@ARTICLE\{IMM2011-06019,
    author       = "B. Wilkowski and L. K. Hansen",
    title        = "Context-dependent literature search: a support for functional imaging coordinate-based databases",
    year         = "2011",
    journal      = "{BMC} Bioinformatics",
    volume       = "",
    editor       = "",
    number       = "",
    publisher    = "",
    url          = "http://www2.compute.dtu.dk/pubdb/pubs/6019-full.html",
    abstract     = "Background: The growth of published neuro-imaging articles and experimental results in neuroscience, brings a
demand for dedicated information retrieval tools for this specific biomedical domain. There is a need to expand
results pool of the functional imaging coordinate-based databases like SumsDB, BrainMap or Brede, because,
despite their limited size, they are arguably richer for neuroimaging than conventional keyword-based retrieval
services.

Results: In this work we propose a general methodology for linking coordinate-based and keyword-based retrieval
systems. The input to the presented pipeline of methods is any paper retrieved by a coordinate-based service.
First, an automatic extraction of significant keywords is performed by mapping noun phrases to Unified Medical
Language System’s ontological concepts using the Metamap software. Each of the extracted keywords is
automatically classified into one or more semantic groups relevant to neuroscience. The semantic groups and,
associated with them, extracted keywords, are later used for construction of logical queries executed on the
PubMed database. The context of the search is defined by selection of semantic groups of interest. A semantic
group representing brain parts is always stipulated in any search to ensure that the retrieved papers are similar
to the input article by brain part or location.

Conclusions: The discussed methodology may be considered as an extension, in the domain of neuroscience, of
the common publication search engines like PubMed or Google Scholar. We propose integration of the
methodology with the BredeQuery plugin for enabling searches directly from within the Statistical Parametric Mapping, a very popular environment in function neuroimaging. Finally, the flexibility of the approach presented
in this work allows adjustment of the methodology also to other biomedical domains."
}