The purpose of this R01 application is to develop an fMRI acquisition informatics tool for the quantitative evaluation of neuroimaging tools routinely used in the analysis of fMRI data. This tool will also allow an investigator to predict what effect their acquisition parameters have on their fMRI data. The project is motivated by the necessity for a metric for comparing fMRI informatics tools. Typically, most validation experiments rely on the use of an actual fMRI data set acquired under experimental conditions or by overly simple models. The use of actual fMRI data can lead to ambiguous results because the baseline image in the absence of the effect in question is an unknown. A more quantitative comparison would use an fMRI acquisition tool where the right answer is known a priori, where the sources of activation, noise and artifact are well characterized and quantified. Such a tool would lead to a more consistent characterization of fMRI tools by replicating identical data sets between research sites. The metric will be constructed through a digital brain model that accurately replicates the images acquired in fMRI. The model will use high-resolution whole brain images and field inhomogeneity maps. Model options will include susceptibility artifacts, the acquisition type (spiral or EPI), physiological and random noise, motion and activation. The output will be in the form of either images or raw data, stored in any of the major public domain file formats (i.e. Analyze) such that it can be readily inserted into most existing fMRI informatics tools. Classes of metrics will be constructed and validations of the model will be performed using standard tools such as AIR/NIS, AFNI, FIASCO, and SPM. Finally, a web server will be created for downloading the model's basic engine, graphical interface, sample data sets, and instructions.
