoun, director of image analysis and MRI research at The MIND Institute and associate professor of electrical and computer engineering at the School of Engineering. With his background in biomedical engineering and years of experience working with psychiatrists at The Johns Hopkins University School of Medicine Department of Psychiatry as well as Yale University, Calhoun is uniquely qualified to apply engineering approaches to an investigation of how the brain works. The MIND Institute is a perfect match for his special qualifications. "The MIND is one of the few centers of its kind in the country," says Calhoun. "What's unique is its focus on mental illness and the fact that it has all of this excellent technology in one place."

Calhoun is also adept at creating successful collaborations between engineers and psychiatrists, who have distinct approaches to research. "I've seen good ideas from engineers go to the point of a paper and then stop because the engineer doesn't know how to address questions of relevance in psychiatry. And psychiatrists don't understand how to read an enginee's paper," explains Calhoun. "So, I've learned how to work with psychiatrists, but think about things from an engineering perspective."

Detecting Differences

Calhoun leads a multidisciplinary team investigating how brain imaging can unlock clues to better diagnosis and treatment techniques for schizophrenia, which affects 3.2 million Americans. The team is using brain scans to look for indicators of the disease. They're also creating powerful new tools to analyze the data they collect.

Using high tech imaging systems, including magnetic resonance imaging (MRI), functional magnetic resonance imaging (fMRI), and electroencephalography (EEG), the team is searching for biomarkers, or measurable changes in the brain that indicate schizophrenia.

Calhoun uses the imaging tools to collect brain scans while subjects perform simple memory or auditory tasks. Each image type yields different information about how the brain works. For instance, fMRI provides a dynamic view of blood flow in the brain, while EEG picks up electrical fields created when neurons fire in the brain. By comparing images from a control group with those from a group of patients with schizophrenia, the researchers can see differences in how the brain works. Those differences could point to possible biomarkers for the disease.

Calhoun and his team are also developing better ways to analyze and synthesize all of the data from hundreds of patients who have had multiple images made during numerous tests. Calhoun already has an analysis toolbox online. Now he and his team are developing more robust tools to first analyze each data type independently, and then to conduct "data fusion," analysis of combinations of images, such as fMRI and MRI, or EEG, fMRI, and MRI. "The goal is to combine the information in a way that gives us a better ability to look for differences in the brain," explains Calhoun. The National Institutes of Health is funding his data fusion study with a $1.5 million grant. His approach, called joint independent component analysis, decomposes the high dimensional data into a smaller set of independent parts. The different types of data are grouped together in the algorithm, and Calhoun then uses an approach based upon information theory to identify which parts of this massive decomposition are the most informative about differences between patients and healthy controls.

Cracking the Code

Calhoun is not alone in his investigation. Terran Lane, assistant professor of computer science, is also on the case. Lane, too, is investigating data fusion and is just starting to develop a sophisticated joint probability model that can combine the different types of data.

Lane is collaborating with Vince Clark, associate professor of psychology and scientific director of The MIND Institute, to study networks in the brain. "We're trying to understand how different regions of the brain influence each other and send messages to each other," says Lane. "The goal is to untangle the story and pull out a network." Lane receives data directly from MIND researchers and loads it on to computers in the Computer Science Department. Then comes the real challenge: the brain's complexity and the immense amount of data involved make teasing out relevant networks very difficult. "It's as if I gave you a list of how many telephone calls enter and leave every building on the UNM campus without telling you who they came from or went to. Then, I asked you to tell me who called who," explains Lane.

Lane is cracking the code using powerful machine learning and statistical techniques. While his tools are decidedly high-tech, he's based his process on a probability theorem, developed in the 18th century, called Bayes' Rule. Using only a small fraction of the available data, software based on Bayesian networks can scan the brain images for possible networks, learn about the data as it goes, and use that information to improve the ongoing analysis process. Lane and senior doctoral student John Burge have created an in-house version of Bayesian network software that includes anatomical and hierarchical information about the brain's structure as it evaluates the structural and temporal data collected by MIND researchers. Lane filters results from the software to improve the statistical confidence, then shares his findings with neuroscientists at The MIND Institute, who provide input and context on the results. "It's a feedback loop. We give them results and they give us back new data and new questions to ask. Hopefully, from this process we'll gain neuroscience consequences and computational consequences," says Lane.

Lane's research, funded by a three-year grant from the National Institutes of Health, is at its midpoint. The team is currently evaluating some new net-works extracted by the software. "In the 20 years that I've been studying the human brain, I have watched the tools for data acquisition become far more sophisticated, while our methods of analysis have remained very simplistic. We only understand about 10 percent of the data we collect; the rest is obscured by its complexity," says Clark. "One of my goals for The MIND Institute is to support the work of scientists like Vince Calhoun and Terran Lane, who can develop more sophisticated methods for analyzing brain data. Their algorithms have already produced a flood of new insights into the kind of data that has been available to us for years, but couldn't be analyzed using available methods."

With clues, collaboration, and cutting-edge technology, UNM's head detectives are piecing together answers that will not only unlock the mysteries of the mind, but could also improve millions of lives.