 |
 |
 |
 |
| |
|||||
|
|
|
Home > News & Events > MITRE Publications > The MITRE Digest > | |||||||||||||||||||
MITRE Researchers Have Brains on Their Minds July 2003
Imagine having clues to a treasure on bits of paper in thousands of different boxes. How could you possibly merge, sort, and organize this information to put together a map? How could you make the information accessible and usable, and thus valuable? Welcome to the world of neuroinformatics—the science of collecting and managing neurological information and imagery, as well as providing computational tools to study the data. Neuroinformatics, in essence, provides ways for researchers to begin drawing a treasure map leading to a greater understanding of human brain structure and function. Over the past few decades, numerous advances in neurological imaging, including MRI (magnetic resonance imaging) and PET (positron emission tomography), have given medical researchers a window into the working human brain. But although this enables neuroscientists to gather more data, a lack of information infrastructure for sharing images and associated information outside individual labs reduces their ability to form and test hypotheses. A wealth of data resides around the country, but it typically remains in its lab of origin. Expertise in managing data is what MITRE brings to the challenge. Researchers Ken Smith, a database engineer, and Monica Carley, a signal processing engineer, are part of a team working with the National Institutes of Health (NIH) to find the best way for the medical community to organize and share its information. MITRE's Center for Enterprise Modernization facilitated the partnership between our neuroinformatics team and NIH. They haven't relied purely on prior experience to tackle the challenge, however. The two have made several visits to neuroscience labs, "field trips" that have granted them a special perspective on the world of brain mapping. "During these visits, we get to observe the experimental protocol and ask the radiologists, neurologists, and neuroscientists questions about their research," Carley says. "This gives us a better idea about what is most important in the data they collect and how it's processed." This attention to detail has paid off in unexpected ways, yielding insights into data and its analysis the team hopes will enable neuroscientists to share basic research. The neuroscience community most likely won't be the only beneficiary of MITRE's research, however. Smith and Carley envision the work one day seeing broader application in areas such as homeland defense, by permitting diverse agencies to share data while maintaining individual privacy.
Sharing and Comparing The problem of sharing the wide variety of brain imagery has concerned the neuroscience community for years. In 1993, a group of topflight medical research institutions founded the International Consortium on Brain Mapping (ICBM), funded by a grant from the National Institute of Mental Health (NIMH). The Consortium's main goal is to develop a reference system for the human brain, a sort of encyclopedia of the physical aspects of the mind. To achieve this, ICBM members knew they'd have to improve data and imagery sharing among research labs worldwide. For several years, they looked unsuccessfully for a research partner in the database/IT community. In 2000, MITRE scientist Jordan Feidler began collaborating with ICBM and created the MITRE neuroinformatics project, the program currently led by Smith and Carley. The work is funded by MITRE (because of the significance of the underlying technologies to our sponsors) and a grant from NIH, administered by NIMH. When the MITRE neuroinformatics team members began their research almost two years ago, they and their colleagues in the neuroscience community worked from the same basic assumption—that MITRE and its ICBM partners would create a single, large database of MRI scans in one repository. In essence, the neuroscience world would come to the database. As the MITRE team came to understand the specific concerns of the neuroscience community better, however, this vision evolved and expanded. "We began looking more deeply at the problem of sharing, because we learned early on that sharing this kind of imagery is extremely hard," explains Smith, principal investigator for the NIH grant. "So over time we formed an idea: what if we developed the tools with open-source code and gave them away, rather than having a monolithic database held by us? That way, if a laboratory doesn't want to send data to us, we can still ship software to them." By essentially taking a miniaturized version of the big database's structure and distributing it to the neuroscience community as free open-source software, the MITRE team hopes neuroscience facilities will choose to develop their own databases locally and share their information peer-to-peer with selected partners. Institutions will also be able to use the tools as a stand-alone data management system, whether or not they enter into data-sharing agreements with others. The centralized repository will no longer be the only way to share data; instead, it will become part of a wider range of resources available to the scientific community. A Two-pronged Attack Smith and Carley, with about six other MITRE staffers, work in parallel on the two main pieces of the program: database management and image processing. Both are areas in which MITRE has years of expertise, working on various projects for our sponsors. The MITRE project focuses on what is called structural MRI data, derived from the noninvasive imaging technique that reveals brain anatomy at a resolution of a single millimeter. With this data, detailed anatomical maps can be formed for individuals or arbitrary groups of people. These maps can in turn be used to compare diseased and normal groups, giving anatomists a deeper understanding of the human brain. The neuroinformatics system's basic architecture consists of two major elements, a digital library and a data warehouse. The first step in the process is the creation of the digital library to hold the images, along with their associated metadata. "Metadata" identifies the demographic qualities of the subject, including age, gender, and ethnic or racial group. The metadata typically also includes other health-related data, such as whether or not the individual is a smoker and results of any neurological tests that have been administered. The images will have no personal identifying information, however; all images maintain the patient's anonymity. The MRIs are also "cleaned up" using sophisticated signal processing techniques to account for variation caused by different equipment, patients' head movements, and so on.
Users will be able to send queries to the library using specific metadata parameters. Before their requests are processed, however, the requesters will be screened based on their affiliation (to a laboratory, peer group, or research institution), so they receive only those images that fit previously agreed-upon sharing and privacy protocols The second major component of MITRE's system architecture, the neuroinformatics warehouse, is more complex. Images from the library will undergo a refinement process, enabling researchers to explore and analyze images using a wide variety of specialized query techniques. Among other approaches, Carley and her team have been investigating the use of content-based retrieval for the system. This works well with large databases and supports a concept called "query by example," where the user submits an image to the data warehouse, which then looks for possible matches. "We want to create tools for testing hypotheses," Carley says. "For example, a researcher might suspect a correlation between the shape of a structure in the brain and some attribute, such as right-handedness or the presence of a disease. If the system could produce a set of scans having a similarly shaped structure, the researcher could then examine the attributes of those subjects and effectively test his or her hypothesis. "Another one of our aims is to dynamically create atlases of the brain. For example, if you were studying the progression of Alzheimer's in an 80-year-old Asian woman, for comparison you may want to get a sense of what the average brain looks like in this demographic group. So the ability to build an atlas on the fly for healthy 80-year-old Asian women would be very beneficial." Thinking about the Future While work in the fields of biotechnology and bioinformation is relatively new to MITRE, the team found most of the necessary technical expertise for the project, such as information technology and signal processing, available within the corporation. "MITRE has a tremendous background in information interoperability, which really helps," Smith says. "We want to avoid creating large inflexible data structures that don't facilitate peer-to-peer sharing of data. We're looking forward to bringing our expertise to bear in neuroinformatics, especially for labs that are 'have nots' in terms of information resources and tools." Ironically, in the hypercompetitive world of academic medical research, MITRE's position as an "outsider" works to our advantage. "I think we're viewed differently from a competitive lab," Carley says. "Our collaborators like our impartiality; we're not competing with them. Our success [as a not-for-profit organization working in the public interest] is measured differently. As a result, we've had several collaborators agree to share their data with us. They view us as a resource." The team members, who are ready to release the first set of database tools to the neuroscience community by mid-2003, have already begun looking ahead. Future plans include improvements to the data querying system (especially the content-based retrievals), applications of data mining techniques (a field in which MITRE is a recognized leader), and explorations of complex issues involving information security and privacy. Team members will also be publishing papers and presenting their findings at technical conferences. They believe the knowledge MITRE offers to the neuroscience community will eventually help our other sponsors as well, many of whom face similar challenges in handling large databases of sensitive or difficult-to-analyze material. "For example, data sharing research could have an impact at the IRS and in homeland security, where it's imperative that information flow occurs in carefully defined ways that respect privacy laws," Smith says. "The content-based retrieval system would also be valuable to our sponsors," Carley adds. "The Intelligence community and the Navy, among others, rely heavily on information that they extract from the various types of imagery they collect." "This is a challenging project and will help us learn a lot, not just in neuroscience, but in many related areas," Smith says. And just like the neuroscientists they're working with, the MITRE neuroinformatics team will keep following the clues in the right direction. —by Alison Stern-Dunyak Related Information Articles and News Websites |
||||||||||||||||||||
| Page last updated: February 12, 2004 | Top of page |
Solutions That Make a Difference.® |
|
|
