Neuroimaging workflow design and data-mining

With the increasing number of neuroimaging studies appearing yearly in the literature, the need to consider the synthesis of the underlying data into new knowledge and research directions has never been more important. The development of large-scale databases and grid-enabled computing has laid the groundwork for mining these rich datasets beyond the scope of their initial collection. Additionally, meta-analyses of the summary results contained in published research articles have provided a powerful way to explore hidden trends in the neuroscience literature. In each case, the processing of data requires a careful consideration of the individual processing steps involved and how they can be assembled into reliable workflows. In results from published studies, the manner in which data were processed may influence meta-analytic results which can have implications on clinical interpretation. Several efforts now exist that provide tools for use in the construction of data processing workflows. However, careful thought must be given to ensuring appropriate, efficient, optimal, and replicable processing. The results obtained from data-mining and meta-analysis must tell a story about a collection of existing data. Also they must suggest novel and testable hypotheses for further investigation with implications for understanding of the brain in health and disease. Where they do, these new results and interpretations often provide fresh insights into the data that extend beyond the rationale for their original collection. In this volume, we have asked leaders in the field of neuroimaging data mining and meta-analysis to provide their thoughts on methods for efficient workflow design, interoperability with large-scale databases, and to discuss their work in exploring the richness of brain imaging data as well as the literature of published research results.