The future appears to promise a more integrative approach to functional brain imaging.

Abstract

This brain movie has the interesting property that any single image from it contains no information about brain function. Instead, information about brain function is encoded in the variance of image intensity over time. How then are such data typically analyzed to yield the now-familiar brain activation maps? The conventional approach to fMRI data analysis consists of three steps: preprocessing, regression, and inference [10]. In preprocessing, the raw images are subjected to spatial registration to correct for small head motions, temporally interpolated to compensate for the fact that different slices are acquired at different times, spatially smoothed to enhance signal to noise, and often spatially normalized or transformed into a common stereotactic space to facilitate group analyses and neuroanatomical labeling. In regression, the paradigm timing is convolved with an estimate of the hemodynamic impulse response function. The resultant hemodynamically lagged and blurred version of the paradigm timing is used as a regressor of interest, forming a general linear model, which is then fit to the data, allocating temporal variance in the (preprocessed) data among such regressors. In inference, the spatial map of regression coefficients resulting from the regression step is thresholded for significance, allowing formation of a map showing the suprathreshold regions as hot spots, often overlaid in color on a (higher resolution) anatomical MR image. This standard inferential univariate fMRI data analysis approach can be seen as an excellent machine for testing prior temporal hypotheses. However, it cannot discover unanticipated structure in the data, i.e., it cannot detect brain activations with timings that were unanticipated by the investigator. In the last decade, investigators have developed a variety of data-driven or exploratory techniques that can discover brain activity not anticipated in advance. It may be useful to view such approaches in the context of the finding [11] that in the resting state, synchronous fluctuations (or covariance) of voxel time courses were found throughout the bilateral motor cortex. Recent years have seen significant progress in paradigm design for fMRI as well as the development of other methods for assessing the functional anatomy of the human brain, such as diffusion tensor imaging [12] for mapping white matter fiber tracts. The future appears to promise a more integrative approach to functional brain imaging, in which data from multiple modalities are entered into comprehensive analyses of brain function and connectivity.