The objective of this chapter is to describe the strategy and procedures behind a commercial software product, EvIdent, which is used to extract information from functional magnetic resonance imaging (fMRI) brain scans.

EvIdent is an example of the application of exploratory data analysis techniques to discover features within highly complex data. A three-stage strategy, called exploring regions of interest with cluster analysis (EROICA), is implemented. The software is model-independent. The first stage preprocesses the data into three sets of time courses (TCs): “trend” TCs (for example, instrument drift), “potentially interesting” TCs in the principal partition, and “reject” TCs (usually noise). It is in the first stage that the speed of the data analysis is achieved. In the second stage, the TCs in the principal partition are further tested, and are grouped using fuzzy clustering. The preprocessing allows the fuzzy clustering to work speedily. In the third stage, further statistical tests are applied to the clustered TCs, in which some may be rejected, and others added in from the other two groups.

Sample data were tested with the software, including prepared sets based on real data as well as unmodified real data. The data sets were provided by the original fMRI investigators, who had previously designed the experiments in which the data were collected to show specific features. This information was withheld from the program’s authors. They reported that the program, which runs without a model or previously provided information, was able to extract these features in all cases.