In Network
By Kathy Hardy
Radiology Today
Vol. 22 No. 6 P. 16
Functional and structural MRI advance brain network research.
Like-minded people can use social media tools to bridge geographic gaps and make connections. Similarly, when it comes to networking within the human brain, researchers are finding validity in a similar theory for connectivity, and they’re using functional and structural MRI (fMRI and sMRI) data as a guide.
Neuroimaging and neuroscience experts from several locations, including The University of Texas Health Science Center at San Antonio (UT Health San Antonio) turned to BrainMap, a database of more than 20,000 published pieces of literature stemming from fMRI and sMRI neuroimaging experiments established by the university’s Research Imaging Institute to learn more about how networks are organized in the brain. Their work, recently published in Communications Biology, helped to confirm a relationship between network-based normal behaviors and brain disorders. They compared the connectivity patterns of large-scale functional networks used in normal behaviors with disease-related coalteration networks and found areas of overlap.
“There has been evidence of coactivation patterns and coalteration patterns in the brain for some time,” says Peter T. Fox, MD, director of The Research Imaging Institute at UT Health San Antonio. “This paper presents evidence that when brain disorders cause structural changes (atrophy), the changes follow the established functional coactivation patterns of normal behaviors. There’s a definite shift now to figure out disease networks. Just like a circuit board, there’s not just one area of the brain that changes. It’s a distributed process.”
The big picture view provided by fMRI and sMRI via the BrainMap database is that, when it comes to learning how the brain is organized, more data is better.
“Functional and structural MRI have been the most dominant approach to studying brain stimulation in the last 30 years,” says Simon B. Eickhoff, MD, PhD, director of the Institute for Systems Neuroscience of the Heinrich-Heine-University in Dusseldorf, Germany. “The problem is, what we can learn from one study is limited. If you combine studies, you can see where there are consistent findings, giving you a much higher level of confidence in those findings.”
Eickhoff, a coresearcher on the article, works on understanding the organizational principles of the human brain by an integrated analysis of structure and function, noting differences and similarities in their connectivity patterns through covariance analyses. Methods used to do this include fMRI, various mapping techniques, connectivity analysis, and modeling.
BrainMap
The meta-analytic network analysis that is valuable in the study of brain organization is enhanced by access to platforms such as BrainMap, a database of published functional and structural neuroimaging experiments. Originally developed in 1988 as a web-based interface, BrainMap has evolved into a broader tool for sharing neuroimaging results. It is designed to enable meta-analysis of studies of human brain function and structure in healthy and diseased subjects.
“It’s not just a library,” Fox says. “It’s a collection of functional activation literature and the as-published data. We clean, validate, and categorize all the data, [utilizing] categorization-applied terms that are useful for retrieving specific data that can be pooled for a meta-analysis. Other researchers looking through the data can see where we have enough and where we need more. This can encourage them to send us additional datasets for entry.”
As Eickhoff explains, the BrainMap database consists of published functional and structural neuroimaging experiments with coordinate-based results (X, Y, Z locations), rather than the actual images. He says this is an easier format in which to search for and combine data.
“Because fMRI is highly standardized, each study reports in the same X, Y, Z coordinates,” he says.
There is also an interactive aspect to BrainMap, and the development team welcomes collaborations with other researchers, sharing data or assistance in the execution of meta-analyses. This can result in further research in specific areas, as well as new literature for the database.
“We have a lot of adult studies in ADHD, but not many involving children,” Fox says. “We can see where we have a deficiency of data. We can’t tell people where to do their research, but we can say where we have a need for data and help direct researchers to focus on that area. It motivates people to help fill the void.”
Eickhoff sees the possibilities of research stemming from BrainMap limited only by the inquisitiveness of those who conduct research on brain networking and disease progression.
“You either know what you want to study, or you can look through the BrainMap database and see what might be a good area to study,” he says. “You can come in looking for a connecting pattern in one region of the brain and ask for all the studies that activate in that region, as well as a coactive region. It’s then just a small step to look at network productivity.”
Roadmap for Disease
Researchers note that system-level functional networks are defined by their functional connectivity, which is a good metric for studying behavior, cognition, and disease. There are 15 to 20 readily identifiable functional networks that can account for much understanding of brain-behavior ontology—concepts and categories in a particular subject area that show their properties and relations between them. These functional circuits lend themselves well to the study of certain psychopathologies and relevance for transdiagnostic investigation.
In their latest work on brain networking, Fox, Eickhoff, and other researchers set out to see whether the brain’s complex architecture of functional, information-processing networks is the target of disease and, if so, which networks are associated with which diseases. The goal is to learn more about the underlying causes of brain disorders.
In the early years of research in this area, Fox says the focus was on what occurred in each area of the brain, eg, which areas of the brain turn on when someone speaks. That’s more of a module-by-module view of the brain. He refers to the emerging understanding of functional network architecture in the brain as a significant advance in the fields of imaging and neuroscience.
“There’s a hypothesis that some disorders show a pattern of progression that follows known brain networks,” Fox says. “Those disorders progress in a manner following functional pathways.”
Their published study of 43 psychiatric and neurologic brain disorders affirmed the hypothesis of network-based degeneration—the idea that disease-related structural damage invades the functional networks in the brain used in human behavior and repeats within “coalteration networks.” Study findings indicate that metabolic stress in high-traffic areas of the brain is a key underlying cause of network-based degeneration.
The network degeneration hypothesis posits that disease-related structural alteration occurs selectively and possibly spreads within system-level functional networks. Neurodegenerative disorders spread along specific neuronal brain networks, leading to functional impairment of these networks. Within that hypothesis is the concept that structural alteration in one area of the brain is influenced by alteration in other areas of the brain. This concept is referred to as coalteration structural connectivity. Researchers continue to look at the level of communication between structural and functional activity, with new evidence suggesting that the variety and unpredictability of diseases that have structural effects on a particular area of the brain may be associated with regions that are important for cognitive and interactive function.
Specifically, the latest research revealed that 14 of the 20 disease-related coalteration networks spatially conformed to functional networks involved in normal behaviors, such as movement, emotion, language, and problem solving, to a high degree. Other findings indicate that the network associations of neurological diseases are stronger than those of psychiatric diseases. In addition, some diseases have broader effects across networks than others. For example, Huntington’s disease affects nine networks, while major depressive disorder affects just two.
“This gives us new ways to look at disease,” Fox says. “If you can learn the network for a particular disease, you can test for it. You can have a roadmap for how to explore diseases.”
Role of MRI
For human neuroscience, fMRI and sMRI, combined with data-driven analytic methods applied at the system level, have been impactful in making progress over the past 30 years. Since its inception in 1990, fMRI has been used in an exceptionally large number of studies in the cognitive neurosciences, clinical psychiatry/psychology, and presurgical planning. Fox says fMRI has been a standard modality for determining brain networks since its inception in 1991.
“To look at a region of the brain, you scan the whole brain over time,” he says. “You would see the areas where activity goes up and what other areas go up with it. This was a method for connectivity mapping.”
Fox says researchers can use these results as a roadmap for more specific investigations, given that biologically specific regions of interest within the brain can be derived from the component maps shared in this study.
Eickhoff anticipates further research in how brain networks function, and he expects BrainMap to play a role in contributing to the future of machine learning in this area.
“Consolidating literature in a database like BrainMap helps prioritize information for use in machine learning,” Eickhoff says. “You can use BrainMap to constrain the search space for analysis and help to better focus on data that are relevant.”
BrainMap Community Portal
Seeing the potential for future successes based on the BrainMap database, the National Institutes of Health recently renewed funding for the program for four additional years. Fox notes that continued funding for BrainMap is crucial for continuance of this program.
One area that will see a benefit from this financial influx is the Texas Advanced Computing Center (TACC) at The University of Texas at Austin’s project to create a BrainMap Community Portal backed by high-performance computing resources. This tool, still in development, is designed to make large-scale, complex, multivariate analyses of this type more accessible to the research community at large.
“The function of the BrainMap Community Portal will be to provide an intuitive and powerful web interface for community researchers to use the BrainMap analysis tools on some of the world’s most powerful computing resources,” says William (Joe) Allen, PhD, a research associate in the Life Sciences Computer Group at TACC. “The design of the portal will be based on our existing portal infrastructure that we have used for other funded projects. Those funded projects include communities of researchers interested in natural hazards, synthetic biology, 3D electron microscopy, and acute pain biomarkers, among others.”
Allen explains that with the currently available BrainMap data and analysis tools, community members download the tools to local machines or lab computers, download data from the website, and run analysis locally. Depending on the complexity of the analysis, this could occupy local computers with expensive calculations and large datasets for days or weeks.
“TACC’s expertise lies in building and operating large supercomputers, as well as designing powerful and intuitive interfaces to those supercomputers,” Allen says. “In this collaboration, we plan to make the BrainMap resources—in particular, the datasets and analysis tools developed by UT Health San Antonio—accessible for anyone to run at a large scale through a simple point-and-click web interface. The data will reside on TACC storage devices, meaning it does not need to occupy individual users’ personal hard drives; analysis tools will run on TACC clusters, meaning they do not occupy the available processing power on researchers’ personal computers.”
In addition, to help promote scientific reproductivity and reliability of results, Allen says their goal is to capture the intent of the original developers of the tools.
“It is not just important to be able to run the analysis tools.” he says. “The user community should be able to run the tools in the exact manner intended by the developers of the tools.”
Plans are to have the portal up and running in a year, and Allen anticipates that most users of the current BrainMap web portal will migrate to this new portal.
— Kathy Hardy is a freelance writer based in Pottstown, Pennsylvania. She is a frequent contributor to Radiology Today.
