No abstract

No abstract

No abstract

As the limits of stimuli presentation rates are explored in event-related fMRI design, there is a greater need to assess the implications of averaging raw fMRI data. Selective averaging assumes that the fMRI signal consists of task-dependent signal, random noise, and non-task dependent brain signal that can be modeled as random noise so that it tends to zero when averaged over a practical number of trials. We recorded a total of four fMRI data series from two normal subjects (subject 1, axially acquired; subject 2, coronally acquired) performing a simple visual event-related task and a water phantom with the same fMRI scanner imaging parameters. To determine which fraction of the fMRI data was deterministic as opposed to random, we created different data subsets by taking the odd or even time points of the full data sets. All data sets were first dimension-reduced with principal component analysis (PCA) and separated into 100 spatially independent components with independent component analysis (ICA). The mutual information between best-matching pairs of components selected from full data set Á/subset comparisons was plotted for each data set. Visual inspection suggested that 45 Á/85 components were reproducible, and hence deterministic, accounting for 79 Á/ 97% of the variance, respectively, in the raw data. The reproducible components exhibited much less trial-to-trial variability than the raw data from even the most activated voxel. Many (22 Á/47) of reproducible components were significantly affected by stimulus presentation (P B/0.001). The most significantly-stimulus-correlated component was strongly time-locked to stimulus presentation and was directly stimulus correlated, corresponding to occipital brain regions. However, other spatially distinct task-related components demonstrated variable temporal relationships with the most significantly stimulus-correlated component. Our results suggest that the majority of the variance in fMRI data is in fact deterministic, and support the notion that the data consist of differing components with differing temporal relationships to visual stimulation. They further suggest roles for restricting interpretations of the spatial extent of activation from event-related designs to a specific region of interest (ROI) and/or first separating the data into spatially independent components. Averaging the time courses of spatially independent components timelocked to stimulus presentation may prevent possible biases in the estimates of the spatial and temporal extent of stimulus-correlated activation and of trial-to-trial variability. #

No abstract