News

Amarnath Gupta, a researcher at the San Diego Supercomputer Center (SDSC) at UC San Diego, is presenting an invited plenary talk at the Imaging, Medical Analysis and Grid Environments (IMAGE) 2003 workshop ( http://www.nesc.ac.uk/esi/events/image03/).

The IMAGE workshop, held September 16 - 18 2003 at the UK e-Science Institute in Edinburgh, Scotland, is investigating key challenges in the handling and use of medical image data distributed across multiple sites, and opportunities for international collaboration including the use of Grid technologies.

Gupta, director of the Advanced Query Processing Lab (AQP) in SDSC's Data and Knowledge Systems (DAKS) program, is giving a presentation titled "Neuroimaging Databases - a Data Engineering Perspective," which covers lessons learned in extensive experience working on technologies to allow efficient searching of multiple neuroscience data collections.

The value of this research is that it will enable medical investigators to simultaneously explore numerous, much larger, and more diverse data collections distributed across different sites, and this will yield new insights into neuroscience and many other medical areas. Potential medical benefits include understanding the basis of diseases such as Parkinson's disease, which is being studied by comparing the human form of the disease with animal models in which parts of the disease are produced by methods such as genetic modification or drugs.

"There is a rapidly growing body of neuroimaging data collections in neuroscience laboratories around the globe," said Gupta. "But our ability to adequately model, query, integrate and derive insight from this important data is still a major challenge." Difficulties stem from the complexity of the data, the variability of data across populations, the lack of precise, machine-usable definitions of conceptual terms, the possible dependence of observations on the particular instrument of observation, and other factors. In addition, there is the presence of significant implicit, unformalized semantics in the data, which makes it very difficult to develop general-purpose data management solutions in such medical imaging applications.

"In the course of our work to build database systems for neuroimaging, we have identified a number of data engineering problems and developed partial solutions for some of them," said Gupta. The talk presents a tour of these problems and proposed solutions, focusing on the management of multi-scale, 3-D, time-varying information.

"An overarching theme is the clear lesson we've learned that success in building real-world neuroimaging databases comes only when neuroscientists and database scientists join hands to overcome these challenges together," said Gupta. "At the same time, database scientists need to venture across their own internal sub-disciplinary boundaries in order to develop useful hybrid data models, query languages, and query evaluation techniques."

Sessions at the IMAGE workshop cover medical imaging, image handling, image analysis and databases, with the associated issues of registration, metadata, ontologies, provenance, data integration, and clinical/research governance. Workshop attendees are producing a document that identifies opportunities for further research and collaborations. - Paul Tooby